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  *      - An icount mechanism is used to keep track of
  24  *        inodes with bad fields and its badness        (ctx->inode_badness)
  25  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
  26  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
  27  *      - A bitmap of which inodes need to be expanded  (expand_eisize_map)
  28  *      - A bitmap of which blocks are in use.          (block_found_map)
  29  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
  30  *      - The data blocks of the directory inodes.      (dir_map)
  31  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
  32  *
  33  * Pass 1 is designed to stash away enough information so that the
  34  * other passes should not need to read in the inode information
  35  * during the normal course of a filesystem check.  (Although if an
  36  * inconsistency is detected, other passes may need to read in an
  37  * inode to fix it.)
  38  *
  39  * Note that pass 1B will be invoked if there are any duplicate blocks
  40  * found.
  41  */
  42
  43 #define _GNU_SOURCE 1 /* get strnlen() */
  44 #include "config.h"
  45 #include <string.h>
  46 #include <time.h>
  47 #ifdef HAVE_ERRNO_H
  48 #include <errno.h>
  49 #endif
  50 #include <assert.h>
  51
  52 #include "e2fsck.h"
  53 #include <ext2fs/ext2_ext_attr.h>
  54 /* todo remove this finally */
  55 #include <ext2fs/ext2fsP.h>
  56 #include <e2p/e2p.h>
  57
  58 #include "problem.h"
  59
  60 #ifdef NO_INLINE_FUNCS
  61 #define _INLINE_
  62 #else
  63 #define _INLINE_ inline
  64 #endif
  65
  66 #undef DEBUG
  67
  68 struct ea_quota {
  69         blk64_t blocks;
  70         __u64 inodes;
  71 };
  72
  73 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  74                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  75                          int ref_offset, void *priv_data);
  76 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  77                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  78                              int ref_offset, void *priv_data);
  79 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  80                          char *block_buf,
  81                          const struct ea_quota *ea_ibody_quota);
  82 static void mark_table_blocks(e2fsck_t ctx);
  83 static void alloc_bb_map(e2fsck_t ctx);
  84 static void alloc_imagic_map(e2fsck_t ctx);
  85 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino);
  86 static void handle_fs_bad_blocks(e2fsck_t ctx);
  87 static void process_inodes(e2fsck_t ctx, char *block_buf);
  88 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  89 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  90                                   dgrp_t group, void * priv_data);
  91 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  92                                     char *block_buf, int adjust_sign);
  93 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  94
  95 struct process_block_struct {
  96         ext2_ino_t      ino;
  97         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
  98                                 fragmented:1, compressed:1, bbcheck:1,
  99                                 inode_modified:1;
 100         blk64_t         num_blocks;
 101         blk64_t         max_blocks;
 102         blk64_t         last_block;
 103         e2_blkcnt_t     last_init_lblock;
 104         e2_blkcnt_t     last_db_block;
 105         int             num_illegal_blocks;
 106         blk64_t         previous_block;
 107         struct ext2_inode *inode;
 108         struct problem_context *pctx;
 109         ext2fs_block_bitmap fs_meta_blocks;
 110         e2fsck_t        ctx;
 111         blk64_t         next_lblock;
 112         struct extent_tree_info eti;
 113 };
 114
 115 struct process_inode_block {
 116         ext2_ino_t ino;
 117         struct ea_quota ea_ibody_quota;
 118         struct ext2_inode_large inode;
 119 };
 120
 121 struct scan_callback_struct {
 122         e2fsck_t        ctx;
 123         char            *block_buf;
 124 };
 125
 126 /*
 127  * For the inodes to process list.
 128  */
 129 static struct process_inode_block *inodes_to_process;
 130 static int process_inode_count;
 131
 132 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 133                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 134
 135 /*
 136  * Free all memory allocated by pass1 in preparation for restarting
 137  * things.
 138  */
 139 static void unwind_pass1(ext2_filsys fs EXT2FS_ATTR((unused)))
 140 {
 141         ext2fs_free_mem(&inodes_to_process);
 142         inodes_to_process = 0;
 143 }
 144
 145 /*
 146  * Check to make sure a device inode is real.  Returns 1 if the device
 147  * checks out, 0 if not.
 148  *
 149  * Note: this routine is now also used to check FIFO's and Sockets,
 150  * since they have the same requirement; the i_block fields should be
 151  * zero.
 152  */
 153 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 154                                     struct ext2_inode *inode)
 155 {
 156         int     i;
 157
 158         /*
 159          * If the index or extents flag is set, then this is a bogus
 160          * device/fifo/socket
 161          */
 162         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 163                 return 0;
 164
 165         /*
 166          * We should be able to do the test below all the time, but
 167          * because the kernel doesn't forcibly clear the device
 168          * inode's additional i_block fields, there are some rare
 169          * occasions when a legitimate device inode will have non-zero
 170          * additional i_block fields.  So for now, we only complain
 171          * when the immutable flag is set, which should never happen
 172          * for devices.  (And that's when the problem is caused, since
 173          * you can't set or clear immutable flags for devices.)  Once
 174          * the kernel has been fixed we can change this...
 175          */
 176         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 177                 for (i=4; i < EXT2_N_BLOCKS; i++)
 178                         if (inode->i_block[i])
 179                                 return 0;
 180         }
 181         return 1;
 182 }
 183
 184 /*
 185  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 186  * checks out, 0 if not.
 187  */
 188 static int check_symlink(e2fsck_t ctx, struct problem_context *pctx,
 189                          ext2_ino_t ino, struct ext2_inode *inode, char *buf)
 190 {
 191         unsigned int buflen;
 192         unsigned int len;
 193         blk64_t blk;
 194
 195         if ((inode->i_size_high || inode->i_size == 0) ||
 196             (inode->i_flags & EXT2_INDEX_FL))
 197                 return 0;
 198
 199         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 200                 size_t inline_size;
 201
 202                 if (inode->i_flags & EXT4_EXTENTS_FL)
 203                         return 0;
 204                 if (ext2fs_inline_data_size(ctx->fs, ino, &inline_size))
 205                         return 0;
 206                 if (inode->i_size != inline_size)
 207                         return 0;
 208
 209                 return 1;
 210         }
 211
 212         if (ext2fs_is_fast_symlink(inode)) {
 213                 if (inode->i_flags & EXT4_EXTENTS_FL)
 214                         return 0;
 215                 buf = (char *)inode->i_block;
 216                 buflen = sizeof(inode->i_block);
 217         } else {
 218                 ext2_extent_handle_t    handle;
 219                 struct ext2_extent_info info;
 220                 struct ext2fs_extent    extent;
 221                 int i;
 222
 223                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 224                         if (ext2fs_extent_open2(ctx->fs, ino, inode, &handle))
 225                                 return 0;
 226                         if (ext2fs_extent_get_info(handle, &info) ||
 227                             (info.num_entries != 1) ||
 228                             (info.max_depth != 0)) {
 229                                 ext2fs_extent_free(handle);
 230                                 return 0;
 231                         }
 232                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 233                                               &extent) ||
 234                             (extent.e_lblk != 0) ||
 235                             (extent.e_len != 1)) {
 236                                 ext2fs_extent_free(handle);
 237                                 return 0;
 238                         }
 239                         blk = extent.e_pblk;
 240                         ext2fs_extent_free(handle);
 241                 } else {
 242                         blk = inode->i_block[0];
 243
 244                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 245                                 if (inode->i_block[i])
 246                                         return 0;
 247                 }
 248
 249                 if (blk < ctx->fs->super->s_first_data_block ||
 250                     blk >= ext2fs_blocks_count(ctx->fs->super))
 251                         return 0;
 252
 253                 if (io_channel_read_blk64(ctx->fs->io, blk, 1, buf))
 254                         return 0;
 255
 256                 buflen = ctx->fs->blocksize;
 257         }
 258
 259         if (inode->i_flags & EXT4_ENCRYPT_FL)
 260                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 261         else {
 262                 len = strnlen(buf, buflen);
 263
 264                 /* Add missing NUL terminator at end of symlink (LU-1540),
 265                  * but only offer to fix this in pass1, not from pass2. */
 266                 if (len > inode->i_size && pctx != NULL &&
 267                     fix_problem(ctx, PR_1_SYMLINK_NUL, pctx)) {
 268                         buf[inode->i_size] = '\0';
 269                         if (ext2fs_is_fast_symlink(inode)) {
 270                                 e2fsck_write_inode(ctx, ino,
 271                                                    inode, "check_ext_attr");
 272                         } else {
 273                                 if (io_channel_write_blk64(ctx->fs->io,
 274                                                            blk, 1, buf))
 275                                         return 0;
 276                         }
 277                         len = inode->i_size;
 278                 }
 279         }
 280
 281         if (len >= buflen)
 282                 return 0;
 283
 284         if (len != inode->i_size)
 285                 return 0;
 286
 287         return 1;
 288 }
 289
 290 int e2fsck_pass1_check_symlink(e2fsck_t ctx, ext2_ino_t ino,
 291                                struct ext2_inode *inode, char *buf)
 292 {
 293         return check_symlink(ctx, NULL, ino, inode, buf);
 294 }
 295
 296 /*
 297  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 298  */
 299 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 300 static void check_extents_inlinedata(e2fsck_t ctx,
 301                                      struct problem_context *pctx)
 302 {
 303         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 304                 return;
 305
 306         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 307                 return;
 308
 309         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 310         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 311 }
 312 #undef BAD_SPECIAL_FLAGS
 313
 314 /*
 315  * If the immutable (or append-only) flag is set on the inode, offer
 316  * to clear it.
 317  */
 318 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 319 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 320 {
 321         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 322                 return;
 323
 324         e2fsck_mark_inode_bad(ctx, pctx->ino, BADNESS_NORMAL);
 325         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 326                 return;
 327
 328         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 329         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 330 }
 331
 332 /*
 333  * If device, fifo or socket, check size is zero -- if not offer to
 334  * clear it
 335  */
 336 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 337 {
 338         struct ext2_inode *inode = pctx->inode;
 339
 340         if (EXT2_I_SIZE(inode) == 0)
 341                 return;
 342
 343         e2fsck_mark_inode_bad(ctx, pctx->ino, BADNESS_NORMAL);
 344         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 345                 return;
 346
 347         ext2fs_inode_size_set(ctx->fs, inode, 0);
 348         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 349 }
 350
 351 /*
 352  * For a given size, calculate how many blocks would be charged towards quota.
 353  */
 354 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 355 {
 356         blk64_t clusters;
 357
 358         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 359         return EXT2FS_C2B(fs, clusters);
 360 }
 361
 362 /*
 363  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 364  * the problem code.
 365  */
 366 static problem_t check_large_ea_inode(e2fsck_t ctx,
 367                                       struct ext2_ext_attr_entry *entry,
 368                                       struct problem_context *pctx,
 369                                       blk64_t *quota_blocks)
 370 {
 371         struct ext2_inode inode;
 372         __u32 hash;
 373         errcode_t retval;
 374
 375         /* Check if inode is within valid range */
 376         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 377             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 378                 pctx->num = entry->e_value_inum;
 379                 return PR_1_ATTR_VALUE_EA_INODE;
 380         }
 381
 382         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 383
 384         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
 385         if (retval) {
 386                 com_err("check_large_ea_inode", retval,
 387                         _("while hashing entry with e_value_inum = %u"),
 388                         entry->e_value_inum);
 389                 fatal_error(ctx, 0);
 390         }
 391
 392         if (hash == entry->e_hash) {
 393                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 394                                                      entry->e_value_size);
 395         } else {
 396                 /* This might be an old Lustre-style ea_inode reference. */
 397                 if (inode.i_mtime == pctx->ino &&
 398                     inode.i_generation == pctx->inode->i_generation) {
 399                         *quota_blocks = 0;
 400                 } else {
 401                         /* If target inode is also missing EA_INODE flag,
 402                          * this is likely to be a bad reference.
 403                          */
 404                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 405                                 pctx->num = entry->e_value_inum;
 406                                 return PR_1_ATTR_VALUE_EA_INODE;
 407                         } else {
 408                                 pctx->num = entry->e_hash;
 409                                 return PR_1_ATTR_HASH;
 410                         }
 411                 }
 412         }
 413
 414         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 415                 pctx->num = entry->e_value_inum;
 416                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 417                         inode.i_flags |= EXT4_EA_INODE_FL;
 418                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 419                                            &inode);
 420                 } else {
 421                         return PR_1_ATTR_NO_EA_INODE_FL;
 422                 }
 423         }
 424         return 0;
 425 }
 426
 427 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 428                               struct ext2_ext_attr_entry *first, void *end)
 429 {
 430         struct ext2_ext_attr_entry *entry;
 431
 432         for (entry = first;
 433              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
 434              entry = EXT2_EXT_ATTR_NEXT(entry)) {
 435                 if (!entry->e_value_inum)
 436                         continue;
 437                 if (!ctx->ea_inode_refs) {
 438                         pctx->errcode = ea_refcount_create(0,
 439                                                            &ctx->ea_inode_refs);
 440                         if (pctx->errcode) {
 441                                 pctx->num = 4;
 442                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 443                                 ctx->flags |= E2F_FLAG_ABORT;
 444                                 return;
 445                         }
 446                 }
 447                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 448                                       0);
 449         }
 450 }
 451
 452 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 453                               struct ea_quota *ea_ibody_quota)
 454 {
 455         struct ext2_super_block *sb = ctx->fs->super;
 456         struct ext2_inode_large *inode;
 457         struct ext2_ext_attr_entry *entry;
 458         char *start, *header, *end;
 459         unsigned int storage_size, remain;
 460         problem_t problem = 0;
 461         region_t region = 0;
 462
 463         ea_ibody_quota->blocks = 0;
 464         ea_ibody_quota->inodes = 0;
 465
 466         inode = (struct ext2_inode_large *) pctx->inode;
 467         storage_size = EXT2_INODE_SIZE(ctx->fs->super) -
 468                 EXT2_GOOD_OLD_INODE_SIZE - inode->i_extra_isize;
 469         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 470                  inode->i_extra_isize;
 471         end = header + storage_size;
 472         entry = &IHDR(inode)->h_first_entry[0];
 473         start = (char *)entry;
 474
 475         /* scan all entry's headers first */
 476
 477         /* take finish entry 0UL into account */
 478         remain = storage_size - sizeof(__u32);
 479
 480         region = region_create(0, storage_size);
 481         if (!region) {
 482                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 483                 problem = 0;
 484                 ctx->flags |= E2F_FLAG_ABORT;
 485                 return;
 486         }
 487         if (region_allocate(region, 0, sizeof(__u32))) {
 488                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 489                 goto fix;
 490         }
 491
 492         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 493                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 494                 __u32 hash;
 495
 496                 if (region_allocate(region, (char *)entry - (char *)header,
 497                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 498                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 499                         goto fix;
 500                 }
 501
 502                 /* header eats this space */
 503                 remain -= sizeof(struct ext2_ext_attr_entry);
 504
 505                 /* is attribute name valid? */
 506                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 507                         pctx->num = entry->e_name_len;
 508                         problem = PR_1_ATTR_NAME_LEN;
 509                         goto fix;
 510                 }
 511
 512                 /* attribute len eats this space */
 513                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 514
 515                 if (entry->e_value_inum == 0) {
 516                         /* check value size */
 517                         if (entry->e_value_size > remain) {
 518                                 pctx->num = entry->e_value_size;
 519                                 problem = PR_1_ATTR_VALUE_SIZE;
 520                                 goto fix;
 521                         }
 522
 523                         if (entry->e_value_size &&
 524                             region_allocate(region,
 525                                             sizeof(__u32) + entry->e_value_offs,
 526                                             EXT2_EXT_ATTR_SIZE(
 527                                                 entry->e_value_size))) {
 528                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 529                                 goto fix;
 530                         }
 531
 532                         hash = ext2fs_ext_attr_hash_entry(entry,
 533                                                           start + entry->e_value_offs);
 534
 535                         /* e_hash may be 0 in older inode's ea */
 536                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 537                                 pctx->num = entry->e_hash;
 538                                 problem = PR_1_ATTR_HASH;
 539                                 goto fix;
 540                         }
 541                 } else {
 542                         blk64_t quota_blocks;
 543
 544                         problem = check_large_ea_inode(ctx, entry, pctx,
 545                                                        &quota_blocks);
 546                         if (problem != 0)
 547                                 goto fix;
 548
 549                         ea_ibody_quota->blocks += quota_blocks;
 550                         ea_ibody_quota->inodes++;
 551                 }
 552
 553                 /* If EA value is stored in external inode then it does not
 554                  * consume space here */
 555                 if (entry->e_value_inum == 0)
 556                         remain -= entry->e_value_size;
 557
 558                 entry = EXT2_EXT_ATTR_NEXT(entry);
 559         }
 560
 561         if (region_allocate(region, (char *)entry - (char *)header,
 562                             sizeof(__u32))) {
 563                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 564                 goto fix;
 565         }
 566 fix:
 567         if (region)
 568                 region_free(region);
 569         /*
 570          * it seems like a corruption. it's very unlikely we could repair
 571          * EA(s) in automatic fashion -bzzz
 572          */
 573         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 574                 inc_ea_inode_refs(ctx, pctx,
 575                                   (struct ext2_ext_attr_entry *)start, end);
 576                 return;
 577         }
 578
 579         /* simply remove all possible EA(s) */
 580         *((__u32 *)header) = 0UL;
 581         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 582                                 EXT2_INODE_SIZE(sb), "pass1");
 583         ea_ibody_quota->blocks = 0;
 584         ea_ibody_quota->inodes = 0;
 585 }
 586
 587 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 588         return (xtime & (1U << 31)) != 0 &&
 589                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 590 }
 591
 592 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 593         check_inode_extra_negative_epoch(inode->i_##xtime, \
 594                                          inode->i_##xtime##_extra)
 595
 596 /* When today's date is earlier than 2242, we assume that atimes,
 597  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 598  * actually pre-1970 dates mis-encoded.
 599  */
 600 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 601
 602 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 603                                     struct ea_quota *ea_ibody_quota)
 604 {
 605         struct ext2_super_block *sb = ctx->fs->super;
 606         struct ext2_inode_large *inode;
 607         __u32 *eamagic;
 608         int min, max, dirty = 0;
 609
 610         ea_ibody_quota->blocks = 0;
 611         ea_ibody_quota->inodes = 0;
 612
 613         inode = (struct ext2_inode_large *) pctx->inode;
 614         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 615                 /* this isn't large inode. so, nothing to check */
 616                 return;
 617         }
 618
 619 #if 0
 620         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 621                         inode->i_extra_isize);
 622 #endif
 623         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 624         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 625         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 626         /*
 627          * For now we will allow i_extra_isize to be 0, but really
 628          * implementations should never allow i_extra_isize to be 0
 629          */
 630         if (inode->i_extra_isize &&
 631             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 632              inode->i_extra_isize & 3)) {
 633                 e2fsck_mark_inode_bad(ctx, pctx->ino, BADNESS_NORMAL);
 634                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 635                         return;
 636                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 637                         inode->i_extra_isize = ctx->want_extra_isize;
 638                 else
 639                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 640                 dirty = 1;
 641
 642                 goto out;
 643         }
 644
 645         /* check if there is no place for an EA header */
 646         if (inode->i_extra_isize >= max - sizeof(__u32))
 647                 return;
 648
 649         eamagic = &IHDR(inode)->h_magic;
 650         if (*eamagic != EXT2_EXT_ATTR_MAGIC &&
 651             (ctx->flags & E2F_FLAG_EXPAND_EISIZE) &&
 652             (inode->i_extra_isize < ctx->want_extra_isize)) {
 653                 fix_problem(ctx, PR_1_EXPAND_EISIZE, pctx);
 654                 memset((char *)inode + EXT2_GOOD_OLD_INODE_SIZE, 0,
 655                         EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE);
 656                 inode->i_extra_isize = ctx->want_extra_isize;
 657                 dirty = 1;
 658                 if (inode->i_extra_isize < ctx->min_extra_isize)
 659                         ctx->min_extra_isize = inode->i_extra_isize;
 660         }
 661
 662         if (*eamagic == EXT2_EXT_ATTR_MAGIC)
 663                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 664
 665         /*
 666          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 667          * the old, invalid format, repair it.
 668          */
 669         if (((sizeof(time_t) <= 4) ||
 670              (((sizeof(time_t) > 4) &&
 671                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 672             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 673              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 674              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 675              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 676
 677                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 678                         return;
 679
 680                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 681                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 682                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 683                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 684                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 685                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 686                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 687                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 688                 dirty = 1;
 689                 e2fsck_mark_inode_bad(ctx, pctx->ino, BADNESS_HIGH);
 690         }
 691
 692 out:
 693         if (dirty)
 694                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 695                                         EXT2_INODE_SIZE(sb), "pass1");
 696 }
 697
 698 /*
 699  * Check to see if the inode might really be a directory, despite i_mode
 700  *
 701  * This is a lot of complexity for something for which I'm not really
 702  * convinced happens frequently in the wild.  If for any reason this
 703  * causes any problems, take this code out.
 704  * [tytso:20070331.0827EDT]
 705  */
 706 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 707                                 char *buf)
 708 {
 709         struct ext2_inode *inode = pctx->inode;
 710         struct ext2_dir_entry   *dirent;
 711         errcode_t               retval;
 712         blk64_t                 blk;
 713         unsigned int            i, rec_len, not_device = 0;
 714         int                     extent_fs;
 715         int                     inlinedata_fs;
 716
 717         /*
 718          * If the mode looks OK, we believe it.  If the first block in
 719          * the i_block array is 0, this cannot be a directory. If the
 720          * inode is extent-mapped, it is still the case that the latter
 721          * cannot be 0 - the magic number in the extent header would make
 722          * it nonzero.
 723          */
 724         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 725             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 726                 return;
 727
 728         /*
 729          * Check the block numbers in the i_block array for validity:
 730          * zero blocks are skipped (but the first one cannot be zero -
 731          * see above), other blocks are checked against the first and
 732          * max data blocks (from the the superblock) and against the
 733          * block bitmap. Any invalid block found means this cannot be
 734          * a directory.
 735          *
 736          * If there are non-zero blocks past the fourth entry, then
 737          * this cannot be a device file: we remember that for the next
 738          * check.
 739          *
 740          * For extent mapped files, we don't do any sanity checking:
 741          * just try to get the phys block of logical block 0 and run
 742          * with it.
 743          *
 744          * For inline data files, we just try to get the size of inline
 745          * data.  If it's true, we will treat it as a directory.
 746          */
 747
 748         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 749         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 750         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 751                 size_t size;
 752                 __u32 dotdot;
 753                 unsigned int rec_len2;
 754                 struct ext2_dir_entry de;
 755
 756                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 757                         return;
 758                 /*
 759                  * If the size isn't a multiple of 4, it's probably not a
 760                  * directory??
 761                  */
 762                 if (size & 3)
 763                         return;
 764                 /*
 765                  * If the first 10 bytes don't look like a directory entry,
 766                  * it's probably not a directory.
 767                  */
 768                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 769                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 770                        EXT2_DIR_NAME_LEN(0));
 771                 dotdot = ext2fs_le32_to_cpu(dotdot);
 772                 de.inode = ext2fs_le32_to_cpu(de.inode);
 773                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 774                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 775                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 776                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 777                      dotdot != EXT2_ROOT_INO) ||
 778                     de.inode >= ctx->fs->super->s_inodes_count ||
 779                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 780                      de.inode != 0) ||
 781                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 782                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 783                         return;
 784                 /* device files never have a "system.data" entry */
 785                 goto isdir;
 786         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 787                 /* extent mapped */
 788                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 789                                  &blk))
 790                         return;
 791                 /* device files are never extent mapped */
 792                 not_device++;
 793         } else {
 794                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 795                         blk = inode->i_block[i];
 796                         if (!blk)
 797                                 continue;
 798                         if (i >= 4)
 799                                 not_device++;
 800
 801                         if (blk < ctx->fs->super->s_first_data_block ||
 802                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 803                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 804                                                            blk))
 805                                 return; /* Invalid block, can't be dir */
 806                 }
 807                 blk = inode->i_block[0];
 808         }
 809
 810         /*
 811          * If the mode says this is a device file and the i_links_count field
 812          * is sane and we have not ruled it out as a device file previously,
 813          * we declare it a device file, not a directory.
 814          */
 815         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 816             (inode->i_links_count == 1) && !not_device)
 817                 return;
 818
 819         /* read the first block */
 820         ehandler_operation(_("reading directory block"));
 821         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 822         ehandler_operation(0);
 823         if (retval)
 824                 return;
 825
 826         dirent = (struct ext2_dir_entry *) buf;
 827         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 828         if (retval)
 829                 return;
 830         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 831             (dirent->name[0] != '.') ||
 832             (dirent->inode != pctx->ino) ||
 833             (rec_len < 12) ||
 834             (rec_len % 4) ||
 835             (rec_len >= ctx->fs->blocksize - 12))
 836                 return;
 837
 838         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 839         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 840         if (retval)
 841                 return;
 842         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 843             (dirent->name[0] != '.') ||
 844             (dirent->name[1] != '.') ||
 845             (rec_len < 12) ||
 846             (rec_len % 4))
 847                 return;
 848
 849 isdir:
 850         e2fsck_mark_inode_bad(ctx, pctx->ino, BADNESS_NORMAL);
 851         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 852                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 853                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 854                                         EXT2_INODE_SIZE(ctx->fs->super),
 855                                         "check_is_really_dir");
 856         }
 857 }
 858
 859 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 860                                      int flags, ext2_icount_t hint,
 861                                      ext2_icount_t *ret)
 862 {
 863         unsigned int            threshold;
 864         unsigned int            save_type;
 865         ext2_ino_t              num_dirs;
 866         errcode_t               retval;
 867         char                    *tdb_dir;
 868         int                     enable;
 869
 870         *ret = 0;
 871
 872         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 873                            &tdb_dir);
 874         profile_get_uint(ctx->profile, "scratch_files",
 875                          "numdirs_threshold", 0, 0, &threshold);
 876         profile_get_boolean(ctx->profile, "scratch_files",
 877                             "icount", 0, 1, &enable);
 878
 879         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 880         if (retval)
 881                 num_dirs = 1024;        /* Guess */
 882
 883         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 884             (!threshold || num_dirs > threshold)) {
 885                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 886                                                   flags, ret);
 887                 if (retval == 0)
 888                         return 0;
 889         }
 890         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 891                                &save_type);
 892         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 893                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 894         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 895         ctx->fs->default_bitmap_type = save_type;
 896         return retval;
 897 }
 898
 899 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 900                                             e2fsck_t ctx,
 901                                             struct problem_context *pctx)
 902 {
 903         errcode_t retval;
 904         struct ext2_inode_large inode;
 905
 906         /*
 907          * Reread inode.  If we don't see checksum error, then this inode
 908          * has been fixed elsewhere.
 909          */
 910         ctx->stashed_ino = 0;
 911         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 912                                         sizeof(inode));
 913         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 914                 return retval;
 915         if (!retval)
 916                 return 0;
 917
 918         /*
 919          * Checksum still doesn't match.  That implies that the inode passes
 920          * all the sanity checks, so maybe the checksum is simply corrupt.
 921          * See if the user will go for fixing that.
 922          */
 923         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 924                 return 0;
 925
 926         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 927                                          sizeof(inode));
 928         return retval;
 929 }
 930
 931 int e2fsck_pass1_delete_attr(e2fsck_t ctx, struct ext2_inode_large *inode,
 932                              struct problem_context *pctx, int needed_size)
 933 {
 934         struct ext2_ext_attr_header *header;
 935         struct ext2_ext_attr_entry *entry_ino, *entry_blk = NULL, *entry;
 936         char *start, name[4096], block_buf[4096];
 937         int len, index = EXT2_ATTR_INDEX_USER, entry_size, ea_size;
 938         int in_inode = 1, error;
 939         unsigned int freed_bytes = inode->i_extra_isize;
 940
 941         entry_ino = &IHDR(inode)->h_first_entry[0];
 942         start = (char *)entry_ino;
 943
 944         if (inode->i_file_acl) {
 945                 error = ext2fs_read_ext_attr(ctx->fs, inode->i_file_acl,
 946                                              block_buf);
 947                 /* We have already checked this block, shouldn't happen */
 948                 if (error) {
 949                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, pctx);
 950                         return 0;
 951                 }
 952                 header = BHDR(block_buf);
 953                 if (header->h_magic != EXT2_EXT_ATTR_MAGIC) {
 954                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, pctx);
 955                         return 0;
 956                 }
 957
 958                 entry_blk = (struct ext2_ext_attr_entry *)(header+1);
 959         }
 960         entry = entry_ino;
 961         len = sizeof(entry->e_name);
 962         entry_size = ext2fs_attr_get_next_attr(entry, index, name, len, 1);
 963
 964         while (freed_bytes < needed_size) {
 965                 if (entry_size && name[0] != '\0') {
 966                         pctx->str = name;
 967                         if (fix_problem(ctx, PR_1_EISIZE_DELETE_EA, pctx)) {
 968                                 ea_size = EXT2_EXT_ATTR_LEN(entry->e_name_len) +
 969                                           EXT2_EXT_ATTR_SIZE(entry->e_value_size);
 970                                 error = ext2fs_attr_set(ctx->fs, pctx->ino,
 971                                                         (struct ext2_inode *)inode,
 972                                                         index, name, 0, 0, 0);
 973                                 if (!error)
 974                                         freed_bytes += ea_size;
 975                         }
 976                 }
 977                 len = sizeof(entry->e_name);
 978                 entry_size = ext2fs_attr_get_next_attr(entry, index,name,len,0);
 979                 entry = EXT2_EXT_ATTR_NEXT(entry);
 980                 if (EXT2_EXT_IS_LAST_ENTRY(entry)) {
 981                         if (in_inode) {
 982                                 entry = entry_blk;
 983                                 len = sizeof(entry->e_name);
 984                                 entry_size = ext2fs_attr_get_next_attr(entry,
 985                                                         index, name, len, 1);
 986                                 in_inode = 0;
 987                         } else {
 988                                 index += 1;
 989                                 in_inode = 1;
 990                                 if (!entry && index < EXT2_ATTR_INDEX_MAX)
 991                                         entry = (struct ext2_ext_attr_entry *)start;
 992                                 else
 993                                         return freed_bytes;
 994                         }
 995                 }
 996         }
 997
 998         return freed_bytes;
 999 }
1000
1001 int e2fsck_pass1_expand_eisize(e2fsck_t ctx, struct ext2_inode_large *inode,
1002                                struct problem_context *pctx)
1003 {
1004         int needed_size = 0, retval, ret = EXT2_EXPAND_EISIZE_UNSAFE;
1005         static int message;
1006
1007 retry:
1008         retval = ext2fs_expand_extra_isize(ctx->fs, pctx->ino, inode,
1009                                            ctx->want_extra_isize, &ret,
1010                                            &needed_size);
1011         if (ret & EXT2_EXPAND_EISIZE_NEW_BLOCK)
1012                 goto mark_expand_eisize_map;
1013         if (!retval) {
1014                 e2fsck_write_inode_full(ctx, pctx->ino,
1015                                         (struct ext2_inode *)inode,
1016                                         EXT2_INODE_SIZE(ctx->fs->super),
1017                                         "pass1");
1018                 return 0;
1019         }
1020
1021         if (ret & EXT2_EXPAND_EISIZE_NOSPC) {
1022                 if (ctx->options & (E2F_OPT_PREEN | E2F_OPT_YES)) {
1023                         fix_problem(ctx, PR_1_EA_BLK_NOSPC, pctx);
1024                         ctx->flags |= E2F_FLAG_ABORT;
1025                         return -1;
1026                 }
1027
1028                 if (!message) {
1029                         pctx->num = ctx->fs->super->s_min_extra_isize;
1030                         fix_problem(ctx, PR_1_EXPAND_EISIZE_WARNING, pctx);
1031                         message = 1;
1032                 }
1033 delete_EA:
1034                 retval = e2fsck_pass1_delete_attr(ctx, inode, pctx,
1035                                                   needed_size);
1036                 if (retval >= ctx->want_extra_isize)
1037                         goto retry;
1038
1039                 needed_size -= retval;
1040
1041                 /*
1042                  * We loop here until either the user deletes EA(s) or
1043                  * EXTRA_ISIZE feature is disabled.
1044                  */
1045                 if (fix_problem(ctx, PR_1_CLEAR_EXTRA_ISIZE, pctx)) {
1046                         ctx->fs->super->s_feature_ro_compat &=
1047                                         ~EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE;
1048                         ext2fs_mark_super_dirty(ctx->fs);
1049                 } else {
1050                         goto delete_EA;
1051                 }
1052                 ctx->fs_unexpanded_inodes++;
1053
1054                 /* No EA was deleted, inode cannot be expanded */
1055                 return -1;
1056         }
1057
1058 mark_expand_eisize_map:
1059         if (!ctx->expand_eisize_map) {
1060                 pctx->errcode = ext2fs_allocate_inode_bitmap(ctx->fs,
1061                                          _("expand extrz isize map"),
1062                                          &ctx->expand_eisize_map);
1063                 if (pctx->errcode) {
1064                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR,
1065                                     pctx);
1066                         exit(1);
1067                 }
1068         }
1069
1070         /* Add this inode to the expand_eisize_map */
1071         ext2fs_mark_inode_bitmap2(ctx->expand_eisize_map, pctx->ino);
1072         return 0;
1073 }
1074
1075 static void reserve_block_for_root_repair(e2fsck_t ctx)
1076 {
1077         blk64_t         blk = 0;
1078         errcode_t       err;
1079         ext2_filsys     fs = ctx->fs;
1080
1081         ctx->root_repair_block = 0;
1082         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
1083                 return;
1084
1085         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
1086         if (err)
1087                 return;
1088         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
1089         ctx->root_repair_block = blk;
1090 }
1091
1092 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
1093 {
1094         blk64_t         blk = 0;
1095         errcode_t       err;
1096         ext2_filsys     fs = ctx->fs;
1097         static const char name[] = "lost+found";
1098         ext2_ino_t      ino;
1099
1100         ctx->lnf_repair_block = 0;
1101         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
1102                 return;
1103
1104         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
1105         if (err)
1106                 return;
1107         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
1108         ctx->lnf_repair_block = blk;
1109 }
1110
1111 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
1112                                          size_t *sz)
1113 {
1114         void *p;
1115         struct ext2_xattr_handle *handle;
1116         errcode_t retval;
1117
1118         retval = ext2fs_xattrs_open(fs, ino, &handle);
1119         if (retval)
1120                 return retval;
1121
1122         retval = ext2fs_xattrs_read(handle);
1123         if (retval)
1124                 goto err;
1125
1126         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
1127         if (retval)
1128                 goto err;
1129         ext2fs_free_mem(&p);
1130 err:
1131         (void) ext2fs_xattrs_close(&handle);
1132         return retval;
1133 }
1134
1135 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
1136                                     struct problem_context *pctx,
1137                                     int failed_csum)
1138 {
1139         if (!failed_csum)
1140                 return;
1141
1142         /*
1143          * If the inode failed the checksum and the user didn't
1144          * clear the inode, test the checksum again -- if it still
1145          * fails, ask the user if the checksum should be corrected.
1146          */
1147         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
1148         if (pctx->errcode)
1149                 ctx->flags |= E2F_FLAG_ABORT;
1150 }
1151 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
1152         do { \
1153                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
1154                 if ((ctx)->flags & E2F_FLAG_ABORT) \
1155                         return; \
1156         } while (0)
1157
1158 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
1159 {
1160         __u32 x;
1161         int i;
1162
1163         for (i = 0; i < EXT2_N_BLOCKS; i++) {
1164                 x = inode->i_block[i];
1165 #ifdef WORDS_BIGENDIAN
1166                 x = ext2fs_swab32(x);
1167 #endif
1168                 if (x >= ext2fs_blocks_count(fs->super))
1169                         return 0;
1170         }
1171
1172         return 1;
1173 }
1174
1175 /*
1176  * Figure out what to do with an inode that has both extents and inline data
1177  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
1178  */
1179 static int fix_inline_data_extents_file(e2fsck_t ctx,
1180                                         ext2_ino_t ino,
1181                                         struct ext2_inode *inode,
1182                                         int inode_size,
1183                                         struct problem_context *pctx)
1184 {
1185         size_t max_inline_ea_size;
1186         ext2_filsys fs = ctx->fs;
1187         int dirty = 0;
1188
1189         /* Both feature flags not set?  Just run the regular checks */
1190         if (!ext2fs_has_feature_extents(fs->super) &&
1191             !ext2fs_has_feature_inline_data(fs->super))
1192                 return 0;
1193
1194         /* Clear both flags if it's a special file */
1195         if (LINUX_S_ISCHR(inode->i_mode) ||
1196             LINUX_S_ISBLK(inode->i_mode) ||
1197             LINUX_S_ISFIFO(inode->i_mode) ||
1198             LINUX_S_ISSOCK(inode->i_mode)) {
1199                 check_extents_inlinedata(ctx, pctx);
1200                 return 0;
1201         }
1202
1203         /* If it looks like an extent tree, try to clear inlinedata */
1204         if (ext2fs_extent_header_verify(inode->i_block,
1205                                  sizeof(inode->i_block)) == 0 &&
1206             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1207                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1208                 dirty = 1;
1209                 goto out;
1210         }
1211
1212         /* If it looks short enough to be inline data, try to clear extents */
1213         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1214                 max_inline_ea_size = inode_size -
1215                                      (EXT2_GOOD_OLD_INODE_SIZE +
1216                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1217         else
1218                 max_inline_ea_size = 0;
1219         if (EXT2_I_SIZE(inode) <
1220             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1221             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1222                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1223                 dirty = 1;
1224                 goto out;
1225         }
1226
1227         /*
1228          * Too big for inline data, but no evidence of extent tree -
1229          * maybe it's a block map file?  If the mappings all look valid?
1230          */
1231         if (could_be_block_map(fs, inode) &&
1232             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1233 #ifdef WORDS_BIGENDIAN
1234                 int i;
1235
1236                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1237                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1238 #endif
1239
1240                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1241                 dirty = 1;
1242                 goto out;
1243         }
1244
1245         /* Oh well, just clear the busted inode. */
1246         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1247                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1248                 return -1;
1249         }
1250
1251 out:
1252         if (dirty)
1253                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1254
1255         return 0;
1256 }
1257
1258 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1259 {
1260         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1261         dgrp_t start = *group, grp;
1262         blk64_t blocks_to_read = 0;
1263         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1264
1265         if (ctx->readahead_kb == 0)
1266                 goto out;
1267
1268         /* Keep iterating groups until we have enough to readahead */
1269         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1270         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1271                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1272                         continue;
1273                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1274                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1275                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1276                                         inodes_per_block;
1277                 if (blocks_to_read * ctx->fs->blocksize >
1278                     ctx->readahead_kb * 1024)
1279                         break;
1280         }
1281
1282         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1283                                grp - start + 1);
1284         if (err == EAGAIN) {
1285                 ctx->readahead_kb /= 2;
1286                 err = 0;
1287         }
1288
1289 out:
1290         if (err) {
1291                 /* Error; disable itable readahead */
1292                 *group = ctx->fs->group_desc_count;
1293                 *next_ino = ctx->fs->super->s_inodes_count;
1294         } else {
1295                 /*
1296                  * Don't do more readahead until we've reached the first inode
1297                  * of the last inode scan buffer block for the last group.
1298                  */
1299                 *group = grp + 1;
1300                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1301                                      ctx->inode_buffer_blocks :
1302                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1303                                     ctx->fs->blocksize /
1304                                     EXT2_INODE_SIZE(ctx->fs->super);
1305                 inodes_in_group--;
1306                 *next_ino = inodes_in_group -
1307                             (inodes_in_group % inodes_per_buffer) + 1 +
1308                             (grp * ctx->fs->super->s_inodes_per_group);
1309         }
1310 }
1311
1312 /*
1313  * Check if the passed ino is one of the used superblock quota inodes.
1314  *
1315  * Before the quota inodes were journaled, older superblock quota inodes
1316  * were just regular files in the filesystem and not reserved inodes.  This
1317  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1318  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1319  */
1320 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1321 {
1322         enum quota_type qtype;
1323
1324         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1325                 if (*quota_sb_inump(sb, qtype) == ino)
1326                         return 1;
1327
1328         return 0;
1329 }
1330
1331 /*
1332  * Check if the passed ino is one of the reserved quota inodes.
1333  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1334  * inodes.  These inodes may or may not be in use by the quota feature.
1335  */
1336 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1337 {
1338         enum quota_type qtype;
1339
1340         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1341                 if (quota_type2inum(qtype, fs->super) == ino)
1342                         return 1;
1343
1344         return 0;
1345 }
1346
1347 /*
1348  * Lustre FS creates special inodes - precreated objects.
1349  * They are zero-sized and have special attributes:
1350  * mode |= S_ISUID | S_ISGID;
1351  * valid |= LA_ATIME | LA_MTIME | LA_CTIME;
1352  * atime = 0;
1353  * mtime = 0;
1354  * ctime = 0;
1355  */
1356 static int precreated_object(struct ext2_inode *inode)
1357 {
1358         if (((inode->i_mode & (S_ISUID | S_ISGID)) == (S_ISUID | S_ISGID)) &&
1359              inode->i_ctime == 0)
1360                 return 1;
1361         return 0;
1362 }
1363
1364 static int e2fsck_should_abort(e2fsck_t ctx)
1365 {
1366         e2fsck_t global_ctx;
1367
1368         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1369                 return 1;
1370
1371         if (ctx->global_ctx) {
1372                 global_ctx = ctx->global_ctx;
1373                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1374                         return 1;
1375         }
1376         return 0;
1377 }
1378
1379 void e2fsck_pass1_thread(e2fsck_t ctx)
1380 {
1381         int     i;
1382         __u64   max_sizes;
1383         ext2_filsys fs = ctx->fs;
1384         ext2_ino_t      ino = 0;
1385         struct ext2_inode *inode = NULL;
1386         ext2_inode_scan scan = NULL;
1387         char            *block_buf = NULL;
1388 #ifdef RESOURCE_TRACK
1389         struct resource_track   rtrack;
1390 #endif
1391         unsigned char   frag, fsize;
1392         struct          problem_context pctx;
1393         struct          scan_callback_struct scan_struct;
1394         struct ext2_super_block *sb = ctx->fs->super;
1395         const char      *old_op;
1396         const char      *eop_next_inode = _("getting next inode from scan");
1397         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1398         int             low_dtime_check = 1;
1399         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1400         unsigned int    bufsize;
1401         int             failed_csum = 0;
1402         ext2_ino_t      ino_threshold = 0;
1403         dgrp_t          ra_group = 0;
1404         struct ea_quota ea_ibody_quota;
1405         int             inode_exp = 0;
1406
1407         init_resource_track(&rtrack, ctx->fs->io);
1408         clear_problem_context(&pctx);
1409
1410         /* If we can do readahead, figure out how many groups to pull in. */
1411         if (!e2fsck_can_readahead(ctx->fs))
1412                 ctx->readahead_kb = 0;
1413         else if (ctx->readahead_kb == ~0ULL)
1414                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1415         pass1_readahead(ctx, &ra_group, &ino_threshold);
1416
1417         if (!(ctx->options & E2F_OPT_PREEN))
1418                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1419
1420         if (ext2fs_has_feature_dir_index(fs->super) &&
1421             !(ctx->options & E2F_OPT_NO)) {
1422                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1423                         ctx->dirs_to_hash = 0;
1424         }
1425
1426 #ifdef MTRACE
1427         mtrace_print("Pass 1");
1428 #endif
1429
1430 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1431
1432         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1433                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1434                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1435                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1436                 max_sizes = (max_sizes * (1UL << i));
1437                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1438         }
1439 #undef EXT2_BPP
1440
1441         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1442         extent_fs = ext2fs_has_feature_extents(sb);
1443         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1444         casefold_fs = ext2fs_has_feature_casefold(sb);
1445
1446         /*
1447          * Allocate bitmaps structures
1448          */
1449         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1450                                                     EXT2FS_BMAP64_RBTREE,
1451                                                     "inode_used_map",
1452                                                     &ctx->inode_used_map);
1453         if (pctx.errcode) {
1454                 pctx.num = 1;
1455                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1456                 ctx->flags |= E2F_FLAG_ABORT;
1457                 return;
1458         }
1459         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1460                         _("directory inode map"),
1461                         EXT2FS_BMAP64_AUTODIR,
1462                         "inode_dir_map", &ctx->inode_dir_map);
1463         if (pctx.errcode) {
1464                 pctx.num = 2;
1465                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1466                 ctx->flags |= E2F_FLAG_ABORT;
1467                 return;
1468         }
1469         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1470                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1471                         "inode_reg_map", &ctx->inode_reg_map);
1472         if (pctx.errcode) {
1473                 pctx.num = 6;
1474                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1475                 ctx->flags |= E2F_FLAG_ABORT;
1476                 return;
1477         }
1478         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1479                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1480                         "block_found_map", &ctx->block_found_map);
1481         if (pctx.errcode) {
1482                 pctx.num = 1;
1483                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1484                 ctx->flags |= E2F_FLAG_ABORT;
1485                 return;
1486         }
1487         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1488                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1489                         "block_metadata_map", &ctx->block_metadata_map);
1490         if (pctx.errcode) {
1491                 pctx.num = 1;
1492                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1493                 ctx->flags |= E2F_FLAG_ABORT;
1494                 return;
1495         }
1496         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1497                                            &ctx->inode_link_info);
1498         if (pctx.errcode) {
1499                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1500                 ctx->flags |= E2F_FLAG_ABORT;
1501                 return;
1502         }
1503         bufsize = inode_size;
1504         if (bufsize < sizeof(struct ext2_inode_large))
1505                 bufsize = sizeof(struct ext2_inode_large);
1506         inode = (struct ext2_inode *)
1507                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1508
1509         inodes_to_process = (struct process_inode_block *)
1510                 e2fsck_allocate_memory(ctx,
1511                                        (ctx->process_inode_size *
1512                                         sizeof(struct process_inode_block)),
1513                                        "array of inodes to process");
1514         process_inode_count = 0;
1515
1516         pctx.errcode = ext2fs_init_dblist(fs, 0);
1517         if (pctx.errcode) {
1518                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1519                 ctx->flags |= E2F_FLAG_ABORT;
1520                 goto endit;
1521         }
1522
1523         /*
1524          * If the last orphan field is set, clear it, since the pass1
1525          * processing will automatically find and clear the orphans.
1526          * In the future, we may want to try using the last_orphan
1527          * linked list ourselves, but for now, we clear it so that the
1528          * ext3 mount code won't get confused.
1529          */
1530         if (!(ctx->options & E2F_OPT_READONLY)) {
1531                 if (fs->super->s_last_orphan) {
1532                         fs->super->s_last_orphan = 0;
1533                         ext2fs_mark_super_dirty(fs);
1534                 }
1535         }
1536
1537         mark_table_blocks(ctx);
1538         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1539                                                 &ctx->block_found_map);
1540         if (pctx.errcode) {
1541                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1542                 ctx->flags |= E2F_FLAG_ABORT;
1543                 goto endit;
1544         }
1545         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1546                                                     "block interate buffer");
1547         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1548                 e2fsck_use_inode_shortcuts(ctx, 1);
1549         e2fsck_intercept_block_allocations(ctx);
1550         old_op = ehandler_operation(_("opening inode scan"));
1551         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1552                                               &scan);
1553         ehandler_operation(old_op);
1554         if (pctx.errcode) {
1555                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1556                 ctx->flags |= E2F_FLAG_ABORT;
1557                 goto endit;
1558         }
1559         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1560                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1561         ctx->stashed_inode = inode;
1562         scan_struct.ctx = ctx;
1563         scan_struct.block_buf = block_buf;
1564         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1565         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1566                                               ctx->fs->group_desc_count)))
1567                 goto endit;
1568         if ((fs->super->s_wtime &&
1569              fs->super->s_wtime < fs->super->s_inodes_count) ||
1570             (fs->super->s_mtime &&
1571              fs->super->s_mtime < fs->super->s_inodes_count) ||
1572             (fs->super->s_mkfs_time &&
1573              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1574                 low_dtime_check = 0;
1575
1576         if (ext2fs_has_feature_mmp(fs->super) &&
1577             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1578             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1579                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1580                                           fs->super->s_mmp_block);
1581
1582         /* Set up ctx->lost_and_found if possible */
1583         (void) e2fsck_get_lost_and_found(ctx, 0);
1584
1585         while (1) {
1586                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1587                         if (e2fsck_mmp_update(fs))
1588                                 fatal_error(ctx, 0);
1589                 }
1590                 old_op = ehandler_operation(eop_next_inode);
1591                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1592                                                           inode, inode_size);
1593                 if (ino > ino_threshold)
1594                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1595                 ehandler_operation(old_op);
1596                 if (e2fsck_should_abort(ctx))
1597                         goto endit;
1598                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1599                         /*
1600                          * If badblocks says badblocks is bad, offer to clear
1601                          * the list, update the in-core bb list, and restart
1602                          * the inode scan.
1603                          */
1604                         if (ino == EXT2_BAD_INO &&
1605                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1606                                         &pctx)) {
1607                                 errcode_t err;
1608
1609                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1610                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1611                                 ctx->fs->badblocks = NULL;
1612                                 err = ext2fs_read_bb_inode(ctx->fs,
1613                                                         &ctx->fs->badblocks);
1614                                 if (err) {
1615                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1616                                                     &pctx);
1617                                         ctx->flags |= E2F_FLAG_ABORT;
1618                                         goto endit;
1619                                 }
1620                                 err = ext2fs_inode_scan_goto_blockgroup(scan,
1621                                                                         0);
1622                                 if (err) {
1623                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1624                                                     &pctx);
1625                                         ctx->flags |= E2F_FLAG_ABORT;
1626                                         goto endit;
1627                                 }
1628                                 continue;
1629                         }
1630                         if (!ctx->inode_bb_map)
1631                                 alloc_bb_map(ctx);
1632                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1633                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1634                         continue;
1635                 }
1636                 if (pctx.errcode &&
1637                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1638                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1639                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1640                         ctx->flags |= E2F_FLAG_ABORT;
1641                         goto endit;
1642                 }
1643                 if (!ino)
1644                         break;
1645                 pctx.ino = ino;
1646                 pctx.inode = inode;
1647                 ctx->stashed_ino = ino;
1648
1649                 /* Clear trashed inode? */
1650                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1651                     inode->i_links_count > 0 &&
1652                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1653                         pctx.errcode = 0;
1654                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1655                 }
1656                 failed_csum = pctx.errcode != 0;
1657
1658                 /*
1659                  * Check for inodes who might have been part of the
1660                  * orphaned list linked list.  They should have gotten
1661                  * dealt with by now, unless the list had somehow been
1662                  * corrupted.
1663                  *
1664                  * FIXME: In the future, inodes which are still in use
1665                  * (and which are therefore) pending truncation should
1666                  * be handled specially.  Right now we just clear the
1667                  * dtime field, and the normal e2fsck handling of
1668                  * inodes where i_size and the inode blocks are
1669                  * inconsistent is to fix i_size, instead of releasing
1670                  * the extra blocks.  This won't catch the inodes that
1671                  * was at the end of the orphan list, but it's better
1672                  * than nothing.  The right answer is that there
1673                  * shouldn't be any bugs in the orphan list handling.  :-)
1674                  */
1675                 if (inode->i_dtime && low_dtime_check &&
1676                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1677                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1678                                 inode->i_dtime = inode->i_links_count ?
1679                                         0 : ctx->now;
1680                                 e2fsck_write_inode(ctx, ino, inode,
1681                                                    "pass1");
1682                                 failed_csum = 0;
1683                         }
1684                 }
1685
1686                 if (inode->i_links_count) {
1687                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1688                                            ino, inode->i_links_count);
1689                         if (pctx.errcode) {
1690                                 pctx.num = inode->i_links_count;
1691                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1692                                 ctx->flags |= E2F_FLAG_ABORT;
1693                                 goto endit;
1694                         }
1695                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1696                            !quota_inum_is_reserved(fs, ino)) {
1697                         if (!inode->i_dtime && inode->i_mode) {
1698                                 if (fix_problem(ctx,
1699                                             PR_1_ZERO_DTIME, &pctx)) {
1700                                         inode->i_dtime = ctx->now;
1701                                         e2fsck_write_inode(ctx, ino, inode,
1702                                                            "pass1");
1703                                         failed_csum = 0;
1704                                 }
1705                         }
1706                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1707                         continue;
1708                 }
1709
1710                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1711                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1712                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1713                      (!casefold_fs &&
1714                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1715                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1716                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1717                 }
1718
1719                 /* Conflicting inlinedata/extents inode flags? */
1720                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1721                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1722                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1723                                                                inode_size,
1724                                                                &pctx);
1725                         if (res < 0) {
1726                                 /* skip FINISH_INODE_LOOP */
1727                                 continue;
1728                         }
1729                 }
1730
1731                 /* Test for incorrect inline_data flags settings. */
1732                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1733                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1734                         size_t size = 0;
1735
1736                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1737                         if (!pctx.errcode &&
1738                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1739                                 ext2fs_set_feature_inline_data(sb);
1740                                 ext2fs_mark_super_dirty(fs);
1741                                 inlinedata_fs = 1;
1742                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1743                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1744                                 /* skip FINISH_INODE_LOOP */
1745                                 continue;
1746                         }
1747                 }
1748
1749                 /* Test for inline data flag but no attr */
1750                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1751                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1752                         size_t size = 0;
1753                         errcode_t err;
1754                         int flags;
1755
1756                         flags = fs->flags;
1757                         if (failed_csum)
1758                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1759                         err = get_inline_data_ea_size(fs, ino, &size);
1760                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1761                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1762
1763                         switch (err) {
1764                         case 0:
1765                                 /* Everything is awesome... */
1766                                 break;
1767                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1768                         case EXT2_ET_BAD_EA_HASH:
1769                         case EXT2_ET_BAD_EA_HEADER:
1770                         case EXT2_ET_EA_BAD_NAME_LEN:
1771                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1772                         case EXT2_ET_EA_KEY_NOT_FOUND:
1773                         case EXT2_ET_EA_NO_SPACE:
1774                         case EXT2_ET_MISSING_EA_FEATURE:
1775                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1776                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1777                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1778                         case EXT2_ET_NO_INLINE_DATA:
1779                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1780                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1781                         case EXT2_ET_EA_INODE_CORRUPTED:
1782                                 /* broken EA or no system.data EA; truncate */
1783                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1784                                                 &pctx)) {
1785                                         err = ext2fs_inode_size_set(fs, inode, 0);
1786                                         if (err) {
1787                                                 pctx.errcode = err;
1788                                                 ctx->flags |= E2F_FLAG_ABORT;
1789                                                 goto endit;
1790                                         }
1791                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1792                                         memset(&inode->i_block, 0,
1793                                                sizeof(inode->i_block));
1794                                         e2fsck_write_inode(ctx, ino, inode,
1795                                                            "pass1");
1796                                         failed_csum = 0;
1797                                 }
1798                                 break;
1799                         default:
1800                                 /* Some other kind of non-xattr error? */
1801                                 pctx.errcode = err;
1802                                 ctx->flags |= E2F_FLAG_ABORT;
1803                                 goto endit;
1804                         }
1805                 }
1806
1807                 /*
1808                  * Test for incorrect extent flag settings.
1809                  *
1810                  * On big-endian machines we must be careful:
1811                  * When the inode is read, the i_block array is not swapped
1812                  * if the extent flag is set.  Therefore if we are testing
1813                  * for or fixing a wrongly-set flag, we must potentially
1814                  * (un)swap before testing, or after fixing.
1815                  */
1816
1817                 /*
1818                  * In this case the extents flag was set when read, so
1819                  * extent_header_verify is ok.  If the inode is cleared,
1820                  * no need to swap... so no extra swapping here.
1821                  */
1822                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1823                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1824                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1825                         if ((ext2fs_extent_header_verify(inode->i_block,
1826                                                  sizeof(inode->i_block)) == 0) &&
1827                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1828                                 ext2fs_set_feature_extents(sb);
1829                                 ext2fs_mark_super_dirty(fs);
1830                                 extent_fs = 1;
1831                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1832                         clear_inode:
1833                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1834                                 if (ino == EXT2_BAD_INO)
1835                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1836                                                                  ino);
1837                                 /* skip FINISH_INODE_LOOP */
1838                                 continue;
1839                         }
1840                 }
1841
1842                 /*
1843                  * For big-endian machines:
1844                  * If the inode didn't have the extents flag set when it
1845                  * was read, then the i_blocks array was swapped.  To test
1846                  * as an extents header, we must swap it back first.
1847                  * IF we then set the extents flag, the entire i_block
1848                  * array must be un/re-swapped to make it proper extents data.
1849                  */
1850                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1851                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1852                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1853                     (LINUX_S_ISREG(inode->i_mode) ||
1854                      LINUX_S_ISDIR(inode->i_mode))) {
1855                         void *ehp;
1856 #ifdef WORDS_BIGENDIAN
1857                         __u32 tmp_block[EXT2_N_BLOCKS];
1858
1859                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1860                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1861                         ehp = tmp_block;
1862 #else
1863                         ehp = inode->i_block;
1864 #endif
1865                         if ((ext2fs_extent_header_verify(ehp,
1866                                          sizeof(inode->i_block)) == 0)) {
1867                                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
1868                                 if (fix_problem(ctx, PR_1_UNSET_EXTENT_FL,
1869                                                 &pctx)) {
1870                                         inode->i_flags |= EXT4_EXTENTS_FL;
1871 #ifdef WORDS_BIGENDIAN
1872                                         memcpy(inode->i_block, tmp_block,
1873                                                sizeof(inode->i_block));
1874 #endif
1875                                         e2fsck_write_inode(ctx, ino, inode,
1876                                                            "pass1");
1877                                         failed_csum = 0;
1878                                 }
1879                         }
1880                 }
1881
1882                 if (ino == EXT2_BAD_INO) {
1883                         struct process_block_struct pb;
1884
1885                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1886                              inode->i_gid || inode->i_links_count ||
1887                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1888                              inode->i_file_acl) &&
1889                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1890                                 memset(inode, 0, sizeof(struct ext2_inode));
1891                                 e2fsck_write_inode(ctx, ino, inode,
1892                                                    "clear bad inode");
1893                                 failed_csum = 0;
1894                         }
1895
1896                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1897                                                           &pb.fs_meta_blocks);
1898                         if (pctx.errcode) {
1899                                 pctx.num = 4;
1900                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1901                                 ctx->flags |= E2F_FLAG_ABORT;
1902                                 goto endit;
1903                         }
1904                         pb.ino = EXT2_BAD_INO;
1905                         pb.num_blocks = pb.last_block = 0;
1906                         pb.last_db_block = -1;
1907                         pb.num_illegal_blocks = 0;
1908                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1909                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1910                         pb.inode = inode;
1911                         pb.pctx = &pctx;
1912                         pb.ctx = ctx;
1913                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1914                                      block_buf, process_bad_block, &pb);
1915                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1916                         if (pctx.errcode) {
1917                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1918                                 ctx->flags |= E2F_FLAG_ABORT;
1919                                 goto endit;
1920                         }
1921                         if (pb.bbcheck)
1922                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1923                                 ctx->flags |= E2F_FLAG_ABORT;
1924                                 goto endit;
1925                         }
1926                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1927                         clear_problem_context(&pctx);
1928                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1929                         continue;
1930                 } else if (ino == EXT2_ROOT_INO) {
1931                         /*
1932                          * Make sure the root inode is a directory; if
1933                          * not, offer to clear it.  It will be
1934                          * regenerated in pass #3.
1935                          */
1936                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1937                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1938                                         goto clear_inode;
1939                         }
1940                         /*
1941                          * If dtime is set, offer to clear it.  mke2fs
1942                          * version 0.2b created filesystems with the
1943                          * dtime field set for the root and lost+found
1944                          * directories.  We won't worry about
1945                          * /lost+found, since that can be regenerated
1946                          * easily.  But we will fix the root directory
1947                          * as a special case.
1948                          */
1949                         if (inode->i_dtime && inode->i_links_count) {
1950                                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
1951                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1952                                         inode->i_dtime = 0;
1953                                         e2fsck_write_inode(ctx, ino, inode,
1954                                                            "pass1");
1955                                         failed_csum = 0;
1956                                 }
1957                         }
1958                 } else if (ino == EXT2_JOURNAL_INO) {
1959                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1960                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1961                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1962                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1963                                                 &pctx)) {
1964                                         inode->i_mode = LINUX_S_IFREG;
1965                                         e2fsck_write_inode(ctx, ino, inode,
1966                                                            "pass1");
1967                                         failed_csum = 0;
1968                                 }
1969                                 check_blocks(ctx, &pctx, block_buf, NULL);
1970                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1971                                 continue;
1972                         }
1973                         if ((inode->i_links_count ||
1974                              inode->i_blocks || inode->i_block[0]) &&
1975                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1976                                         &pctx)) {
1977                                 memset(inode, 0, inode_size);
1978                                 ext2fs_icount_store(ctx->inode_link_info,
1979                                                     ino, 0);
1980                                 e2fsck_write_inode_full(ctx, ino, inode,
1981                                                         inode_size, "pass1");
1982                                 failed_csum = 0;
1983                         }
1984                 } else if (quota_inum_is_reserved(fs, ino)) {
1985                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1986                         if (ext2fs_has_feature_quota(fs->super) &&
1987                             quota_inum_is_super(fs->super, ino)) {
1988                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1989                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1990                                                         &pctx)) {
1991                                         inode->i_mode = LINUX_S_IFREG;
1992                                         e2fsck_write_inode(ctx, ino, inode,
1993                                                         "pass1");
1994                                         failed_csum = 0;
1995                                 }
1996                                 check_blocks(ctx, &pctx, block_buf, NULL);
1997                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1998                                 continue;
1999                         }
2000                         if ((inode->i_links_count ||
2001                              inode->i_blocks || inode->i_block[0]) &&
2002                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
2003                                         &pctx)) {
2004                                 memset(inode, 0, inode_size);
2005                                 ext2fs_icount_store(ctx->inode_link_info,
2006                                                     ino, 0);
2007                                 e2fsck_write_inode_full(ctx, ino, inode,
2008                                                         inode_size, "pass1");
2009                                 failed_csum = 0;
2010                         }
2011                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
2012                         problem_t problem = 0;
2013
2014                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2015                         if (ino == EXT2_BOOT_LOADER_INO) {
2016                                 if (LINUX_S_ISDIR(inode->i_mode))
2017                                         problem = PR_1_RESERVED_BAD_MODE;
2018                         } else if (ino == EXT2_RESIZE_INO) {
2019                                 if (inode->i_mode &&
2020                                     !LINUX_S_ISREG(inode->i_mode))
2021                                         problem = PR_1_RESERVED_BAD_MODE;
2022                         } else {
2023                                 if (inode->i_mode != 0)
2024                                         problem = PR_1_RESERVED_BAD_MODE;
2025                         }
2026                         if (problem) {
2027                                 if (fix_problem(ctx, problem, &pctx)) {
2028                                         inode->i_mode = 0;
2029                                         e2fsck_write_inode(ctx, ino, inode,
2030                                                            "pass1");
2031                                         failed_csum = 0;
2032                                 }
2033                         }
2034                         check_blocks(ctx, &pctx, block_buf, NULL);
2035                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2036                         continue;
2037                 }
2038
2039                 if (!inode->i_links_count) {
2040                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2041                         continue;
2042                 }
2043                 /*
2044                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
2045                  * deleted files.  Oops.
2046                  *
2047                  * Since all new ext2 implementations get this right,
2048                  * we now assume that the case of non-zero
2049                  * i_links_count and non-zero dtime means that we
2050                  * should keep the file, not delete it.
2051                  *
2052                  */
2053                 if (inode->i_dtime) {
2054                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2055                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
2056                                 inode->i_dtime = 0;
2057                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
2058                                 failed_csum = 0;
2059                         }
2060                 }
2061
2062                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
2063                 switch (fs->super->s_creator_os) {
2064                     case EXT2_OS_HURD:
2065                         frag = inode->osd2.hurd2.h_i_frag;
2066                         fsize = inode->osd2.hurd2.h_i_fsize;
2067                         break;
2068                     default:
2069                         frag = fsize = 0;
2070                 }
2071
2072                 /* Fixed in pass2, e2fsck_process_bad_inode(). */
2073                 if (inode->i_faddr || frag || fsize ||
2074                     (!ext2fs_has_feature_largedir(fs->super) &&
2075                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
2076                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2077                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2078                     !ext2fs_has_feature_64bit(fs->super) &&
2079                     inode->osd2.linux2.l_i_file_acl_high != 0)
2080                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2081                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
2082                     !ext2fs_has_feature_huge_file(fs->super) &&
2083                     (inode->osd2.linux2.l_i_blocks_hi != 0))
2084                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2085                 if (inode->i_flags & EXT2_IMAGIC_FL) {
2086                         if (imagic_fs) {
2087                                 if (!ctx->inode_imagic_map)
2088                                         alloc_imagic_map(ctx);
2089                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
2090                                                          ino);
2091                         } else {
2092                                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2093                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
2094                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
2095                                         e2fsck_write_inode(ctx, ino,
2096                                                            inode, "pass1");
2097                                         failed_csum = 0;
2098                                 }
2099                         }
2100                 }
2101
2102                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
2103                 check_is_really_dir(ctx, &pctx, block_buf);
2104
2105                 /*
2106                  * ext2fs_inode_has_valid_blocks2 does not actually look
2107                  * at i_block[] values, so not endian-sensitive here.
2108                  */
2109                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
2110                     LINUX_S_ISLNK(inode->i_mode) &&
2111                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
2112                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
2113                         inode->i_flags &= ~EXT4_EXTENTS_FL;
2114                         e2fsck_write_inode(ctx, ino, inode, "pass1");
2115                         failed_csum = 0;
2116                 }
2117
2118                 if (LINUX_S_ISDIR(inode->i_mode)) {
2119                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
2120                         e2fsck_add_dir_info(ctx, ino, 0);
2121                         ctx->fs_directory_count++;
2122                         if (inode->i_flags & EXT4_ENCRYPT_FL)
2123                                 add_encrypted_dir(ctx, ino);
2124                 } else if (LINUX_S_ISREG (inode->i_mode)) {
2125                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
2126                         ctx->fs_regular_count++;
2127                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
2128                            e2fsck_pass1_check_device_inode(fs, inode)) {
2129                         check_extents_inlinedata(ctx, &pctx);
2130                         check_immutable(ctx, &pctx);
2131                         check_size(ctx, &pctx);
2132                         ctx->fs_chardev_count++;
2133                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
2134                            e2fsck_pass1_check_device_inode(fs, inode)) {
2135                         check_extents_inlinedata(ctx, &pctx);
2136                         check_immutable(ctx, &pctx);
2137                         check_size(ctx, &pctx);
2138                         ctx->fs_blockdev_count++;
2139                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
2140                            check_symlink(ctx, &pctx, ino, inode, block_buf)) {
2141                         check_immutable(ctx, &pctx);
2142                         ctx->fs_symlinks_count++;
2143                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
2144                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2145                                 continue;
2146                         } else if (ext2fs_is_fast_symlink(inode)) {
2147                                 ctx->fs_fast_symlinks_count++;
2148                                 check_blocks(ctx, &pctx, block_buf,
2149                                              &ea_ibody_quota);
2150                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2151                                 continue;
2152                         }
2153                 }
2154                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2155                          e2fsck_pass1_check_device_inode(fs, inode)) {
2156                         check_extents_inlinedata(ctx, &pctx);
2157                         check_immutable(ctx, &pctx);
2158                         check_size(ctx, &pctx);
2159                         ctx->fs_fifo_count++;
2160                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2161                            e2fsck_pass1_check_device_inode(fs, inode)) {
2162                         check_extents_inlinedata(ctx, &pctx);
2163                         check_immutable(ctx, &pctx);
2164                         check_size(ctx, &pctx);
2165                         ctx->fs_sockets_count++;
2166                 } else {
2167                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2168                 }
2169
2170                 if (EXT4_XTIME_FUTURE(ctx, sb, inode->i_atime, ctx->time_fudge))
2171                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2172                 else if (EXT4_XTIME_FUTURE(ctx, sb, inode->i_mtime,
2173                                            ctx->time_fudge))
2174                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
2175
2176                 if (EXT4_XTIME_FUTURE(ctx, sb, inode->i_ctime, ctx->time_fudge))
2177                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_HIGH);
2178                 else if (!precreated_object(inode) &&
2179                          EXT4_XTIME_ANCIENT(ctx, sb, inode->i_ctime,
2180                                             ctx->time_fudge))
2181                         e2fsck_mark_inode_bad(ctx, ino, BADNESS_HIGH);
2182
2183                 /* i_crtime is checked in check_inode_extra_space() */
2184
2185                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2186                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2187                         if (inode->i_block[EXT2_IND_BLOCK])
2188                                 ctx->fs_ind_count++;
2189                         if (inode->i_block[EXT2_DIND_BLOCK])
2190                                 ctx->fs_dind_count++;
2191                         if (inode->i_block[EXT2_TIND_BLOCK])
2192                                 ctx->fs_tind_count++;
2193                 }
2194                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2195                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2196                     (inode->i_block[EXT2_IND_BLOCK] ||
2197                      inode->i_block[EXT2_DIND_BLOCK] ||
2198                      inode->i_block[EXT2_TIND_BLOCK] ||
2199                      ext2fs_file_acl_block(fs, inode))) {
2200                         struct process_inode_block *itp;
2201
2202                         itp = &inodes_to_process[process_inode_count];
2203                         itp->ino = ino;
2204                         itp->ea_ibody_quota = ea_ibody_quota;
2205                         if (inode_size < sizeof(struct ext2_inode_large))
2206                                 memcpy(&itp->inode, inode, inode_size);
2207                         else
2208                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2209                         process_inode_count++;
2210                 } else
2211                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2212
2213                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2214
2215                 if (ctx->flags & E2F_FLAG_EXPAND_EISIZE) {
2216                         struct ext2_inode_large *inode_l;
2217
2218                         inode_l = (struct ext2_inode_large *)inode;
2219
2220                         if (inode_l->i_extra_isize < ctx->want_extra_isize) {
2221                                 fix_problem(ctx, PR_1_EXPAND_EISIZE, &pctx);
2222                                 inode_exp = e2fsck_pass1_expand_eisize(ctx,
2223                                                                        inode_l,
2224                                                                        &pctx);
2225                         }
2226                         if ((inode_l->i_extra_isize < ctx->min_extra_isize) &&
2227                             inode_exp == 0)
2228                                 ctx->min_extra_isize = inode_l->i_extra_isize;
2229                 }
2230
2231                 if (e2fsck_should_abort(ctx))
2232                         goto endit;
2233
2234                 if (process_inode_count >= ctx->process_inode_size) {
2235                         process_inodes(ctx, block_buf);
2236
2237                         if (e2fsck_should_abort(ctx))
2238                                 goto endit;
2239                 }
2240         }
2241         process_inodes(ctx, block_buf);
2242         ext2fs_close_inode_scan(scan);
2243         scan = NULL;
2244
2245         reserve_block_for_root_repair(ctx);
2246         reserve_block_for_lnf_repair(ctx);
2247
2248         /*
2249          * If any extended attribute blocks' reference counts need to
2250          * be adjusted, either up (ctx->refcount_extra), or down
2251          * (ctx->refcount), then fix them.
2252          */
2253         if (ctx->refcount) {
2254                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2255                 ea_refcount_free(ctx->refcount);
2256                 ctx->refcount = 0;
2257         }
2258         if (ctx->refcount_extra) {
2259                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2260                                         block_buf, +1);
2261                 ea_refcount_free(ctx->refcount_extra);
2262                 ctx->refcount_extra = 0;
2263         }
2264
2265         if (ctx->ea_block_quota_blocks) {
2266                 ea_refcount_free(ctx->ea_block_quota_blocks);
2267                 ctx->ea_block_quota_blocks = 0;
2268         }
2269
2270         if (ctx->ea_block_quota_inodes) {
2271                 ea_refcount_free(ctx->ea_block_quota_inodes);
2272                 ctx->ea_block_quota_inodes = 0;
2273         }
2274
2275         if (ctx->invalid_bitmaps)
2276                 handle_fs_bad_blocks(ctx);
2277
2278         /* We don't need the block_ea_map any more */
2279         if (ctx->block_ea_map) {
2280                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2281                 ctx->block_ea_map = 0;
2282         }
2283
2284         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2285                 clear_problem_context(&pctx);
2286                 pctx.errcode = ext2fs_create_resize_inode(fs);
2287                 if (pctx.errcode) {
2288                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2289                                          &pctx)) {
2290                                 ctx->flags |= E2F_FLAG_ABORT;
2291                                 goto endit;
2292                         }
2293                         pctx.errcode = 0;
2294                 }
2295                 if (!pctx.errcode) {
2296                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2297                                           "recreate inode");
2298                         inode->i_mtime = ctx->now;
2299                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2300                                            "recreate inode");
2301                 }
2302                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2303         }
2304
2305         if (ctx->flags & E2F_FLAG_RESTART) {
2306                 /*
2307                  * Only the master copy of the superblock and block
2308                  * group descriptors are going to be written during a
2309                  * restart, so set the superblock to be used to be the
2310                  * master superblock.
2311                  */
2312                 ctx->use_superblock = 0;
2313                 unwind_pass1(fs);
2314                 goto endit;
2315         }
2316
2317         if (ctx->block_dup_map) {
2318                 if (ctx->options & E2F_OPT_PREEN) {
2319                         clear_problem_context(&pctx);
2320                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2321                 }
2322                 e2fsck_pass1_dupblocks(ctx, block_buf);
2323         }
2324         ctx->flags |= E2F_FLAG_ALLOC_OK;
2325         ext2fs_free_mem(&inodes_to_process);
2326 endit:
2327         e2fsck_use_inode_shortcuts(ctx, 0);
2328
2329         if (scan)
2330                 ext2fs_close_inode_scan(scan);
2331         if (block_buf)
2332                 ext2fs_free_mem(&block_buf);
2333         if (inode)
2334                 ext2fs_free_mem(&inode);
2335
2336         /*
2337          * The l+f inode may have been cleared, so zap it now and
2338          * later passes will recalculate it if necessary
2339          */
2340         ctx->lost_and_found = 0;
2341
2342         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2343                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2344         else
2345                 ctx->invalid_bitmaps++;
2346 }
2347
2348 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2349                                           ext2fs_generic_bitmap *dest)
2350 {
2351         errcode_t ret;
2352
2353         ret = ext2fs_copy_bitmap(*src, dest);
2354         if (ret)
2355                 return ret;
2356
2357         (*dest)->fs = fs;
2358         ext2fs_free_generic_bmap(*src);
2359         *src = NULL;
2360
2361         return 0;
2362 }
2363
2364
2365 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2366                                       ext2_filsys src)
2367 {
2368         errcode_t       retval;
2369
2370         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2371         if (dest->dblist)
2372                 dest->dblist->fs = dest;
2373         if (src->block_map) {
2374                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2375                                                   &dest->block_map);
2376                 if (retval)
2377                         return retval;
2378         }
2379         if (src->inode_map) {
2380                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2381                                                   &dest->inode_map);
2382                 if (retval)
2383                         return retval;
2384         }
2385
2386         if (src->badblocks) {
2387                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2388                 if (retval)
2389                         return retval;
2390         }
2391
2392         /* disable it for now */
2393         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2394         retval = ext2fs_open_channel(dest, src_context->io_options,
2395                                      src_context->io_manager,
2396                                      src_context->openfs_flags,
2397                                      src->io->block_size);
2398         if (retval)
2399                 return retval;
2400
2401         /* Block size might not be default */
2402         io_channel_set_blksize(dest->io, src->io->block_size);
2403         ehandler_init(dest->io);
2404
2405         assert(dest->io->magic == src->io->magic);
2406         assert(dest->io->manager == src->io->manager);
2407         assert(strcmp(dest->io->name, src->io->name) == 0);
2408         assert(dest->io->block_size == src->io->block_size);
2409         assert(dest->io->read_error == src->io->read_error);
2410         assert(dest->io->write_error == src->io->write_error);
2411         assert(dest->io->refcount == src->io->refcount);
2412         assert(dest->io->flags == src->io->flags);
2413         assert(dest->io->app_data == dest);
2414         assert(src->io->app_data == src);
2415         assert(dest->io->align == src->io->align);
2416
2417         /* The data should be written to disk immediately */
2418         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2419         /* icache will be rebuilt if needed, so do not copy from @src */
2420         src->icache = NULL;
2421         return 0;
2422 }
2423
2424 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2425 {
2426         struct ext2_inode_cache *icache = dest->icache;
2427         errcode_t retval = 0;
2428         io_channel dest_io;
2429         io_channel dest_image_io;
2430
2431         dest_io = dest->io;
2432         dest_image_io = dest->image_io;
2433
2434         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2435         dest->io = dest_io;
2436         dest->image_io = dest_image_io;
2437         dest->icache = icache;
2438         if (dest->dblist)
2439                 dest->dblist->fs = dest;
2440         if (src->inode_map) {
2441                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2442                                                   &dest->inode_map);
2443                 if (retval)
2444                         return retval;
2445         }
2446         if (src->block_map) {
2447                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2448                                                   &dest->block_map);
2449                 if (retval)
2450                         return retval;
2451         }
2452
2453         if (src->icache) {
2454                 ext2fs_free_inode_cache(src->icache);
2455                 src->icache = NULL;
2456         }
2457
2458         if (src->badblocks) {
2459                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2460
2461                 ext2fs_badblocks_list_free(src->badblocks);
2462                 src->badblocks = NULL;
2463         }
2464
2465         io_channel_close(src->io);
2466         return retval;
2467 }
2468
2469 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx)
2470 {
2471         errcode_t       retval;
2472         e2fsck_t        thread_context;
2473         ext2_filsys     thread_fs;
2474         ext2_filsys     global_fs = global_ctx->fs;
2475
2476         assert(global_ctx->inode_used_map == NULL);
2477         assert(global_ctx->inode_dir_map == NULL);
2478         assert(global_ctx->inode_bb_map == NULL);
2479         assert(global_ctx->inode_imagic_map == NULL);
2480         assert(global_ctx->inode_reg_map == NULL);
2481         assert(global_ctx->inodes_to_rebuild == NULL);
2482
2483         assert(global_ctx->block_found_map == NULL);
2484         assert(global_ctx->block_dup_map == NULL);
2485         assert(global_ctx->block_ea_map == NULL);
2486         assert(global_ctx->block_metadata_map == NULL);
2487         assert(global_ctx->fs->dblist == NULL);
2488
2489         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2490         if (retval) {
2491                 com_err(global_ctx->program_name, retval, "while allocating memory");
2492                 return retval;
2493         }
2494         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2495         thread_context->global_ctx = global_ctx;
2496
2497         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2498         if (retval) {
2499                 com_err(global_ctx->program_name, retval, "while allocating memory");
2500                 goto out_context;
2501         }
2502
2503         io_channel_flush_cleanup(global_fs->io);
2504         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2505         if (retval) {
2506                 com_err(global_ctx->program_name, retval, "while copying fs");
2507                 goto out_fs;
2508         }
2509         thread_fs->priv_data = thread_context;
2510
2511         thread_context->fs = thread_fs;
2512         *thread_ctx = thread_context;
2513         return 0;
2514 out_fs:
2515         ext2fs_free_mem(&thread_fs);
2516 out_context:
2517         ext2fs_free_mem(&thread_context);
2518         return retval;
2519 }
2520
2521 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2522 {
2523         errcode_t       retval;
2524         int             flags = global_ctx->flags;
2525         ext2_filsys     thread_fs = thread_ctx->fs;
2526         ext2_filsys     global_fs = global_ctx->fs;
2527 #ifdef HAVE_SETJMP_H
2528         jmp_buf         old_jmp;
2529
2530         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2531 #endif
2532         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2533 #ifdef HAVE_SETJMP_H
2534         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2535 #endif
2536         /* Keep the global singal flags*/
2537         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2538                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2539
2540         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2541         if (retval) {
2542                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2543                 return retval;
2544         }
2545         global_fs->priv_data = global_ctx;
2546         global_ctx->fs = global_fs;
2547
2548         if (thread_ctx->inode_used_map) {
2549                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2550                                         &thread_ctx->inode_used_map,
2551                                         &global_ctx->inode_used_map);
2552                 if (retval)
2553                         return retval;
2554         }
2555         if (thread_ctx->inode_bad_map) {
2556                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2557                                         &thread_ctx->inode_bad_map,
2558                                         &global_ctx->inode_bad_map);
2559                 if (retval)
2560                         return retval;
2561         }
2562         if (thread_ctx->inode_dir_map) {
2563                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2564                                         &thread_ctx->inode_dir_map,
2565                                         &global_ctx->inode_dir_map);
2566                 if (retval)
2567                         return retval;
2568         }
2569         if (thread_ctx->inode_bb_map) {
2570                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2571                                         &thread_ctx->inode_bb_map,
2572                                         &global_ctx->inode_bb_map);
2573                 if (retval)
2574                         return retval;
2575         }
2576         if (thread_ctx->inode_imagic_map) {
2577                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2578                                         &thread_ctx->inode_imagic_map,
2579                                         &global_ctx->inode_imagic_map);
2580                 if (retval)
2581                         return retval;
2582         }
2583         if (thread_ctx->inode_reg_map) {
2584                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2585                                         &thread_ctx->inode_reg_map,
2586                                         &global_ctx->inode_reg_map);
2587                 if (retval)
2588                         return retval;
2589         }
2590         if (thread_ctx->inodes_to_rebuild) {
2591                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2592                                         &thread_ctx->inodes_to_rebuild,
2593                                         &global_ctx->inodes_to_rebuild);
2594                 if (retval)
2595                         return retval;
2596         }
2597         if (thread_ctx->block_found_map) {
2598                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2599                                         &thread_ctx->block_found_map,
2600                                         &global_ctx->block_found_map);
2601                 if (retval)
2602                         return retval;
2603         }
2604         if (thread_ctx->block_dup_map) {
2605                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2606                                         &thread_ctx->block_dup_map,
2607                                         &global_ctx->block_dup_map);
2608                 if (retval)
2609                         return retval;
2610         }
2611         if (thread_ctx->block_ea_map) {
2612                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2613                                         &thread_ctx->block_ea_map,
2614                                         &global_ctx->block_ea_map);
2615                 if (retval)
2616                         return retval;
2617         }
2618         if (thread_ctx->block_metadata_map) {
2619                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2620                                         &thread_ctx->block_metadata_map,
2621                                         &global_ctx->block_metadata_map);
2622                 if (retval)
2623                         return retval;
2624         }
2625
2626         return 0;
2627 }
2628
2629 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2630 {
2631         errcode_t       retval;
2632
2633         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2634         ext2fs_free_mem(&thread_ctx->fs);
2635         ext2fs_free_mem(&thread_ctx);
2636
2637         return retval;
2638 }
2639
2640 void e2fsck_pass1_multithread(e2fsck_t ctx)
2641 {
2642         errcode_t       retval;
2643         e2fsck_t        thread_ctx;
2644
2645         retval = e2fsck_pass1_thread_prepare(ctx, &thread_ctx);
2646         if (retval) {
2647                 com_err(ctx->program_name, 0,
2648                         _("while preparing pass1 thread\n"));
2649                 ctx->flags |= E2F_FLAG_ABORT;
2650                 return;
2651         }
2652
2653 #ifdef HAVE_SETJMP_H
2654         /*
2655          * When fatal_error() happens, jump to here. The thread
2656          * context's flags will be saved, but its abort_loc will
2657          * be overwritten by original jump buffer for the later
2658          * tests.
2659          */
2660         if (setjmp(thread_ctx->abort_loc)) {
2661                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2662                 e2fsck_pass1_thread_join(ctx, thread_ctx);
2663                 return;
2664         }
2665         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2666 #endif
2667
2668         e2fsck_pass1_thread(thread_ctx);
2669         retval = e2fsck_pass1_thread_join(ctx, thread_ctx);
2670         if (retval) {
2671                 com_err(ctx->program_name, 0,
2672                         _("while joining pass1 thread\n"));
2673                 ctx->flags |= E2F_FLAG_ABORT;
2674                 return;
2675         }
2676 }
2677
2678 void e2fsck_pass1(e2fsck_t ctx)
2679 {
2680         e2fsck_pass1_multithread(ctx);
2681 }
2682
2683 #undef FINISH_INODE_LOOP
2684
2685 /*
2686  * When the inode_scan routines call this callback at the end of the
2687  * glock group, call process_inodes.
2688  */
2689 static errcode_t scan_callback(ext2_filsys fs,
2690                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2691                                dgrp_t group, void * priv_data)
2692 {
2693         struct scan_callback_struct *scan_struct;
2694         e2fsck_t ctx;
2695
2696         scan_struct = (struct scan_callback_struct *) priv_data;
2697         ctx = scan_struct->ctx;
2698
2699         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2700
2701         if (ctx->progress)
2702                 if ((ctx->progress)(ctx, 1, group+1,
2703                                     ctx->fs->group_desc_count))
2704                         return EXT2_ET_CANCEL_REQUESTED;
2705
2706         return 0;
2707 }
2708
2709 /*
2710  * Process the inodes in the "inodes to process" list.
2711  */
2712 static void process_inodes(e2fsck_t ctx, char *block_buf)
2713 {
2714         int                     i;
2715         struct ext2_inode       *old_stashed_inode;
2716         ext2_ino_t              old_stashed_ino;
2717         const char              *old_operation;
2718         char                    buf[80];
2719         struct problem_context  pctx;
2720
2721 #if 0
2722         printf("begin process_inodes: ");
2723 #endif
2724         if (process_inode_count == 0)
2725                 return;
2726         old_operation = ehandler_operation(0);
2727         old_stashed_inode = ctx->stashed_inode;
2728         old_stashed_ino = ctx->stashed_ino;
2729         qsort(inodes_to_process, process_inode_count,
2730                       sizeof(struct process_inode_block), process_inode_cmp);
2731         clear_problem_context(&pctx);
2732         for (i=0; i < process_inode_count; i++) {
2733                 pctx.inode = ctx->stashed_inode =
2734                         (struct ext2_inode *) &inodes_to_process[i].inode;
2735                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2736
2737 #if 0
2738                 printf("%u ", pctx.ino);
2739 #endif
2740                 sprintf(buf, _("reading indirect blocks of inode %u"),
2741                         pctx.ino);
2742                 ehandler_operation(buf);
2743                 check_blocks(ctx, &pctx, block_buf,
2744                              &inodes_to_process[i].ea_ibody_quota);
2745                 if (e2fsck_should_abort(ctx))
2746                         break;
2747         }
2748         ctx->stashed_inode = old_stashed_inode;
2749         ctx->stashed_ino = old_stashed_ino;
2750         process_inode_count = 0;
2751 #if 0
2752         printf("end process inodes\n");
2753 #endif
2754         ehandler_operation(old_operation);
2755 }
2756
2757 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2758 {
2759         const struct process_inode_block *ib_a =
2760                 (const struct process_inode_block *) a;
2761         const struct process_inode_block *ib_b =
2762                 (const struct process_inode_block *) b;
2763         int     ret;
2764
2765         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2766                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2767         if (ret == 0)
2768                 /*
2769                  * We only call process_inodes() for non-extent
2770                  * inodes, so it's OK to pass NULL to
2771                  * ext2fs_file_acl_block() here.
2772                  */
2773                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2774                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2775         if (ret == 0)
2776                 ret = ib_a->ino - ib_b->ino;
2777         return ret;
2778 }
2779
2780 /*
2781  * Mark an inode as being bad and increment its badness counter.
2782  */
2783 void e2fsck_mark_inode_bad_loc(e2fsck_t ctx, ino_t ino, int count,
2784                                const char *func, const int line)
2785 {
2786         struct          problem_context pctx;
2787         __u16           result;
2788
2789         if (!ctx->inode_badness) {
2790                 clear_problem_context(&pctx);
2791
2792                 pctx.errcode = ext2fs_create_icount2(ctx->fs, 0, 0, NULL,
2793                                                      &ctx->inode_badness);
2794                 if (pctx.errcode) {
2795                         fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
2796                         ctx->flags |= E2F_FLAG_ABORT;
2797                         return;
2798                 }
2799         }
2800         ext2fs_icount_fetch(ctx->inode_badness, ino, &result);
2801         ext2fs_icount_store(ctx->inode_badness, ino, count + result);
2802
2803         if (ctx->options & E2F_OPT_DEBUG)
2804                 fprintf(stderr, "%s:%d: increase inode %lu badness %u to %u\n",
2805                         func, line, (unsigned long)ino, result, count + result);
2806 }
2807
2808 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino)
2809 {
2810         struct          problem_context pctx;
2811
2812         if (!ctx->encrypted_dirs) {
2813                 pctx.errcode = ext2fs_u32_list_create(&ctx->encrypted_dirs, 0);
2814                 if (pctx.errcode)
2815                         goto error;
2816         }
2817         pctx.errcode = ext2fs_u32_list_add(ctx->encrypted_dirs, ino);
2818         if (pctx.errcode == 0)
2819                 return;
2820 error:
2821         fix_problem(ctx, PR_1_ALLOCATE_ENCRYPTED_DIRLIST, &pctx);
2822         /* Should never get here */
2823         ctx->flags |= E2F_FLAG_ABORT;
2824 }
2825
2826 /*
2827  * This procedure will allocate the inode "bb" (badblock) map table
2828  */
2829 static void alloc_bb_map(e2fsck_t ctx)
2830 {
2831         struct          problem_context pctx;
2832
2833         clear_problem_context(&pctx);
2834         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2835                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2836                         "inode_bb_map", &ctx->inode_bb_map);
2837         if (pctx.errcode) {
2838                 pctx.num = 4;
2839                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2840                 /* Should never get here */
2841                 ctx->flags |= E2F_FLAG_ABORT;
2842                 return;
2843         }
2844 }
2845
2846 /*
2847  * This procedure will allocate the inode imagic table
2848  */
2849 static void alloc_imagic_map(e2fsck_t ctx)
2850 {
2851         struct          problem_context pctx;
2852
2853         clear_problem_context(&pctx);
2854         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2855                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2856                         "inode_imagic_map", &ctx->inode_imagic_map);
2857         if (pctx.errcode) {
2858                 pctx.num = 5;
2859                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2860                 /* Should never get here */
2861                 ctx->flags |= E2F_FLAG_ABORT;
2862                 return;
2863         }
2864 }
2865
2866 /*
2867  * Marks a block as in use, setting the dup_map if it's been set
2868  * already.  Called by process_block and process_bad_block.
2869  *
2870  * WARNING: Assumes checks have already been done to make sure block
2871  * is valid.  This is true in both process_block and process_bad_block.
2872  */
2873 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2874 {
2875         struct          problem_context pctx;
2876
2877         clear_problem_context(&pctx);
2878
2879         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2880                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2881                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2882                         return;
2883                 }
2884                 if (!ctx->block_dup_map) {
2885                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2886                                         _("multiply claimed block map"),
2887                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2888                                         &ctx->block_dup_map);
2889                         if (pctx.errcode) {
2890                                 pctx.num = 3;
2891                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2892                                             &pctx);
2893                                 /* Should never get here */
2894                                 ctx->flags |= E2F_FLAG_ABORT;
2895                                 return;
2896                         }
2897                 }
2898                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2899         } else {
2900                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2901         }
2902 }
2903
2904 /*
2905  * When cluster size is greater than one block, it is caller's responsibility
2906  * to make sure block parameter starts at a cluster boundary.
2907  */
2908 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2909                                       unsigned int num)
2910 {
2911         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2912                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2913         else {
2914                 unsigned int i;
2915
2916                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2917                         mark_block_used(ctx, block + i);
2918         }
2919 }
2920
2921 /*
2922  * Adjust the extended attribute block's reference counts at the end
2923  * of pass 1, either by subtracting out references for EA blocks that
2924  * are still referenced in ctx->refcount, or by adding references for
2925  * EA blocks that had extra references as accounted for in
2926  * ctx->refcount_extra.
2927  */
2928 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2929                                     char *block_buf, int adjust_sign)
2930 {
2931         struct ext2_ext_attr_header     *header;
2932         struct problem_context          pctx;
2933         ext2_filsys                     fs = ctx->fs;
2934         blk64_t                         blk;
2935         __u32                           should_be;
2936         ea_value_t                      count;
2937
2938         clear_problem_context(&pctx);
2939
2940         ea_refcount_intr_begin(refcount);
2941         while (1) {
2942                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2943                         break;
2944                 pctx.blk = blk;
2945                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2946                                                      pctx.ino);
2947                 /* We already checked this block, shouldn't happen */
2948                 if (pctx.errcode) {
2949                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2950                         return;
2951                 }
2952                 header = BHDR(block_buf);
2953                 if (header->h_magic != EXT2_EXT_ATTR_MAGIC) {
2954                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2955                         return;
2956                 }
2957
2958                 pctx.blkcount = header->h_refcount;
2959                 should_be = header->h_refcount + adjust_sign * (int)count;
2960                 pctx.num = should_be;
2961                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2962                         header->h_refcount = should_be;
2963                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2964                                                              block_buf,
2965                                                              pctx.ino);
2966                         if (pctx.errcode) {
2967                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2968                                             &pctx);
2969                                 continue;
2970                         }
2971                 }
2972         }
2973 }
2974
2975 /*
2976  * Handle processing the extended attribute blocks
2977  */
2978 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2979                            char *block_buf, struct ea_quota *ea_block_quota)
2980 {
2981         ext2_filsys fs = ctx->fs;
2982         ext2_ino_t      ino = pctx->ino;
2983         struct ext2_inode *inode = pctx->inode;
2984         blk64_t         blk;
2985         char *          end;
2986         struct ext2_ext_attr_header *header;
2987         struct ext2_ext_attr_entry *first, *entry;
2988         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2989         __u64           quota_inodes = 0;
2990         region_t        region = 0;
2991         int             failed_csum = 0;
2992
2993         ea_block_quota->blocks = 0;
2994         ea_block_quota->inodes = 0;
2995
2996         blk = ext2fs_file_acl_block(fs, inode);
2997         if (blk == 0)
2998                 return 0;
2999
3000         /*
3001          * If the Extended attribute flag isn't set, then a non-zero
3002          * file acl means that the inode is corrupted.
3003          *
3004          * Or if the extended attribute block is an invalid block,
3005          * then the inode is also corrupted.
3006          */
3007         if (!ext2fs_has_feature_xattr(fs->super) ||
3008             (blk < fs->super->s_first_data_block) ||
3009             (blk >= ext2fs_blocks_count(fs->super))) {
3010                 /* Fixed in pass2, e2fsck_process_bad_inode(). */
3011                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
3012                 return 0;
3013         }
3014
3015         /* If ea bitmap hasn't been allocated, create it */
3016         if (!ctx->block_ea_map) {
3017                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
3018                                         _("ext attr block map"),
3019                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
3020                                         &ctx->block_ea_map);
3021                 if (pctx->errcode) {
3022                         pctx->num = 2;
3023                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3024                         ctx->flags |= E2F_FLAG_ABORT;
3025                         return 0;
3026                 }
3027         }
3028
3029         /* Create the EA refcount structure if necessary */
3030         if (!ctx->refcount) {
3031                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3032                 if (pctx->errcode) {
3033                         pctx->num = 1;
3034                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3035                         ctx->flags |= E2F_FLAG_ABORT;
3036                         return 0;
3037                 }
3038         }
3039
3040 #if 0
3041         /* Debugging text */
3042         printf("Inode %u has EA block %u\n", ino, blk);
3043 #endif
3044
3045         /* Have we seen this EA block before? */
3046         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3047                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3048                 ea_block_quota->inodes = 0;
3049
3050                 if (ctx->ea_block_quota_blocks) {
3051                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3052                                           &quota_blocks);
3053                         if (quota_blocks)
3054                                 ea_block_quota->blocks = quota_blocks;
3055                 }
3056
3057                 if (ctx->ea_block_quota_inodes)
3058                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3059                                           &ea_block_quota->inodes);
3060
3061                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3062                         return 1;
3063                 /* Ooops, this EA was referenced more than it stated */
3064                 if (!ctx->refcount_extra) {
3065                         pctx->errcode = ea_refcount_create(0,
3066                                            &ctx->refcount_extra);
3067                         if (pctx->errcode) {
3068                                 pctx->num = 2;
3069                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3070                                 ctx->flags |= E2F_FLAG_ABORT;
3071                                 return 0;
3072                         }
3073                 }
3074                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3075                 return 1;
3076         }
3077
3078         /*
3079          * OK, we haven't seen this EA block yet.  So we need to
3080          * validate it
3081          */
3082         pctx->blk = blk;
3083         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3084         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3085                 pctx->errcode = 0;
3086                 failed_csum = 1;
3087         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3088                 pctx->errcode = 0;
3089
3090         if (pctx->errcode &&
3091             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3092                 pctx->errcode = 0;
3093                 goto clear_extattr;
3094         }
3095         header = BHDR(block_buf);
3096         pctx->blk = ext2fs_file_acl_block(fs, inode);
3097         if (((ctx->ext_attr_ver == 1) &&
3098              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3099             ((ctx->ext_attr_ver == 2) &&
3100              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3101                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3102                         goto clear_extattr;
3103         }
3104
3105         if (header->h_blocks != 1) {
3106                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3107                         goto clear_extattr;
3108         }
3109
3110         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3111                 goto clear_extattr;
3112
3113         region = region_create(0, fs->blocksize);
3114         if (!region) {
3115                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3116                 ctx->flags |= E2F_FLAG_ABORT;
3117                 return 0;
3118         }
3119         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3120                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3121                         goto clear_extattr;
3122         }
3123
3124         first = (struct ext2_ext_attr_entry *)(header+1);
3125         end = block_buf + fs->blocksize;
3126         entry = first;
3127         while ((char *)entry < end && *(__u32 *)entry) {
3128                 __u32 hash;
3129
3130                 if (region_allocate(region, (char *)entry - (char *)header,
3131                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3132                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3133                                 goto clear_extattr;
3134                         break;
3135                 }
3136                 if ((ctx->ext_attr_ver == 1 &&
3137                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3138                     (ctx->ext_attr_ver == 2 &&
3139                      entry->e_name_index == 0)) {
3140                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3141                                 goto clear_extattr;
3142                         break;
3143                 }
3144                 if (entry->e_value_inum == 0) {
3145                         if (entry->e_value_offs + entry->e_value_size >
3146                             fs->blocksize) {
3147                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3148                                         goto clear_extattr;
3149                                 break;
3150                         }
3151                         if (entry->e_value_size &&
3152                             region_allocate(region, entry->e_value_offs,
3153                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3154                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3155                                                 pctx))
3156                                         goto clear_extattr;
3157                         }
3158
3159                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3160                                                           entry->e_value_offs);
3161
3162                         if (entry->e_hash != hash) {
3163                                 pctx->num = entry->e_hash;
3164                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3165                                         goto clear_extattr;
3166                                 entry->e_hash = hash;
3167                         }
3168                 } else {
3169                         problem_t problem;
3170                         blk64_t entry_quota_blocks;
3171
3172                         problem = check_large_ea_inode(ctx, entry, pctx,
3173                                                        &entry_quota_blocks);
3174                         if (problem && fix_problem(ctx, problem, pctx))
3175                                 goto clear_extattr;
3176
3177                         quota_blocks += entry_quota_blocks;
3178                         quota_inodes++;
3179                 }
3180
3181                 entry = EXT2_EXT_ATTR_NEXT(entry);
3182         }
3183         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3184                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3185                         goto clear_extattr;
3186         }
3187         region_free(region);
3188
3189         /*
3190          * We only get here if there was no other errors that were fixed.
3191          * If there was a checksum fail, ask to correct it.
3192          */
3193         if (failed_csum &&
3194             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3195                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3196                                                        pctx->ino);
3197                 if (pctx->errcode)
3198                         return 0;
3199         }
3200
3201         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3202                 if (!ctx->ea_block_quota_blocks) {
3203                         pctx->errcode = ea_refcount_create(0,
3204                                                 &ctx->ea_block_quota_blocks);
3205                         if (pctx->errcode) {
3206                                 pctx->num = 3;
3207                                 goto refcount_fail;
3208                         }
3209                 }
3210                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3211                                   quota_blocks);
3212         }
3213
3214         if (quota_inodes) {
3215                 if (!ctx->ea_block_quota_inodes) {
3216                         pctx->errcode = ea_refcount_create(0,
3217                                                 &ctx->ea_block_quota_inodes);
3218                         if (pctx->errcode) {
3219                                 pctx->num = 4;
3220 refcount_fail:
3221                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3222                                 ctx->flags |= E2F_FLAG_ABORT;
3223                                 return 0;
3224                         }
3225                 }
3226
3227                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3228                                   quota_inodes);
3229         }
3230         ea_block_quota->blocks = quota_blocks;
3231         ea_block_quota->inodes = quota_inodes;
3232
3233         inc_ea_inode_refs(ctx, pctx, first, end);
3234         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3235         mark_block_used(ctx, blk);
3236         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3237         return 1;
3238
3239 clear_extattr:
3240         if (region)
3241                 region_free(region);
3242         ext2fs_file_acl_block_set(fs, inode, 0);
3243         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3244         return 0;
3245 }
3246
3247 /* Returns 1 if bad htree, 0 if OK */
3248 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3249                         ext2_ino_t ino, struct ext2_inode *inode,
3250                         char *block_buf)
3251 {
3252         struct ext2_dx_root_info        *root;
3253         ext2_filsys                     fs = ctx->fs;
3254         errcode_t                       retval;
3255         blk64_t                         blk;
3256
3257         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3258              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3259             (!ext2fs_has_feature_dir_index(fs->super))) {
3260                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
3261                 if (fix_problem(ctx, PR_1_HTREE_SET, pctx))
3262                         return 1;
3263         }
3264
3265         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3266
3267         if ((pctx->errcode) ||
3268             (blk == 0) ||
3269             (blk < fs->super->s_first_data_block) ||
3270             (blk >= ext2fs_blocks_count(fs->super))) {
3271                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
3272                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3273                         return 1;
3274                 else
3275                         return 0;
3276         }
3277
3278         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3279         if (retval) {
3280                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
3281                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3282                         return 1;
3283         }
3284
3285         /* XXX should check that beginning matches a directory */
3286         root = get_ext2_dx_root_info(fs, block_buf);
3287
3288         if ((root->reserved_zero || root->info_length < 8) &&
3289             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3290                 return 1;
3291
3292         pctx->num = root->hash_version;
3293         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3294             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3295             (root->hash_version != EXT2_HASH_TEA) &&
3296             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3297                 return 1;
3298
3299         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3300             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3301                 return 1;
3302
3303         pctx->num = root->indirect_levels;
3304         if ((root->indirect_levels > ext2_dir_htree_level(fs)) &&
3305             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3306                 return 1;
3307
3308         return 0;
3309 }
3310
3311 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3312                         struct ext2_inode *inode, int restart_flag,
3313                         const char *source)
3314 {
3315         inode->i_flags = 0;
3316         inode->i_links_count = 0;
3317         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3318         inode->i_dtime = ctx->now;
3319
3320         /*
3321          * If a special inode has such rotten block mappings that we
3322          * want to clear the whole inode, be sure to actually zap
3323          * the block maps because i_links_count isn't checked for
3324          * special inodes, and we'll end up right back here the next
3325          * time we run fsck.
3326          */
3327         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3328                 memset(inode->i_block, 0, sizeof(inode->i_block));
3329
3330         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3331         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3332         if (ctx->inode_reg_map)
3333                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3334         if (ctx->inode_badness)
3335                 ext2fs_icount_store(ctx->inode_badness, ino, 0);
3336
3337         /*
3338          * If the inode was partially accounted for before processing
3339          * was aborted, we need to restart the pass 1 scan.
3340          */
3341         ctx->flags |= restart_flag;
3342
3343         if (ino == EXT2_BAD_INO)
3344                 memset(inode, 0, sizeof(struct ext2_inode));
3345
3346         e2fsck_write_inode(ctx, ino, inode, source);
3347 }
3348
3349 /*
3350  * Use the multiple-blocks reclamation code to fix alignment problems in
3351  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3352  * physical cluster, and we want the block offsets within that cluster to
3353  * line up.
3354  */
3355 static int has_unaligned_cluster_map(e2fsck_t ctx,
3356                                      blk64_t last_pblk, blk64_t last_lblk,
3357                                      blk64_t pblk, blk64_t lblk)
3358 {
3359         blk64_t cluster_mask;
3360
3361         if (!ctx->fs->cluster_ratio_bits)
3362                 return 0;
3363         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3364
3365         /*
3366          * If the block in the logical cluster doesn't align with the block in
3367          * the physical cluster...
3368          */
3369         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3370                 return 1;
3371
3372         /*
3373          * If we cross a physical cluster boundary within a logical cluster...
3374          */
3375         if (last_pblk && (lblk & cluster_mask) != 0 &&
3376             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3377             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3378                 return 1;
3379
3380         return 0;
3381 }
3382
3383 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3384                              struct process_block_struct *pb,
3385                              blk64_t start_block, blk64_t end_block,
3386                              blk64_t eof_block,
3387                              ext2_extent_handle_t ehandle,
3388                              int try_repairs)
3389 {
3390         struct ext2fs_extent    extent;
3391         blk64_t                 blk, last_lblk;
3392         unsigned int            i, n;
3393         int                     is_dir, is_leaf;
3394         problem_t               problem;
3395         struct ext2_extent_info info;
3396         int                     failed_csum = 0;
3397
3398         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3399                 failed_csum = 1;
3400
3401         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3402         if (pctx->errcode)
3403                 return;
3404         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3405             !pb->eti.force_rebuild) {
3406                 struct extent_tree_level *etl;
3407
3408                 etl = pb->eti.ext_info + info.curr_level;
3409                 etl->num_extents += info.num_entries;
3410                 etl->max_extents += info.max_entries;
3411                 /*
3412                  * Implementation wart: Splitting extent blocks when appending
3413                  * will leave the old block with one free entry.  Therefore
3414                  * unless the node is totally full, pretend that a non-root
3415                  * extent block can hold one fewer entry than it actually does,
3416                  * so that we don't repeatedly rebuild the extent tree.
3417                  */
3418                 if (info.curr_level && info.num_entries < info.max_entries)
3419                         etl->max_extents--;
3420         }
3421
3422         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3423                                           &extent);
3424         while ((pctx->errcode == 0 ||
3425                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3426                info.num_entries-- > 0) {
3427                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3428                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3429                 last_lblk = extent.e_lblk + extent.e_len - 1;
3430
3431                 problem = 0;
3432                 pctx->blk = extent.e_pblk;
3433                 pctx->blk2 = extent.e_lblk;
3434                 pctx->num = extent.e_len;
3435                 pctx->blkcount = extent.e_lblk + extent.e_len;
3436
3437                 if (extent.e_pblk == 0 ||
3438                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3439                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3440                         problem = PR_1_EXTENT_BAD_START_BLK;
3441                 else if (extent.e_lblk < start_block)
3442                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3443                 else if ((end_block && last_lblk > end_block) &&
3444                          !(last_lblk > eof_block &&
3445                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3446                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3447                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3448                 else if (is_leaf && extent.e_len == 0)
3449                         problem = PR_1_EXTENT_LENGTH_ZERO;
3450                 else if (is_leaf &&
3451                          (extent.e_pblk + extent.e_len) >
3452                          ext2fs_blocks_count(ctx->fs->super))
3453                         problem = PR_1_EXTENT_ENDS_BEYOND;
3454                 else if (is_leaf && is_dir &&
3455                          ((extent.e_lblk + extent.e_len) >
3456                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3457                         problem = PR_1_TOOBIG_DIR;
3458
3459                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3460 #if 0
3461                         printf("extent_region(ino=%u, expect=%llu, "
3462                                "lblk=%llu, len=%u)\n",
3463                                pb->ino, pb->next_lblock,
3464                                extent.e_lblk, extent.e_len);
3465 #endif
3466                         if (extent.e_lblk < pb->next_lblock)
3467                                 problem = PR_1_EXTENT_COLLISION;
3468                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3469                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3470                 }
3471
3472                 /*
3473                  * Uninitialized blocks in a directory?  Clear the flag and
3474                  * we'll interpret the blocks later.
3475                  */
3476                 if (try_repairs && is_dir && problem == 0 &&
3477                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3478                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3479                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3480                         pb->inode_modified = 1;
3481                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3482                                                               &extent);
3483                         if (pctx->errcode)
3484                                 return;
3485                         failed_csum = 0;
3486                 }
3487
3488                 if (try_repairs && problem) {
3489                         /* To ensure that extent is in inode */
3490                         if (info.curr_level == 0)
3491                                 e2fsck_mark_inode_bad(ctx, pctx->ino,
3492                                                       BADNESS_HIGH);
3493 report_problem:
3494                         if (fix_problem(ctx, problem, pctx)) {
3495                                 if (ctx->invalid_bitmaps) {
3496                                         /*
3497                                          * If fsck knows the bitmaps are bad,
3498                                          * skip to the next extent and
3499                                          * try to clear this extent again
3500                                          * after fixing the bitmaps, by
3501                                          * restarting fsck.
3502                                          */
3503                                         pctx->errcode = ext2fs_extent_get(
3504                                                           ehandle,
3505                                                           EXT2_EXTENT_NEXT_SIB,
3506                                                           &extent);
3507                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3508                                         if (pctx->errcode ==
3509                                                     EXT2_ET_NO_CURRENT_NODE) {
3510                                                 pctx->errcode = 0;
3511                                                 break;
3512                                         }
3513                                         continue;
3514                                 }
3515                                 e2fsck_read_bitmaps(ctx);
3516                                 pb->inode_modified = 1;
3517                                 pctx->errcode =
3518                                         ext2fs_extent_delete(ehandle, 0);
3519                                 if (pctx->errcode) {
3520                                         pctx->str = "ext2fs_extent_delete";
3521                                         return;
3522                                 }
3523                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3524                                 if (pctx->errcode &&
3525                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3526                                         pctx->str = "ext2fs_extent_fix_parents";
3527                                         return;
3528                                 }
3529                                 pctx->errcode = ext2fs_extent_get(ehandle,
3530                                                                   EXT2_EXTENT_CURRENT,
3531                                                                   &extent);
3532                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3533                                         pctx->errcode = 0;
3534                                         break;
3535                                 }
3536                                 failed_csum = 0;
3537                                 continue;
3538                         }
3539                         goto next;
3540                 }
3541
3542                 if (!is_leaf) {
3543                         blk64_t lblk = extent.e_lblk;
3544                         int next_try_repairs = 1;
3545
3546                         blk = extent.e_pblk;
3547
3548                         /*
3549                          * If this lower extent block collides with critical
3550                          * metadata, don't try to repair the damage.  Pass 1b
3551                          * will reallocate the block; then we can try again.
3552                          */
3553                         if (pb->ino != EXT2_RESIZE_INO &&
3554                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3555                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3556                                                       extent.e_pblk)) {
3557                                 next_try_repairs = 0;
3558                                 pctx->blk = blk;
3559                                 fix_problem(ctx,
3560                                             PR_1_CRITICAL_METADATA_COLLISION,
3561                                             pctx);
3562                                 if ((ctx->options & E2F_OPT_NO) == 0)
3563                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3564                         }
3565                         pctx->errcode = ext2fs_extent_get(ehandle,
3566                                                   EXT2_EXTENT_DOWN, &extent);
3567                         if (pctx->errcode &&
3568                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3569                                 pctx->str = "EXT2_EXTENT_DOWN";
3570                                 problem = PR_1_EXTENT_HEADER_INVALID;
3571                                 if (!next_try_repairs)
3572                                         return;
3573                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3574                                         goto report_problem;
3575                                 return;
3576                         }
3577                         /* The next extent should match this index's logical start */
3578                         if (extent.e_lblk != lblk) {
3579                                 struct ext2_extent_info e_info;
3580
3581                                 ext2fs_extent_get_info(ehandle, &e_info);
3582                                 pctx->blk = lblk;
3583                                 pctx->blk2 = extent.e_lblk;
3584                                 pctx->num = e_info.curr_level - 1;
3585                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3586                                 if (fix_problem(ctx, problem, pctx)) {
3587                                         pb->inode_modified = 1;
3588                                         pctx->errcode =
3589                                                 ext2fs_extent_fix_parents(ehandle);
3590                                         if (pctx->errcode) {
3591                                                 pctx->str = "ext2fs_extent_fix_parents";
3592                                                 return;
3593                                         }
3594                                 }
3595                         }
3596                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3597                                          last_lblk, eof_block, ehandle,
3598                                          next_try_repairs);
3599                         if (pctx->errcode)
3600                                 return;
3601                         pctx->errcode = ext2fs_extent_get(ehandle,
3602                                                   EXT2_EXTENT_UP, &extent);
3603                         if (pctx->errcode) {
3604                                 pctx->str = "EXT2_EXTENT_UP";
3605                                 return;
3606                         }
3607                         mark_block_used(ctx, blk);
3608                         pb->num_blocks++;
3609                         goto next;
3610                 }
3611
3612                 if ((pb->previous_block != 0) &&
3613                     (pb->previous_block+1 != extent.e_pblk)) {
3614                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3615                                 char type = '?';
3616
3617                                 if (pb->is_dir)
3618                                         type = 'd';
3619                                 else if (pb->is_reg)
3620                                         type = 'f';
3621
3622                                 printf(("%6lu(%c): expecting %6lu "
3623                                         "actual extent "
3624                                         "phys %6lu log %lu len %lu\n"),
3625                                        (unsigned long) pctx->ino, type,
3626                                        (unsigned long) pb->previous_block+1,
3627                                        (unsigned long) extent.e_pblk,
3628                                        (unsigned long) extent.e_lblk,
3629                                        (unsigned long) extent.e_len);
3630                         }
3631                         pb->fragmented = 1;
3632                 }
3633                 /*
3634                  * If we notice a gap in the logical block mappings of an
3635                  * extent-mapped directory, offer to close the hole by
3636                  * moving the logical block down, otherwise we'll go mad in
3637                  * pass 3 allocating empty directory blocks to fill the hole.
3638                  */
3639                 if (try_repairs && is_dir &&
3640                     pb->last_block + 1 < extent.e_lblk) {
3641                         blk64_t new_lblk;
3642
3643                         new_lblk = pb->last_block + 1;
3644                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3645                                 new_lblk = ((new_lblk +
3646                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3647                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3648                                            (extent.e_pblk &
3649                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3650                         pctx->blk = extent.e_lblk;
3651                         pctx->blk2 = new_lblk;
3652                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3653                                 extent.e_lblk = new_lblk;
3654                                 pb->inode_modified = 1;
3655                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3656                                                                 0, &extent);
3657                                 if (pctx->errcode) {
3658                                         pctx->errcode = 0;
3659                                         goto alloc_later;
3660                                 }
3661                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3662                                 if (pctx->errcode)
3663                                         goto failed_add_dir_block;
3664                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3665                                                                 extent.e_lblk);
3666                                 if (pctx->errcode)
3667                                         goto failed_add_dir_block;
3668                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3669                                 failed_csum = 0;
3670                         }
3671                 }
3672 alloc_later:
3673                 if (is_dir) {
3674                         while (++pb->last_db_block <
3675                                (e2_blkcnt_t) extent.e_lblk) {
3676                                 pctx->errcode = ext2fs_add_dir_block2(
3677                                                         ctx->fs->dblist,
3678                                                         pb->ino, 0,
3679                                                         pb->last_db_block);
3680                                 if (pctx->errcode) {
3681                                         pctx->blk = 0;
3682                                         pctx->num = pb->last_db_block;
3683                                         goto failed_add_dir_block;
3684                                 }
3685                         }
3686
3687                         for (i = 0; i < extent.e_len; i++) {
3688                                 pctx->errcode = ext2fs_add_dir_block2(
3689                                                         ctx->fs->dblist,
3690                                                         pctx->ino,
3691                                                         extent.e_pblk + i,
3692                                                         extent.e_lblk + i);
3693                                 if (pctx->errcode) {
3694                                         pctx->blk = extent.e_pblk + i;
3695                                         pctx->num = extent.e_lblk + i;
3696                                 failed_add_dir_block:
3697                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3698                                         /* Should never get here */
3699                                         ctx->flags |= E2F_FLAG_ABORT;
3700                                         return;
3701                                 }
3702                         }
3703                         if (extent.e_len > 0)
3704                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3705                 }
3706                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3707                                               pb->last_block,
3708                                               extent.e_pblk,
3709                                               extent.e_lblk)) {
3710                         for (i = 0; i < extent.e_len; i++) {
3711                                 pctx->blk = extent.e_lblk + i;
3712                                 pctx->blk2 = extent.e_pblk + i;
3713                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3714                                 mark_block_used(ctx, extent.e_pblk + i);
3715                                 mark_block_used(ctx, extent.e_pblk + i);
3716                         }
3717                 }
3718
3719                 /*
3720                  * Check whether first cluster got marked in previous iteration.
3721                  */
3722                 if (ctx->fs->cluster_ratio_bits &&
3723                     pb->previous_block &&
3724                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3725                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3726                         /* Set blk to the beginning of next cluster. */
3727                         blk = EXT2FS_C2B(
3728                                 ctx->fs,
3729                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3730                 else
3731                         /* Set blk to the beginning of current cluster. */
3732                         blk = EXT2FS_C2B(ctx->fs,
3733                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3734
3735                 if (blk < extent.e_pblk + extent.e_len) {
3736                         mark_blocks_used(ctx, blk,
3737                                          extent.e_pblk + extent.e_len - blk);
3738                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3739                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3740                         pb->num_blocks += n;
3741                 }
3742                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3743                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3744                 start_block = pb->last_block = last_lblk;
3745                 if (is_leaf && !is_dir &&
3746                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3747                         pb->last_init_lblock = last_lblk;
3748         next:
3749                 pctx->errcode = ext2fs_extent_get(ehandle,
3750                                                   EXT2_EXTENT_NEXT_SIB,
3751                                                   &extent);
3752         }
3753
3754         /* Failed csum but passes checks?  Ask to fix checksum. */
3755         if (failed_csum &&
3756             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3757                 pb->inode_modified = 1;
3758                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3759                 if (pctx->errcode)
3760                         return;
3761         }
3762
3763         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3764                 pctx->errcode = 0;
3765 }
3766
3767 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3768                                  struct process_block_struct *pb)
3769 {
3770         struct ext2_extent_info info;
3771         struct ext2_inode       *inode = pctx->inode;
3772         ext2_extent_handle_t    ehandle;
3773         ext2_filsys             fs = ctx->fs;
3774         ext2_ino_t              ino = pctx->ino;
3775         errcode_t               retval;
3776         blk64_t                 eof_lblk;
3777         struct ext3_extent_header       *eh;
3778
3779         /* Check for a proper extent header... */
3780         eh = (struct ext3_extent_header *) &inode->i_block[0];
3781         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3782         if (retval) {
3783                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3784                         e2fsck_clear_inode(ctx, ino, inode, 0,
3785                                            "check_blocks_extents");
3786                 pctx->errcode = 0;
3787                 return;
3788         }
3789
3790         /* ...since this function doesn't fail if i_block is zeroed. */
3791         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3792         if (pctx->errcode) {
3793                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3794                         e2fsck_clear_inode(ctx, ino, inode, 0,
3795                                            "check_blocks_extents");
3796                 pctx->errcode = 0;
3797                 return;
3798         }
3799
3800         retval = ext2fs_extent_get_info(ehandle, &info);
3801         if (retval == 0) {
3802                 int max_depth = info.max_depth;
3803
3804                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3805                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3806                 ctx->extent_depth_count[max_depth]++;
3807         }
3808
3809         /* Check maximum extent depth */
3810         pctx->blk = info.max_depth;
3811         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3812         if (pctx->blk2 < pctx->blk &&
3813             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3814                 pb->eti.force_rebuild = 1;
3815
3816         /* Can we collect extent tree level stats? */
3817         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3818         if (pctx->blk2 > pctx->blk)
3819                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3820         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3821         pb->eti.ino = pb->ino;
3822
3823         pb->next_lblock = 0;
3824
3825         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3826                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3827         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3828         if (pctx->errcode &&
3829             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3830                 pb->num_blocks = 0;
3831                 inode->i_blocks = 0;
3832                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3833                                    "check_blocks_extents");
3834                 pctx->errcode = 0;
3835         }
3836         ext2fs_extent_free(ehandle);
3837
3838         /* Rebuild unless it's a dir and we're rehashing it */
3839         if (LINUX_S_ISDIR(inode->i_mode) &&
3840             e2fsck_dir_will_be_rehashed(ctx, ino))
3841                 return;
3842
3843         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3844                 e2fsck_rebuild_extents_later(ctx, ino);
3845         else
3846                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3847 }
3848
3849 /*
3850  * In fact we don't need to check blocks for an inode with inline data
3851  * because this inode doesn't have any blocks.  In this function all
3852  * we need to do is add this inode into dblist when it is a directory.
3853  */
3854 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3855                                      struct process_block_struct *pb)
3856 {
3857         int     flags;
3858         size_t  inline_data_size = 0;
3859
3860         if (!pb->is_dir) {
3861                 pctx->errcode = 0;
3862                 return;
3863         }
3864
3865         /* Process the dirents in i_block[] as the "first" block. */
3866         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3867         if (pctx->errcode)
3868                 goto err;
3869
3870         /* Process the dirents in the EA as a "second" block. */
3871         flags = ctx->fs->flags;
3872         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3873         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3874                                                 &inline_data_size);
3875         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3876                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3877         if (pctx->errcode) {
3878                 pctx->errcode = 0;
3879                 return;
3880         }
3881
3882         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3883                 return;
3884
3885         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3886         if (pctx->errcode)
3887                 goto err;
3888
3889         return;
3890 err:
3891         pctx->blk = 0;
3892         pctx->num = 0;
3893         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3894         ctx->flags |= E2F_FLAG_ABORT;
3895 }
3896
3897 /*
3898  * This subroutine is called on each inode to account for all of the
3899  * blocks used by that inode.
3900  */
3901 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3902                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3903 {
3904         ext2_filsys fs = ctx->fs;
3905         struct process_block_struct pb;
3906         ext2_ino_t      ino = pctx->ino;
3907         struct ext2_inode *inode = pctx->inode;
3908         unsigned        bad_size = 0;
3909         int             dirty_inode = 0;
3910         int             extent_fs;
3911         int             inlinedata_fs;
3912         __u64           size;
3913         struct ea_quota ea_block_quota;
3914
3915         pb.ino = ino;
3916         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3917                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3918         pb.last_block = ~0;
3919         pb.last_init_lblock = -1;
3920         pb.last_db_block = -1;
3921         pb.num_illegal_blocks = 0;
3922         pb.suppress = 0; pb.clear = 0;
3923         pb.fragmented = 0;
3924         pb.compressed = 0;
3925         pb.previous_block = 0;
3926         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3927         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3928         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3929         pb.inode = inode;
3930         pb.pctx = pctx;
3931         pb.ctx = ctx;
3932         pb.inode_modified = 0;
3933         pb.eti.force_rebuild = 0;
3934         pctx->ino = ino;
3935         pctx->errcode = 0;
3936
3937         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3938         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3939
3940         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3941                 if (e2fsck_should_abort(ctx))
3942                         goto out;
3943                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3944         }
3945
3946         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3947                 check_blocks_inline_data(ctx, pctx, &pb);
3948         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3949                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3950                         check_blocks_extents(ctx, pctx, &pb);
3951                 else {
3952                         int flags;
3953                         /*
3954                          * If we've modified the inode, write it out before
3955                          * iterate() tries to use it.
3956                          */
3957                         if (dirty_inode) {
3958                                 e2fsck_write_inode(ctx, ino, inode,
3959                                                    "check_blocks");
3960                                 dirty_inode = 0;
3961                         }
3962                         flags = fs->flags;
3963                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3964                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3965                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3966                                                 block_buf, process_block, &pb);
3967                         /*
3968                          * We do not have uninitialized extents in non extent
3969                          * files.
3970                          */
3971                         pb.last_init_lblock = pb.last_block;
3972                         /*
3973                          * If iterate() changed a block mapping, we have to
3974                          * re-read the inode.  If we decide to clear the
3975                          * inode after clearing some stuff, we'll re-write the
3976                          * bad mappings into the inode!
3977                          */
3978                         if (pb.inode_modified)
3979                                 e2fsck_read_inode(ctx, ino, inode,
3980                                                   "check_blocks");
3981                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3982                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3983
3984                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3985 #ifdef DEBUG
3986                                 printf("bmap rebuild ino=%d\n", ino);
3987 #endif
3988                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3989                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3990                                         e2fsck_rebuild_extents_later(ctx, ino);
3991                         }
3992                 }
3993         }
3994         end_problem_latch(ctx, PR_LATCH_BLOCK);
3995         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3996         if (e2fsck_should_abort(ctx))
3997                 goto out;
3998         if (pctx->errcode)
3999                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4000
4001         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4002                 if (LINUX_S_ISDIR(inode->i_mode))
4003                         ctx->fs_fragmented_dir++;
4004                 else
4005                         ctx->fs_fragmented++;
4006         }
4007
4008         if (pb.clear) {
4009                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4010                                    "check_blocks");
4011                 return;
4012         }
4013
4014         if (inode->i_flags & EXT2_INDEX_FL) {
4015                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4016                         inode->i_flags &= ~EXT2_INDEX_FL;
4017                         dirty_inode++;
4018                 } else {
4019                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4020                 }
4021         }
4022
4023         if (!pb.num_blocks && pb.is_dir &&
4024             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4025                 /*
4026                  * The mode might be in-correct. Increasing the badness by
4027                  * small amount won't hurt much.
4028                  */
4029                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
4030                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4031                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4032                         ctx->fs_directory_count--;
4033                         return;
4034                 }
4035         }
4036
4037         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4038             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4039             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4040                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4041                                ino,
4042                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4043                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4044                                   ino, (ea_ibody_quota ?
4045                                         ea_ibody_quota->inodes : 0) +
4046                                                 ea_block_quota.inodes + 1);
4047         }
4048
4049         if (!ext2fs_has_feature_huge_file(fs->super) ||
4050             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4051                 pb.num_blocks *= (fs->blocksize / 512);
4052         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4053 #if 0
4054         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4055                ino, inode->i_size, pb.last_block, ext2fs_inode_i_blocks(fs, inode),
4056                pb.num_blocks);
4057 #endif
4058         if (pb.is_dir) {
4059                 unsigned nblock = inode->i_size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4060                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4061                         int flags;
4062                         size_t sz = 0;
4063                         errcode_t err;
4064
4065                         flags = ctx->fs->flags;
4066                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4067                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4068                                                       &sz);
4069                         ctx->fs->flags = (flags &
4070                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4071                                          (ctx->fs->flags &
4072                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4073                         if (err || sz != inode->i_size) {
4074                                 bad_size = 7;
4075                                 pctx->num = sz;
4076                         }
4077                 } else if (inode->i_size & (fs->blocksize - 1))
4078                         bad_size = 5;
4079                 else if (nblock > (pb.last_block + 1))
4080                         bad_size = 1;
4081                 else if (nblock < (pb.last_block + 1)) {
4082                         if (((pb.last_block + 1) - nblock) >
4083                             fs->super->s_prealloc_dir_blocks)
4084                                 bad_size = 2;
4085                 }
4086         } else {
4087                 size = EXT2_I_SIZE(inode);
4088                 if ((pb.last_init_lblock >= 0) &&
4089                     /* Do not allow initialized allocated blocks past i_size*/
4090                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4091                     !(inode->i_flags & EXT4_VERITY_FL))
4092                         bad_size = 3;
4093                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4094                          size > ext2_max_sizes[fs->super->s_log_block_size])
4095                         /* too big for a direct/indirect-mapped file */
4096                         bad_size = 4;
4097                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4098                          size >
4099                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4100                         /* too big for an extent-based file - 32bit ee_block */
4101                         bad_size = 6;
4102         }
4103         /* i_size for symlinks is checked elsewhere */
4104         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4105                 /* Did inline_data set pctx->num earlier? */
4106                 if (bad_size != 7)
4107                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4108                 pctx->group = bad_size;
4109                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
4110                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4111                         if (LINUX_S_ISDIR(inode->i_mode))
4112                                 pctx->num &= 0xFFFFFFFFULL;
4113                         ext2fs_inode_size_set(fs, inode, pctx->num);
4114                         if (EXT2_I_SIZE(inode) == 0 &&
4115                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4116                                 memset(inode->i_block, 0,
4117                                        sizeof(inode->i_block));
4118                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4119                         }
4120                         dirty_inode++;
4121                 }
4122                 pctx->num = 0;
4123         }
4124         if (LINUX_S_ISREG(inode->i_mode) &&
4125             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4126                 ctx->large_files++;
4127         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4128             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4129              (ext2fs_has_feature_huge_file(fs->super) &&
4130               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4131               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4132                 pctx->num = pb.num_blocks;
4133                 e2fsck_mark_inode_bad(ctx, ino, BADNESS_NORMAL);
4134                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4135                         inode->i_blocks = pb.num_blocks;
4136                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4137                         dirty_inode++;
4138                 }
4139                 pctx->num = 0;
4140         }
4141
4142         /*
4143          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4144          * a block mapped file, so rebuild it as an extent file.  We can skip
4145          * symlinks because they're never rewritten.
4146          */
4147         if (ext2fs_has_feature_bigalloc(fs->super) &&
4148             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4149             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4150             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4151             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4152             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4153                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4154                 if (pctx->errcode)
4155                         goto out;
4156         }
4157
4158         if (ctx->dirs_to_hash && pb.is_dir &&
4159             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4160             !(inode->i_flags & EXT2_INDEX_FL) &&
4161             ((inode->i_size / fs->blocksize) >= 3))
4162                 e2fsck_rehash_dir_later(ctx, ino);
4163
4164 out:
4165         if (dirty_inode)
4166                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4167 }
4168
4169 #if 0
4170 /*
4171  * Helper function called by process block when an illegal block is
4172  * found.  It returns a description about why the block is illegal
4173  */
4174 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4175 {
4176         blk64_t super;
4177         int     i;
4178         static char     problem[80];
4179
4180         super = fs->super->s_first_data_block;
4181         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4182         if (block < super) {
4183                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4184                 return(problem);
4185         } else if (block >= ext2fs_blocks_count(fs->super)) {
4186                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4187                 return(problem);
4188         }
4189         for (i = 0; i < fs->group_desc_count; i++) {
4190                 if (block == super) {
4191                         sprintf(problem, "is the superblock in group %d", i);
4192                         break;
4193                 }
4194                 if (block > super &&
4195                     block <= (super + fs->desc_blocks)) {
4196                         sprintf(problem, "is in the group descriptors "
4197                                 "of group %d", i);
4198                         break;
4199                 }
4200                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4201                         sprintf(problem, "is the block bitmap of group %d", i);
4202                         break;
4203                 }
4204                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4205                         sprintf(problem, "is the inode bitmap of group %d", i);
4206                         break;
4207                 }
4208                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4209                     (block < ext2fs_inode_table_loc(fs, i)
4210                      + fs->inode_blocks_per_group)) {
4211                         sprintf(problem, "is in the inode table of group %d",
4212                                 i);
4213                         break;
4214                 }
4215                 super += fs->super->s_blocks_per_group;
4216         }
4217         return(problem);
4218 }
4219 #endif
4220
4221 /*
4222  * This is a helper function for check_blocks().
4223  */
4224 static int process_block(ext2_filsys fs,
4225                   blk64_t       *block_nr,
4226                   e2_blkcnt_t blockcnt,
4227                   blk64_t ref_block EXT2FS_ATTR((unused)),
4228                   int ref_offset EXT2FS_ATTR((unused)),
4229                   void *priv_data)
4230 {
4231         struct process_block_struct *p;
4232         struct problem_context *pctx;
4233         blk64_t blk = *block_nr;
4234         int     ret_code = 0;
4235         problem_t       problem = 0;
4236         e2fsck_t        ctx;
4237
4238         p = (struct process_block_struct *) priv_data;
4239         pctx = p->pctx;
4240         ctx = p->ctx;
4241
4242         /*
4243          * For a directory, add logical block zero for processing even if it's
4244          * not mapped or we'll be perennially stuck with broken "." and ".."
4245          * entries.
4246          */
4247         if (p->is_dir && blockcnt == 0 && blk == 0) {
4248                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4249                 if (pctx->errcode) {
4250                         pctx->blk = blk;
4251                         pctx->num = blockcnt;
4252                         goto failed_add_dir_block;
4253                 }
4254                 p->last_db_block++;
4255         }
4256
4257         if (blk == 0)
4258                 return 0;
4259
4260 #if 0
4261         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4262                blockcnt);
4263 #endif
4264
4265         /*
4266          * Simplistic fragmentation check.  We merely require that the
4267          * file be contiguous.  (Which can never be true for really
4268          * big files that are greater than a block group.)
4269          */
4270         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4271                 if (p->previous_block+1 != blk) {
4272                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4273                                 char type = '?';
4274
4275                                 if (p->is_dir)
4276                                         type = 'd';
4277                                 else if (p->is_reg)
4278                                         type = 'f';
4279
4280                                 printf(_("%6lu(%c): expecting %6lu "
4281                                          "got phys %6lu (blkcnt %lld)\n"),
4282                                        (unsigned long) pctx->ino, type,
4283                                        (unsigned long) p->previous_block+1,
4284                                        (unsigned long) blk,
4285                                        blockcnt);
4286                         }
4287                         p->fragmented = 1;
4288                 }
4289         }
4290
4291         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4292             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4293                 problem = PR_1_TOOBIG_DIR;
4294         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4295                 problem = PR_1_TOOBIG_DIR;
4296         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4297                 problem = PR_1_TOOBIG_REG;
4298         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4299                 problem = PR_1_TOOBIG_SYMLINK;
4300
4301         if (blk < fs->super->s_first_data_block ||
4302             blk >= ext2fs_blocks_count(fs->super)) {
4303                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4304                 e2fsck_mark_inode_bad(ctx, pctx->ino, BADNESS_NORMAL);
4305         }
4306
4307         /*
4308          * If this IND/DIND/TIND block is squatting atop some critical metadata
4309          * (group descriptors, superblock, bitmap, inode table), any write to
4310          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4311          * fix that and restart fsck.
4312          */
4313         if (blockcnt < 0 &&
4314             p->ino != EXT2_RESIZE_INO &&
4315             blk < ctx->fs->super->s_blocks_count &&
4316             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4317                 pctx->blk = blk;
4318                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4319                 if ((ctx->options & E2F_OPT_NO) == 0)
4320                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4321         }
4322
4323         if (problem) {
4324                 p->num_illegal_blocks++;
4325                 /*
4326                  * A bit of subterfuge here -- we're trying to fix a block
4327                  * mapping, but the IND/DIND/TIND block could have collided
4328                  * with some critical metadata.  So, fix the in-core mapping so
4329                  * iterate won't go insane, but return 0 instead of
4330                  * BLOCK_CHANGED so that it won't write the remapping out to
4331                  * our multiply linked block.
4332                  *
4333                  * Even if we previously determined that an *IND block
4334                  * conflicts with critical metadata, we must still try to
4335                  * iterate the *IND block as if it is an *IND block to find and
4336                  * mark the blocks it points to.  Better to be overly cautious
4337                  * with the used_blocks map so that we don't move the *IND
4338                  * block to a block that's really in use!
4339                  */
4340                 if (p->ino != EXT2_RESIZE_INO &&
4341                     ref_block != 0 &&
4342                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4343                                               ref_block)) {
4344                         *block_nr = 0;
4345                         return 0;
4346                 }
4347                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4348                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4349                                 p->clear = 1;
4350                                 return BLOCK_ABORT;
4351                         }
4352                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4353                                 p->suppress = 1;
4354                                 set_latch_flags(PR_LATCH_BLOCK,
4355                                                 PRL_SUPPRESS, 0);
4356                         }
4357                 }
4358                 pctx->blk = blk;
4359                 pctx->blkcount = blockcnt;
4360                 if (fix_problem(ctx, problem, pctx)) {
4361                         blk = *block_nr = 0;
4362                         ret_code = BLOCK_CHANGED;
4363                         p->inode_modified = 1;
4364                         /*
4365                          * If the directory block is too big and is beyond the
4366                          * end of the FS, don't bother trying to add it for
4367                          * processing -- the kernel would never have created a
4368                          * directory this large, and we risk an ENOMEM abort.
4369                          * In any case, the toobig handler for extent-based
4370                          * directories also doesn't feed toobig blocks to
4371                          * pass 2.
4372                          */
4373                         if (problem == PR_1_TOOBIG_DIR)
4374                                 return ret_code;
4375                         goto mark_dir;
4376                 } else
4377                         return 0;
4378         }
4379
4380         if (p->ino == EXT2_RESIZE_INO) {
4381                 /*
4382                  * The resize inode has already be sanity checked
4383                  * during pass #0 (the superblock checks).  All we
4384                  * have to do is mark the double indirect block as
4385                  * being in use; all of the other blocks are handled
4386                  * by mark_table_blocks()).
4387                  */
4388                 if (blockcnt == BLOCK_COUNT_DIND)
4389                         mark_block_used(ctx, blk);
4390                 p->num_blocks++;
4391         } else if (!(ctx->fs->cluster_ratio_bits &&
4392                      p->previous_block &&
4393                      (EXT2FS_B2C(ctx->fs, blk) ==
4394                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4395                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4396                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4397                 mark_block_used(ctx, blk);
4398                 p->num_blocks++;
4399         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4400                                              p->last_block, blk, blockcnt)) {
4401                 pctx->blk = blockcnt;
4402                 pctx->blk2 = blk;
4403                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4404                 mark_block_used(ctx, blk);
4405                 mark_block_used(ctx, blk);
4406         }
4407         if (blockcnt >= 0)
4408                 p->last_block = blockcnt;
4409         p->previous_block = blk;
4410 mark_dir:
4411         if (p->is_dir && (blockcnt >= 0)) {
4412                 while (++p->last_db_block < blockcnt) {
4413                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4414                                                               p->ino, 0,
4415                                                               p->last_db_block);
4416                         if (pctx->errcode) {
4417                                 pctx->blk = 0;
4418                                 pctx->num = p->last_db_block;
4419                                 goto failed_add_dir_block;
4420                         }
4421                 }
4422                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4423                                                       blk, blockcnt);
4424                 if (pctx->errcode) {
4425                         pctx->blk = blk;
4426                         pctx->num = blockcnt;
4427                 failed_add_dir_block:
4428                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4429                         /* Should never get here */
4430                         ctx->flags |= E2F_FLAG_ABORT;
4431                         return BLOCK_ABORT;
4432                 }
4433         }
4434         return ret_code;
4435 }
4436
4437 static int process_bad_block(ext2_filsys fs,
4438                       blk64_t *block_nr,
4439                       e2_blkcnt_t blockcnt,
4440                       blk64_t ref_block EXT2FS_ATTR((unused)),
4441                       int ref_offset EXT2FS_ATTR((unused)),
4442                       void *priv_data)
4443 {
4444         struct process_block_struct *p;
4445         blk64_t         blk = *block_nr;
4446         blk64_t         first_block;
4447         dgrp_t          i;
4448         struct problem_context *pctx;
4449         e2fsck_t        ctx;
4450
4451         if (!blk)
4452                 return 0;
4453
4454         p = (struct process_block_struct *) priv_data;
4455         ctx = p->ctx;
4456         pctx = p->pctx;
4457
4458         pctx->ino = EXT2_BAD_INO;
4459         pctx->blk = blk;
4460         pctx->blkcount = blockcnt;
4461
4462         if ((blk < fs->super->s_first_data_block) ||
4463             (blk >= ext2fs_blocks_count(fs->super))) {
4464                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4465                         *block_nr = 0;
4466                         return BLOCK_CHANGED;
4467                 } else
4468                         return 0;
4469         }
4470
4471         if (blockcnt < 0) {
4472                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4473                         p->bbcheck = 1;
4474                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4475                                 *block_nr = 0;
4476                                 return BLOCK_CHANGED;
4477                         }
4478                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4479                                                     blk)) {
4480                         p->bbcheck = 1;
4481                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4482                                         pctx)) {
4483                                 *block_nr = 0;
4484                                 return BLOCK_CHANGED;
4485                         }
4486                         if (e2fsck_should_abort(ctx))
4487                                 return BLOCK_ABORT;
4488                 } else
4489                         mark_block_used(ctx, blk);
4490                 return 0;
4491         }
4492 #if 0
4493         printf ("DEBUG: Marking %u as bad.\n", blk);
4494 #endif
4495         ctx->fs_badblocks_count++;
4496         /*
4497          * If the block is not used, then mark it as used and return.
4498          * If it is already marked as found, this must mean that
4499          * there's an overlap between the filesystem table blocks
4500          * (bitmaps and inode table) and the bad block list.
4501          */
4502         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4503                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4504                 return 0;
4505         }
4506         /*
4507          * Try to find the where the filesystem block was used...
4508          */
4509         first_block = fs->super->s_first_data_block;
4510
4511         for (i = 0; i < fs->group_desc_count; i++ ) {
4512                 pctx->group = i;
4513                 pctx->blk = blk;
4514                 if (!ext2fs_bg_has_super(fs, i))
4515                         goto skip_super;
4516                 if (blk == first_block) {
4517                         if (i == 0) {
4518                                 if (fix_problem(ctx,
4519                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4520                                                 pctx)) {
4521                                         *block_nr = 0;
4522                                         return BLOCK_CHANGED;
4523                                 }
4524                                 return 0;
4525                         }
4526                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4527                         return 0;
4528                 }
4529                 if ((blk > first_block) &&
4530                     (blk <= first_block + fs->desc_blocks)) {
4531                         if (i == 0) {
4532                                 pctx->blk = *block_nr;
4533                                 if (fix_problem(ctx,
4534                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4535                                         *block_nr = 0;
4536                                         return BLOCK_CHANGED;
4537                                 }
4538                                 return 0;
4539                         }
4540                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4541                         return 0;
4542                 }
4543         skip_super:
4544                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4545                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4546                                 ctx->invalid_block_bitmap_flag[i]++;
4547                                 ctx->invalid_bitmaps++;
4548                         }
4549                         return 0;
4550                 }
4551                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4552                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4553                                 ctx->invalid_inode_bitmap_flag[i]++;
4554                                 ctx->invalid_bitmaps++;
4555                         }
4556                         return 0;
4557                 }
4558                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4559                     (blk < (ext2fs_inode_table_loc(fs, i) +
4560                             fs->inode_blocks_per_group))) {
4561                         /*
4562                          * If there are bad blocks in the inode table,
4563                          * the inode scan code will try to do
4564                          * something reasonable automatically.
4565                          */
4566                         return 0;
4567                 }
4568                 first_block += fs->super->s_blocks_per_group;
4569         }
4570         /*
4571          * If we've gotten to this point, then the only
4572          * possibility is that the bad block inode meta data
4573          * is using a bad block.
4574          */
4575         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4576             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4577             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4578                 p->bbcheck = 1;
4579                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4580                         *block_nr = 0;
4581                         return BLOCK_CHANGED;
4582                 }
4583                 if (e2fsck_should_abort(ctx))
4584                         return BLOCK_ABORT;
4585                 return 0;
4586         }
4587
4588         pctx->group = -1;
4589
4590         /* Warn user that the block wasn't claimed */
4591         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4592
4593         return 0;
4594 }
4595
4596 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4597                             const char *name, int num, blk64_t *new_block)
4598 {
4599         ext2_filsys fs = ctx->fs;
4600         dgrp_t          last_grp;
4601         blk64_t         old_block = *new_block;
4602         blk64_t         last_block;
4603         dgrp_t          flexbg;
4604         unsigned        flexbg_size;
4605         int             i, is_flexbg;
4606         char            *buf;
4607         struct problem_context  pctx;
4608
4609         clear_problem_context(&pctx);
4610
4611         pctx.group = group;
4612         pctx.blk = old_block;
4613         pctx.str = name;
4614
4615         /*
4616          * For flex_bg filesystems, first try to allocate the metadata
4617          * within the flex_bg, and if that fails then try finding the
4618          * space anywhere in the filesystem.
4619          */
4620         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4621         if (is_flexbg) {
4622                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
4623                 flexbg = group / flexbg_size;
4624                 first_block = ext2fs_group_first_block2(fs,
4625                                                         flexbg_size * flexbg);
4626                 last_grp = group | (flexbg_size - 1);
4627                 if (last_grp >= fs->group_desc_count)
4628                         last_grp = fs->group_desc_count - 1;
4629                 last_block = ext2fs_group_last_block2(fs, last_grp);
4630         } else
4631                 last_block = ext2fs_group_last_block2(fs, group);
4632         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4633                                                num, ctx->block_found_map,
4634                                                new_block);
4635         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4636                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4637                                 fs->super->s_first_data_block,
4638                                 ext2fs_blocks_count(fs->super),
4639                                 num, ctx->block_found_map, new_block);
4640         if (pctx.errcode) {
4641                 pctx.num = num;
4642                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4643                 ext2fs_unmark_valid(fs);
4644                 ctx->flags |= E2F_FLAG_ABORT;
4645                 return;
4646         }
4647         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4648         if (pctx.errcode) {
4649                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4650                 ext2fs_unmark_valid(fs);
4651                 ctx->flags |= E2F_FLAG_ABORT;
4652                 return;
4653         }
4654         ext2fs_mark_super_dirty(fs);
4655         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4656         pctx.blk2 = *new_block;
4657         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4658                           PR_1_RELOC_TO), &pctx);
4659         pctx.blk2 = 0;
4660         for (i = 0; i < num; i++) {
4661                 pctx.blk = i;
4662                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4663                 if (old_block) {
4664                         pctx.errcode = io_channel_read_blk64(fs->io,
4665                                    old_block + i, 1, buf);
4666                         if (pctx.errcode)
4667                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4668                         pctx.blk = (*new_block) + i;
4669                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4670                                                               1, buf);
4671                 } else {
4672                         pctx.blk = (*new_block) + i;
4673                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4674                                                            NULL, NULL);
4675                 }
4676
4677                 if (pctx.errcode)
4678                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4679         }
4680         ext2fs_free_mem(&buf);
4681 }
4682
4683 /*
4684  * This routine gets called at the end of pass 1 if bad blocks are
4685  * detected in the superblock, group descriptors, inode_bitmaps, or
4686  * block bitmaps.  At this point, all of the blocks have been mapped
4687  * out, so we can try to allocate new block(s) to replace the bad
4688  * blocks.
4689  */
4690 static void handle_fs_bad_blocks(e2fsck_t ctx)
4691 {
4692         ext2_filsys fs = ctx->fs;
4693         dgrp_t          i;
4694         blk64_t         first_block;
4695         blk64_t         new_blk;
4696
4697         for (i = 0; i < fs->group_desc_count; i++) {
4698                 first_block = ext2fs_group_first_block2(fs, i);
4699
4700                 if (ctx->invalid_block_bitmap_flag[i]) {
4701                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4702                         new_table_block(ctx, first_block, i, _("block bitmap"),
4703                                         1, &new_blk);
4704                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4705                 }
4706                 if (ctx->invalid_inode_bitmap_flag[i]) {
4707                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4708                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4709                                         1, &new_blk);
4710                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4711                 }
4712                 if (ctx->invalid_inode_table_flag[i]) {
4713                         new_blk = ext2fs_inode_table_loc(fs, i);
4714                         new_table_block(ctx, first_block, i, _("inode table"),
4715                                         fs->inode_blocks_per_group,
4716                                         &new_blk);
4717                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4718                         ctx->flags |= E2F_FLAG_RESTART;
4719                 }
4720         }
4721         ctx->invalid_bitmaps = 0;
4722 }
4723
4724 /*
4725  * This routine marks all blocks which are used by the superblock,
4726  * group descriptors, inode bitmaps, and block bitmaps.
4727  */
4728 static void mark_table_blocks(e2fsck_t ctx)
4729 {
4730         ext2_filsys fs = ctx->fs;
4731         blk64_t b;
4732         dgrp_t  i;
4733         unsigned int    j;
4734         struct problem_context pctx;
4735
4736         clear_problem_context(&pctx);
4737
4738         for (i = 0; i < fs->group_desc_count; i++) {
4739                 pctx.group = i;
4740
4741                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4742                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4743
4744                 /*
4745                  * Mark the blocks used for the inode table
4746                  */
4747                 if (ext2fs_inode_table_loc(fs, i)) {
4748                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4749                              j < fs->inode_blocks_per_group;
4750                              j++, b++) {
4751                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4752                                                              b)) {
4753                                         pctx.blk = b;
4754                                         if (!ctx->invalid_inode_table_flag[i] &&
4755                                             fix_problem(ctx,
4756                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4757                                                 ctx->invalid_inode_table_flag[i]++;
4758                                                 ctx->invalid_bitmaps++;
4759                                         }
4760                                 } else {
4761                                     ext2fs_mark_block_bitmap2(
4762                                                 ctx->block_found_map, b);
4763                                     ext2fs_mark_block_bitmap2(
4764                                                 ctx->block_metadata_map, b);
4765                                 }
4766                         }
4767                 }
4768
4769                 /*
4770                  * Mark block used for the block bitmap
4771                  */
4772                 if (ext2fs_block_bitmap_loc(fs, i)) {
4773                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4774                                      ext2fs_block_bitmap_loc(fs, i))) {
4775                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4776                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4777                                         ctx->invalid_block_bitmap_flag[i]++;
4778                                         ctx->invalid_bitmaps++;
4779                                 }
4780                         } else {
4781                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4782                                      ext2fs_block_bitmap_loc(fs, i));
4783                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4784                                      ext2fs_block_bitmap_loc(fs, i));
4785                         }
4786                 }
4787                 /*
4788                  * Mark block used for the inode bitmap
4789                  */
4790                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4791                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4792                                      ext2fs_inode_bitmap_loc(fs, i))) {
4793                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4794                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4795                                         ctx->invalid_inode_bitmap_flag[i]++;
4796                                         ctx->invalid_bitmaps++;
4797                                 }
4798                         } else {
4799                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4800                                      ext2fs_inode_bitmap_loc(fs, i));
4801                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4802                                      ext2fs_inode_bitmap_loc(fs, i));
4803                         }
4804                 }
4805         }
4806 }
4807
4808 /*
4809  * These subroutines short circuits ext2fs_get_blocks and
4810  * ext2fs_check_directory; we use them since we already have the inode
4811  * structure, so there's no point in letting the ext2fs library read
4812  * the inode again.
4813  */
4814 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4815                                   blk_t *blocks)
4816 {
4817         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4818         int     i;
4819
4820         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4821                 return EXT2_ET_CALLBACK_NOTHANDLED;
4822
4823         for (i=0; i < EXT2_N_BLOCKS; i++)
4824                 blocks[i] = ctx->stashed_inode->i_block[i];
4825         return 0;
4826 }
4827
4828 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4829                                   struct ext2_inode *inode)
4830 {
4831         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4832
4833         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4834                 return EXT2_ET_CALLBACK_NOTHANDLED;
4835         *inode = *ctx->stashed_inode;
4836         return 0;
4837 }
4838
4839 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4840                             struct ext2_inode *inode)
4841 {
4842         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4843
4844         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4845                 (inode != ctx->stashed_inode))
4846                 *ctx->stashed_inode = *inode;
4847         return EXT2_ET_CALLBACK_NOTHANDLED;
4848 }
4849
4850 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4851 {
4852         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4853
4854         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4855                 return EXT2_ET_CALLBACK_NOTHANDLED;
4856
4857         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4858                 return EXT2_ET_NO_DIRECTORY;
4859         return 0;
4860 }
4861
4862 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4863                                         blk64_t *ret)
4864 {
4865         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4866         errcode_t       retval;
4867         blk64_t         new_block;
4868
4869         if (ctx->block_found_map) {
4870                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4871                                            &new_block);
4872                 if (retval)
4873                         return retval;
4874                 if (fs->block_map) {
4875                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4876                         ext2fs_mark_bb_dirty(fs);
4877                 }
4878         } else {
4879                 if (!fs->block_map) {
4880                         retval = ext2fs_read_block_bitmap(fs);
4881                         if (retval)
4882                                 return retval;
4883                 }
4884
4885                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4886                 if (retval)
4887                         return retval;
4888         }
4889
4890         *ret = new_block;
4891         return (0);
4892 }
4893
4894 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4895                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4896 {
4897         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4898         errcode_t       retval;
4899
4900         if (ctx->block_found_map)
4901                 return ext2fs_new_range(fs, flags, goal, len,
4902                                         ctx->block_found_map, pblk, plen);
4903
4904         if (!fs->block_map) {
4905                 retval = ext2fs_read_block_bitmap(fs);
4906                 if (retval)
4907                         return retval;
4908         }
4909
4910         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4911                                 pblk, plen);
4912 }
4913
4914 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4915 {
4916         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4917
4918         /* Never free a critical metadata block */
4919         if (ctx->block_found_map &&
4920             ctx->block_metadata_map &&
4921             inuse < 0 &&
4922             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4923                 return;
4924
4925         if (ctx->block_found_map) {
4926                 if (inuse > 0)
4927                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4928                 else
4929                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4930         }
4931 }
4932
4933 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4934                                            blk_t num, int inuse)
4935 {
4936         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4937
4938         /* Never free a critical metadata block */
4939         if (ctx->block_found_map &&
4940             ctx->block_metadata_map &&
4941             inuse < 0 &&
4942             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4943                 return;
4944
4945         if (ctx->block_found_map) {
4946                 if (inuse > 0)
4947                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4948                                                         blk, num);
4949                 else
4950                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4951                                                         blk, num);
4952         }
4953 }
4954
4955 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4956 {
4957         ext2_filsys fs = ctx->fs;
4958
4959         if (use_shortcuts) {
4960                 fs->get_blocks = pass1_get_blocks;
4961                 fs->check_directory = pass1_check_directory;
4962                 fs->read_inode = pass1_read_inode;
4963                 fs->write_inode = pass1_write_inode;
4964                 ctx->stashed_ino = 0;
4965         } else {
4966                 fs->get_blocks = 0;
4967                 fs->check_directory = 0;
4968                 fs->read_inode = 0;
4969                 fs->write_inode = 0;
4970         }
4971 }
4972
4973 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4974 {
4975         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4976         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4977                                                 e2fsck_block_alloc_stats, 0);
4978         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4979         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4980                                         e2fsck_block_alloc_stats_range, NULL);
4981 }