e2fsck/pass1.c

   1 /*
   2  * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
   3  *
   4  * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
   5  *
   6  * %Begin-Header%
   7  * This file may be redistributed under the terms of the GNU Public
   8  * License.
   9  * %End-Header%
  10  *
  11  * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
  12  * and applies the following tests to each inode:
  13  *
  14  *      - The mode field of the inode must be legal.
  15  *      - The size and block count fields of the inode are correct.
  16  *      - A data block must not be used by another inode
  17  *
  18  * Pass 1 also gathers the collects the following information:
  19  *
  20  *      - A bitmap of which inodes are in use.          (inode_used_map)
  21  *      - A bitmap of which inodes are directories.     (inode_dir_map)
  22  *      - A bitmap of which inodes are regular files.   (inode_reg_map)
  23  *      - A bitmap of which inodes have bad fields.     (inode_bad_map)
  24  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
  25  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
  26  *      - A bitmap of which inodes are casefolded.      (inode_casefold_map)
  27  *      - A bitmap of which blocks are in use.          (block_found_map)
  28  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
  29  *      - The data blocks of the directory inodes.      (dir_map)
  30  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
  31  *      - The encryption policy ID of each encrypted inode. (encrypted_files)
  32  *
  33  * Pass 1 is designed to stash away enough information so that the
  34  * other passes should not need to read in the inode information
  35  * during the normal course of a filesystem check.  (Although if an
  36  * inconsistency is detected, other passes may need to read in an
  37  * inode to fix it.)
  38  *
  39  * Note that pass 1B will be invoked if there are any duplicate blocks
  40  * found.
  41  */
  42
  43 #define _GNU_SOURCE 1 /* get strnlen() */
  44 #include "config.h"
  45 #include <string.h>
  46 #include <time.h>
  47 #ifdef HAVE_ERRNO_H
  48 #include <errno.h>
  49 #endif
  50 #include <assert.h>
  51 #ifdef HAVE_PTHREAD
  52 #include <pthread.h>
  53 #endif
  54
  55 #include "e2fsck.h"
  56 #include <ext2fs/ext2_ext_attr.h>
  57 /* todo remove this finally */
  58 #include <ext2fs/ext2fsP.h>
  59 #include <e2p/e2p.h>
  60
  61 #include "problem.h"
  62
  63 #ifdef NO_INLINE_FUNCS
  64 #define _INLINE_
  65 #else
  66 #define _INLINE_ inline
  67 #endif
  68
  69 #undef DEBUG
  70
  71 struct ea_quota {
  72         blk64_t blocks;
  73         __u64 inodes;
  74 };
  75
  76 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  77                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  78                          int ref_offset, void *priv_data);
  79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  80                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  81                              int ref_offset, void *priv_data);
  82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  83                          char *block_buf,
  84                          const struct ea_quota *ea_ibody_quota);
  85 static void mark_table_blocks(e2fsck_t ctx);
  86 static void alloc_bb_map(e2fsck_t ctx);
  87 static void alloc_imagic_map(e2fsck_t ctx);
  88 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino);
  89 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino);
  90 static void handle_fs_bad_blocks(e2fsck_t ctx);
  91 static void process_inodes(e2fsck_t ctx, char *block_buf);
  92 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  93 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  94                                   dgrp_t group, void * priv_data);
  95 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  96                                     char *block_buf, int adjust_sign);
  97 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  98
  99 struct process_block_struct {
 100         ext2_ino_t      ino;
 101         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
 102                                 fragmented:1, compressed:1, bbcheck:1,
 103                                 inode_modified:1;
 104         blk64_t         num_blocks;
 105         blk64_t         max_blocks;
 106         blk64_t         last_block;
 107         e2_blkcnt_t     last_init_lblock;
 108         e2_blkcnt_t     last_db_block;
 109         int             num_illegal_blocks;
 110         blk64_t         previous_block;
 111         struct ext2_inode *inode;
 112         struct problem_context *pctx;
 113         ext2fs_block_bitmap fs_meta_blocks;
 114         e2fsck_t        ctx;
 115         blk64_t         next_lblock;
 116         struct extent_tree_info eti;
 117 };
 118
 119 struct process_inode_block {
 120         ext2_ino_t ino;
 121         struct ea_quota ea_ibody_quota;
 122         struct ext2_inode_large inode;
 123 };
 124
 125 struct scan_callback_struct {
 126         e2fsck_t        ctx;
 127         char            *block_buf;
 128 };
 129
 130 /*
 131  * For the inodes to process list.
 132  */
 133 static struct process_inode_block *inodes_to_process;
 134 static int process_inode_count;
 135
 136 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 137                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 138
 139 /*
 140  * Check to make sure a device inode is real.  Returns 1 if the device
 141  * checks out, 0 if not.
 142  *
 143  * Note: this routine is now also used to check FIFO's and Sockets,
 144  * since they have the same requirement; the i_block fields should be
 145  * zero.
 146  */
 147 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 148                                     struct ext2_inode *inode)
 149 {
 150         int     i;
 151
 152         /*
 153          * If the index or extents flag is set, then this is a bogus
 154          * device/fifo/socket
 155          */
 156         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 157                 return 0;
 158
 159         /*
 160          * We should be able to do the test below all the time, but
 161          * because the kernel doesn't forcibly clear the device
 162          * inode's additional i_block fields, there are some rare
 163          * occasions when a legitimate device inode will have non-zero
 164          * additional i_block fields.  So for now, we only complain
 165          * when the immutable flag is set, which should never happen
 166          * for devices.  (And that's when the problem is caused, since
 167          * you can't set or clear immutable flags for devices.)  Once
 168          * the kernel has been fixed we can change this...
 169          */
 170         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 171                 for (i=4; i < EXT2_N_BLOCKS; i++)
 172                         if (inode->i_block[i])
 173                                 return 0;
 174         }
 175         return 1;
 176 }
 177
 178 /*
 179  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 180  * checks out, 0 if not.
 181  */
 182 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 183                                struct ext2_inode *inode, char *buf)
 184 {
 185         unsigned int buflen;
 186         unsigned int len;
 187
 188         if ((inode->i_size_high || inode->i_size == 0) ||
 189             (inode->i_flags & EXT2_INDEX_FL))
 190                 return 0;
 191
 192         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 193                 size_t inline_size;
 194
 195                 if (inode->i_flags & EXT4_EXTENTS_FL)
 196                         return 0;
 197                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 198                         return 0;
 199                 if (inode->i_size != inline_size)
 200                         return 0;
 201
 202                 return 1;
 203         }
 204
 205         if (ext2fs_is_fast_symlink(inode)) {
 206                 if (inode->i_flags & EXT4_EXTENTS_FL)
 207                         return 0;
 208                 buf = (char *)inode->i_block;
 209                 buflen = sizeof(inode->i_block);
 210         } else {
 211                 ext2_extent_handle_t    handle;
 212                 struct ext2_extent_info info;
 213                 struct ext2fs_extent    extent;
 214                 blk64_t blk;
 215                 int i;
 216
 217                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 218                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 219                                 return 0;
 220                         if (ext2fs_extent_get_info(handle, &info) ||
 221                             (info.num_entries != 1) ||
 222                             (info.max_depth != 0)) {
 223                                 ext2fs_extent_free(handle);
 224                                 return 0;
 225                         }
 226                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 227                                               &extent) ||
 228                             (extent.e_lblk != 0) ||
 229                             (extent.e_len != 1)) {
 230                                 ext2fs_extent_free(handle);
 231                                 return 0;
 232                         }
 233                         blk = extent.e_pblk;
 234                         ext2fs_extent_free(handle);
 235                 } else {
 236                         blk = inode->i_block[0];
 237
 238                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 239                                 if (inode->i_block[i])
 240                                         return 0;
 241                 }
 242
 243                 if (blk < fs->super->s_first_data_block ||
 244                     blk >= ext2fs_blocks_count(fs->super))
 245                         return 0;
 246
 247                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 248                         return 0;
 249
 250                 buflen = fs->blocksize;
 251         }
 252
 253         if (inode->i_flags & EXT4_ENCRYPT_FL)
 254                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 255         else
 256                 len = strnlen(buf, buflen);
 257
 258         if (len >= buflen)
 259                 return 0;
 260
 261         if (len != inode->i_size)
 262                 return 0;
 263         return 1;
 264 }
 265
 266 /*
 267  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 268  */
 269 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 270 static void check_extents_inlinedata(e2fsck_t ctx,
 271                                      struct problem_context *pctx)
 272 {
 273         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 274                 return;
 275
 276         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 277                 return;
 278
 279         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 280         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 281 }
 282 #undef BAD_SPECIAL_FLAGS
 283
 284 /*
 285  * If the immutable (or append-only) flag is set on the inode, offer
 286  * to clear it.
 287  */
 288 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 289 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 290 {
 291         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 292                 return;
 293
 294         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 295                 return;
 296
 297         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 298         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 299 }
 300
 301 /*
 302  * If device, fifo or socket, check size is zero -- if not offer to
 303  * clear it
 304  */
 305 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 306 {
 307         struct ext2_inode *inode = pctx->inode;
 308
 309         if (EXT2_I_SIZE(inode) == 0)
 310                 return;
 311
 312         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 313                 return;
 314
 315         ext2fs_inode_size_set(ctx->fs, inode, 0);
 316         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 317 }
 318
 319 /*
 320  * For a given size, calculate how many blocks would be charged towards quota.
 321  */
 322 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 323 {
 324         blk64_t clusters;
 325
 326         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 327         return EXT2FS_C2B(fs, clusters);
 328 }
 329
 330 /*
 331  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 332  * the problem code.
 333  */
 334 static problem_t check_large_ea_inode(e2fsck_t ctx,
 335                                       struct ext2_ext_attr_entry *entry,
 336                                       struct problem_context *pctx,
 337                                       blk64_t *quota_blocks)
 338 {
 339         struct ext2_inode inode;
 340         __u32 hash, signed_hash;
 341         errcode_t retval;
 342
 343         /* Check if inode is within valid range */
 344         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 345             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 346                 pctx->num = entry->e_value_inum;
 347                 return PR_1_ATTR_VALUE_EA_INODE;
 348         }
 349
 350         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 351
 352         retval = ext2fs_ext_attr_hash_entry3(ctx->fs, entry, NULL, &hash,
 353                                              &signed_hash);
 354         if (retval) {
 355                 com_err("check_large_ea_inode", retval,
 356                         _("while hashing entry with e_value_inum = %u"),
 357                         entry->e_value_inum);
 358                 fatal_error(ctx, 0);
 359         }
 360
 361         if ((hash == entry->e_hash) || (signed_hash == entry->e_hash)) {
 362                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 363                                                      entry->e_value_size);
 364         } else {
 365                 /* This might be an old Lustre-style ea_inode reference. */
 366                 if (inode.i_mtime == pctx->ino &&
 367                     inode.i_generation == pctx->inode->i_generation) {
 368                         *quota_blocks = 0;
 369                 } else {
 370                         /* If target inode is also missing EA_INODE flag,
 371                          * this is likely to be a bad reference.
 372                          */
 373                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 374                                 pctx->num = entry->e_value_inum;
 375                                 return PR_1_ATTR_VALUE_EA_INODE;
 376                         } else {
 377                                 pctx->num = entry->e_hash;
 378                                 return PR_1_ATTR_HASH;
 379                         }
 380                 }
 381         }
 382
 383         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 384                 pctx->num = entry->e_value_inum;
 385                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 386                         inode.i_flags |= EXT4_EA_INODE_FL;
 387                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 388                                            &inode);
 389                 } else {
 390                         return PR_1_ATTR_NO_EA_INODE_FL;
 391                 }
 392         }
 393         return 0;
 394 }
 395
 396 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 397                               struct ext2_ext_attr_entry *first, void *end)
 398 {
 399         struct ext2_ext_attr_entry *entry = first;
 400         struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
 401
 402         while ((void *) entry < end && (void *) np < end &&
 403                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 404                 if (!entry->e_value_inum)
 405                         goto next;
 406                 if (!ctx->ea_inode_refs) {
 407                         pctx->errcode = ea_refcount_create(0,
 408                                                            &ctx->ea_inode_refs);
 409                         if (pctx->errcode) {
 410                                 pctx->num = 4;
 411                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 412                                 ctx->flags |= E2F_FLAG_ABORT;
 413                                 return;
 414                         }
 415                 }
 416                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 417                                       0);
 418         next:
 419                 entry = np;
 420                 np = EXT2_EXT_ATTR_NEXT(entry);
 421         }
 422 }
 423
 424 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 425                               struct ea_quota *ea_ibody_quota)
 426 {
 427         struct ext2_super_block *sb = ctx->fs->super;
 428         struct ext2_inode_large *inode;
 429         struct ext2_ext_attr_entry *entry;
 430         char *start, *header, *end;
 431         unsigned int storage_size, remain;
 432         problem_t problem = 0;
 433         region_t region = 0;
 434
 435         ea_ibody_quota->blocks = 0;
 436         ea_ibody_quota->inodes = 0;
 437
 438         inode = (struct ext2_inode_large *) pctx->inode;
 439         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 440                 inode->i_extra_isize;
 441         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 442                  inode->i_extra_isize;
 443         end = header + storage_size;
 444         start = header + sizeof(__u32);
 445         entry = (struct ext2_ext_attr_entry *) start;
 446
 447         /* scan all entry's headers first */
 448
 449         /* take finish entry 0UL into account */
 450         remain = storage_size - sizeof(__u32);
 451
 452         region = region_create(0, storage_size);
 453         if (!region) {
 454                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 455                 problem = 0;
 456                 ctx->flags |= E2F_FLAG_ABORT;
 457                 return;
 458         }
 459         if (region_allocate(region, 0, sizeof(__u32))) {
 460                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 461                 goto fix;
 462         }
 463
 464         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 465                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 466                 __u32 hash;
 467
 468                 if (region_allocate(region, (char *)entry - (char *)header,
 469                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 470                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 471                         goto fix;
 472                 }
 473
 474                 /* header eats this space */
 475                 remain -= sizeof(struct ext2_ext_attr_entry);
 476
 477                 /* is attribute name valid? */
 478                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 479                         pctx->num = entry->e_name_len;
 480                         problem = PR_1_ATTR_NAME_LEN;
 481                         goto fix;
 482                 }
 483
 484                 /* attribute len eats this space */
 485                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 486
 487                 if (entry->e_value_inum == 0) {
 488                         /* check value size */
 489                         if (entry->e_value_size > remain) {
 490                                 pctx->num = entry->e_value_size;
 491                                 problem = PR_1_ATTR_VALUE_SIZE;
 492                                 goto fix;
 493                         }
 494
 495                         if (entry->e_value_size &&
 496                             region_allocate(region,
 497                                             sizeof(__u32) + entry->e_value_offs,
 498                                             EXT2_EXT_ATTR_SIZE(
 499                                                 entry->e_value_size))) {
 500                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 501                                 goto fix;
 502                         }
 503
 504                         hash = ext2fs_ext_attr_hash_entry(entry,
 505                                                 start + entry->e_value_offs);
 506                         if (entry->e_hash != 0 && entry->e_hash != hash)
 507                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
 508                                                 start + entry->e_value_offs);
 509
 510                         /* e_hash may be 0 in older inode's ea */
 511                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 512                                 pctx->num = entry->e_hash;
 513                                 problem = PR_1_ATTR_HASH;
 514                                 goto fix;
 515                         }
 516                 } else {
 517                         blk64_t quota_blocks;
 518
 519                         problem = check_large_ea_inode(ctx, entry, pctx,
 520                                                        &quota_blocks);
 521                         if (problem != 0)
 522                                 goto fix;
 523
 524                         ea_ibody_quota->blocks += quota_blocks;
 525                         ea_ibody_quota->inodes++;
 526                 }
 527
 528                 /* If EA value is stored in external inode then it does not
 529                  * consume space here */
 530                 if (entry->e_value_inum == 0)
 531                         remain -= entry->e_value_size;
 532
 533                 entry = EXT2_EXT_ATTR_NEXT(entry);
 534         }
 535
 536         if (region_allocate(region, (char *)entry - (char *)header,
 537                             sizeof(__u32))) {
 538                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 539                 goto fix;
 540         }
 541 fix:
 542         if (region)
 543                 region_free(region);
 544         /*
 545          * it seems like a corruption. it's very unlikely we could repair
 546          * EA(s) in automatic fashion -bzzz
 547          */
 548         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 549                 inc_ea_inode_refs(ctx, pctx,
 550                                   (struct ext2_ext_attr_entry *)start, end);
 551                 return;
 552         }
 553
 554         /* simply remove all possible EA(s) */
 555         *((__u32 *)header) = 0UL;
 556         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 557                                 EXT2_INODE_SIZE(sb), "pass1");
 558         ea_ibody_quota->blocks = 0;
 559         ea_ibody_quota->inodes = 0;
 560 }
 561
 562 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 563         return (xtime & (1U << 31)) != 0 &&
 564                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 565 }
 566
 567 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 568         check_inode_extra_negative_epoch(inode->i_##xtime, \
 569                                          inode->i_##xtime##_extra)
 570
 571 /* When today's date is earlier than 2242, we assume that atimes,
 572  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 573  * actually pre-1970 dates mis-encoded.
 574  */
 575 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 576
 577 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 578                                     struct ea_quota *ea_ibody_quota)
 579 {
 580         struct ext2_super_block *sb = ctx->fs->super;
 581         struct ext2_inode_large *inode;
 582         __u32 *eamagic;
 583         int min, max;
 584
 585         ea_ibody_quota->blocks = 0;
 586         ea_ibody_quota->inodes = 0;
 587
 588         inode = (struct ext2_inode_large *) pctx->inode;
 589         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 590                 /* this isn't large inode. so, nothing to check */
 591                 return;
 592         }
 593
 594 #if 0
 595         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 596                         inode->i_extra_isize);
 597 #endif
 598         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 599         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 600         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 601         /*
 602          * For now we will allow i_extra_isize to be 0, but really
 603          * implementations should never allow i_extra_isize to be 0
 604          */
 605         if (inode->i_extra_isize &&
 606             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 607              inode->i_extra_isize & 3)) {
 608                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 609                         return;
 610                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 611                         inode->i_extra_isize = sb->s_want_extra_isize;
 612                 else
 613                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 614                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 615                                         EXT2_INODE_SIZE(sb), "pass1");
 616         }
 617
 618         /* check if there is no place for an EA header */
 619         if (inode->i_extra_isize >= max - sizeof(__u32))
 620                 return;
 621
 622         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 623                         inode->i_extra_isize);
 624         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 625                 /* it seems inode has an extended attribute(s) in body */
 626                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 627         }
 628
 629         /*
 630          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 631          * the old, invalid format, repair it.
 632          */
 633         if (((sizeof(time_t) <= 4) ||
 634              (((sizeof(time_t) > 4) &&
 635                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 636             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 637              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 638              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 639              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 640
 641                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 642                         return;
 643
 644                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 645                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 646                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 647                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 648                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 649                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 650                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 651                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 652                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 653                                         EXT2_INODE_SIZE(sb), "pass1");
 654         }
 655
 656 }
 657
 658 /*
 659  * Check to see if the inode might really be a directory, despite i_mode
 660  *
 661  * This is a lot of complexity for something for which I'm not really
 662  * convinced happens frequently in the wild.  If for any reason this
 663  * causes any problems, take this code out.
 664  * [tytso:20070331.0827EDT]
 665  */
 666 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 667                                 char *buf)
 668 {
 669         struct ext2_inode *inode = pctx->inode;
 670         struct ext2_dir_entry   *dirent;
 671         errcode_t               retval;
 672         blk64_t                 blk;
 673         unsigned int            i, rec_len, not_device = 0;
 674         int                     extent_fs;
 675         int                     inlinedata_fs;
 676
 677         /*
 678          * If the mode looks OK, we believe it.  If the first block in
 679          * the i_block array is 0, this cannot be a directory. If the
 680          * inode is extent-mapped, it is still the case that the latter
 681          * cannot be 0 - the magic number in the extent header would make
 682          * it nonzero.
 683          */
 684         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 685             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 686                 return;
 687
 688         /*
 689          * Check the block numbers in the i_block array for validity:
 690          * zero blocks are skipped (but the first one cannot be zero -
 691          * see above), other blocks are checked against the first and
 692          * max data blocks (from the the superblock) and against the
 693          * block bitmap. Any invalid block found means this cannot be
 694          * a directory.
 695          *
 696          * If there are non-zero blocks past the fourth entry, then
 697          * this cannot be a device file: we remember that for the next
 698          * check.
 699          *
 700          * For extent mapped files, we don't do any sanity checking:
 701          * just try to get the phys block of logical block 0 and run
 702          * with it.
 703          *
 704          * For inline data files, we just try to get the size of inline
 705          * data.  If it's true, we will treat it as a directory.
 706          */
 707
 708         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 709         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 710         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 711                 size_t size;
 712                 __u32 dotdot;
 713                 unsigned int rec_len2;
 714                 struct ext2_dir_entry de;
 715
 716                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 717                         return;
 718                 /*
 719                  * If the size isn't a multiple of 4, it's probably not a
 720                  * directory??
 721                  */
 722                 if (size & 3)
 723                         return;
 724                 /*
 725                  * If the first 10 bytes don't look like a directory entry,
 726                  * it's probably not a directory.
 727                  */
 728                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 729                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 730                        EXT2_DIR_REC_LEN(0));
 731                 dotdot = ext2fs_le32_to_cpu(dotdot);
 732                 de.inode = ext2fs_le32_to_cpu(de.inode);
 733                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 734                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 735                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 736                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 737                      dotdot != EXT2_ROOT_INO) ||
 738                     de.inode >= ctx->fs->super->s_inodes_count ||
 739                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 740                      de.inode != 0) ||
 741                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 742                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 743                         return;
 744                 /* device files never have a "system.data" entry */
 745                 goto isdir;
 746         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 747                 /* extent mapped */
 748                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 749                                  &blk))
 750                         return;
 751                 /* device files are never extent mapped */
 752                 not_device++;
 753         } else {
 754                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 755                         blk = inode->i_block[i];
 756                         if (!blk)
 757                                 continue;
 758                         if (i >= 4)
 759                                 not_device++;
 760
 761                         if (blk < ctx->fs->super->s_first_data_block ||
 762                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 763                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 764                                                            blk))
 765                                 return; /* Invalid block, can't be dir */
 766                 }
 767                 blk = inode->i_block[0];
 768         }
 769
 770         /*
 771          * If the mode says this is a device file and the i_links_count field
 772          * is sane and we have not ruled it out as a device file previously,
 773          * we declare it a device file, not a directory.
 774          */
 775         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 776             (inode->i_links_count == 1) && !not_device)
 777                 return;
 778
 779         /* read the first block */
 780         ehandler_operation(_("reading directory block"));
 781         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 782         ehandler_operation(0);
 783         if (retval)
 784                 return;
 785
 786         dirent = (struct ext2_dir_entry *) buf;
 787         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 788         if (retval)
 789                 return;
 790         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 791             (dirent->name[0] != '.') ||
 792             (dirent->inode != pctx->ino) ||
 793             (rec_len < 12) ||
 794             (rec_len % 4) ||
 795             (rec_len >= ctx->fs->blocksize - 12))
 796                 return;
 797
 798         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 799         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 800         if (retval)
 801                 return;
 802         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 803             (dirent->name[0] != '.') ||
 804             (dirent->name[1] != '.') ||
 805             (rec_len < 12) ||
 806             (rec_len % 4))
 807                 return;
 808
 809 isdir:
 810         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 811                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 812                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 813                                         EXT2_INODE_SIZE(ctx->fs->super),
 814                                         "check_is_really_dir");
 815         }
 816 }
 817
 818 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 819                                      int flags, ext2_icount_t hint,
 820                                      ext2_icount_t *ret)
 821 {
 822         unsigned int            threshold;
 823         unsigned int            save_type;
 824         ext2_ino_t              num_dirs;
 825         errcode_t               retval;
 826         char                    *tdb_dir;
 827         int                     enable;
 828
 829         *ret = 0;
 830
 831         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 832                            &tdb_dir);
 833         profile_get_uint(ctx->profile, "scratch_files",
 834                          "numdirs_threshold", 0, 0, &threshold);
 835         profile_get_boolean(ctx->profile, "scratch_files",
 836                             "icount", 0, 1, &enable);
 837
 838         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 839         if (retval)
 840                 num_dirs = 1024;        /* Guess */
 841
 842         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 843             (!threshold || num_dirs > threshold)) {
 844                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 845                                                   flags, ret);
 846                 if (retval == 0)
 847                         return 0;
 848         }
 849         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 850                                &save_type);
 851         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 852                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 853         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 854         ctx->fs->default_bitmap_type = save_type;
 855         return retval;
 856 }
 857
 858 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 859                                             e2fsck_t ctx,
 860                                             struct problem_context *pctx)
 861 {
 862         errcode_t retval;
 863         struct ext2_inode_large inode;
 864
 865         /*
 866          * Reread inode.  If we don't see checksum error, then this inode
 867          * has been fixed elsewhere.
 868          */
 869         ctx->stashed_ino = 0;
 870         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 871                                         sizeof(inode));
 872         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 873                 return retval;
 874         if (!retval)
 875                 return 0;
 876
 877         /*
 878          * Checksum still doesn't match.  That implies that the inode passes
 879          * all the sanity checks, so maybe the checksum is simply corrupt.
 880          * See if the user will go for fixing that.
 881          */
 882         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 883                 return 0;
 884
 885         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 886                                          sizeof(inode));
 887         return retval;
 888 }
 889
 890 static void reserve_block_for_root_repair(e2fsck_t ctx)
 891 {
 892         blk64_t         blk = 0;
 893         errcode_t       err;
 894         ext2_filsys     fs = ctx->fs;
 895
 896         ctx->root_repair_block = 0;
 897         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 898                 return;
 899
 900         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 901         if (err)
 902                 return;
 903         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 904         ctx->root_repair_block = blk;
 905 }
 906
 907 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 908 {
 909         blk64_t         blk = 0;
 910         errcode_t       err;
 911         ext2_filsys     fs = ctx->fs;
 912         static const char name[] = "lost+found";
 913         ext2_ino_t      ino;
 914
 915         ctx->lnf_repair_block = 0;
 916         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 917                 return;
 918
 919         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 920         if (err)
 921                 return;
 922         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 923         ctx->lnf_repair_block = blk;
 924 }
 925
 926 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 927                                          struct ext2_inode *inode,
 928                                          size_t *sz)
 929 {
 930         void *p;
 931         struct ext2_xattr_handle *handle;
 932         errcode_t retval;
 933
 934         retval = ext2fs_xattrs_open(fs, ino, &handle);
 935         if (retval)
 936                 return retval;
 937
 938         retval = ext2fs_xattrs_read_inode(handle,
 939                                           (struct ext2_inode_large *)inode);
 940         if (retval)
 941                 goto err;
 942
 943         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 944         if (retval)
 945                 goto err;
 946         ext2fs_free_mem(&p);
 947 err:
 948         (void) ext2fs_xattrs_close(&handle);
 949         return retval;
 950 }
 951
 952 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 953                                     struct problem_context *pctx,
 954                                     int failed_csum)
 955 {
 956         if (!failed_csum)
 957                 return;
 958
 959         /*
 960          * If the inode failed the checksum and the user didn't
 961          * clear the inode, test the checksum again -- if it still
 962          * fails, ask the user if the checksum should be corrected.
 963          */
 964         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 965         if (pctx->errcode)
 966                 ctx->flags |= E2F_FLAG_ABORT;
 967 }
 968 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 969         do { \
 970                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 971                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 972                         return; \
 973         } while (0)
 974
 975 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 976 {
 977         __u32 x;
 978         int i;
 979
 980         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 981                 x = inode->i_block[i];
 982 #ifdef WORDS_BIGENDIAN
 983                 x = ext2fs_swab32(x);
 984 #endif
 985                 if (x >= ext2fs_blocks_count(fs->super))
 986                         return 0;
 987         }
 988
 989         return 1;
 990 }
 991
 992 /*
 993  * Figure out what to do with an inode that has both extents and inline data
 994  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 995  */
 996 static int fix_inline_data_extents_file(e2fsck_t ctx,
 997                                         ext2_ino_t ino,
 998                                         struct ext2_inode *inode,
 999                                         int inode_size,
1000                                         struct problem_context *pctx)
1001 {
1002         size_t max_inline_ea_size;
1003         ext2_filsys fs = ctx->fs;
1004         int dirty = 0;
1005
1006         /* Both feature flags not set?  Just run the regular checks */
1007         if (!ext2fs_has_feature_extents(fs->super) &&
1008             !ext2fs_has_feature_inline_data(fs->super))
1009                 return 0;
1010
1011         /* Clear both flags if it's a special file */
1012         if (LINUX_S_ISCHR(inode->i_mode) ||
1013             LINUX_S_ISBLK(inode->i_mode) ||
1014             LINUX_S_ISFIFO(inode->i_mode) ||
1015             LINUX_S_ISSOCK(inode->i_mode)) {
1016                 check_extents_inlinedata(ctx, pctx);
1017                 return 0;
1018         }
1019
1020         /* If it looks like an extent tree, try to clear inlinedata */
1021         if (ext2fs_extent_header_verify(inode->i_block,
1022                                  sizeof(inode->i_block)) == 0 &&
1023             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1024                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1025                 dirty = 1;
1026                 goto out;
1027         }
1028
1029         /* If it looks short enough to be inline data, try to clear extents */
1030         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1031                 max_inline_ea_size = inode_size -
1032                                      (EXT2_GOOD_OLD_INODE_SIZE +
1033                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1034         else
1035                 max_inline_ea_size = 0;
1036         if (EXT2_I_SIZE(inode) <
1037             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1038             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1039                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1040                 dirty = 1;
1041                 goto out;
1042         }
1043
1044         /*
1045          * Too big for inline data, but no evidence of extent tree -
1046          * maybe it's a block map file?  If the mappings all look valid?
1047          */
1048         if (could_be_block_map(fs, inode) &&
1049             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1050 #ifdef WORDS_BIGENDIAN
1051                 int i;
1052
1053                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1054                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1055 #endif
1056
1057                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1058                 dirty = 1;
1059                 goto out;
1060         }
1061
1062         /* Oh well, just clear the busted inode. */
1063         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1064                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1065                 return -1;
1066         }
1067
1068 out:
1069         if (dirty)
1070                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1071
1072         return 0;
1073 }
1074
1075 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1076 {
1077         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1078         dgrp_t start = *group, grp;
1079         blk64_t blocks_to_read = 0;
1080         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1081
1082         if (ctx->readahead_kb == 0)
1083                 goto out;
1084
1085         /* Keep iterating groups until we have enough to readahead */
1086         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1087         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1088                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1089                         continue;
1090                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1091                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1092                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1093                                         inodes_per_block;
1094                 if (blocks_to_read * ctx->fs->blocksize >
1095                     ctx->readahead_kb * 1024)
1096                         break;
1097         }
1098
1099         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1100                                grp - start + 1);
1101         if (err == EAGAIN) {
1102                 ctx->readahead_kb /= 2;
1103                 err = 0;
1104         }
1105
1106 out:
1107         if (err) {
1108                 /* Error; disable itable readahead */
1109                 *group = ctx->fs->group_desc_count;
1110                 *next_ino = ctx->fs->super->s_inodes_count;
1111         } else {
1112                 /*
1113                  * Don't do more readahead until we've reached the first inode
1114                  * of the last inode scan buffer block for the last group.
1115                  */
1116                 *group = grp + 1;
1117                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1118                                      ctx->inode_buffer_blocks :
1119                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1120                                     ctx->fs->blocksize /
1121                                     EXT2_INODE_SIZE(ctx->fs->super);
1122                 inodes_in_group--;
1123                 *next_ino = inodes_in_group -
1124                             (inodes_in_group % inodes_per_buffer) + 1 +
1125                             (grp * ctx->fs->super->s_inodes_per_group);
1126         }
1127 }
1128
1129 /*
1130  * Check if the passed ino is one of the used superblock quota inodes.
1131  *
1132  * Before the quota inodes were journaled, older superblock quota inodes
1133  * were just regular files in the filesystem and not reserved inodes.  This
1134  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1135  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1136  */
1137 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1138 {
1139         enum quota_type qtype;
1140
1141         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1142                 if (*quota_sb_inump(sb, qtype) == ino)
1143                         return 1;
1144
1145         return 0;
1146 }
1147
1148 /*
1149  * Check if the passed ino is one of the reserved quota inodes.
1150  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1151  * inodes.  These inodes may or may not be in use by the quota feature.
1152  */
1153 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1154 {
1155         enum quota_type qtype;
1156
1157         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1158                 if (quota_type2inum(qtype, fs->super) == ino)
1159                         return 1;
1160
1161         return 0;
1162 }
1163
1164 static int e2fsck_should_abort(e2fsck_t ctx)
1165 {
1166         e2fsck_t global_ctx;
1167
1168         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1169                 return 1;
1170
1171         if (ctx->global_ctx) {
1172                 global_ctx = ctx->global_ctx;
1173                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1174                         return 1;
1175         }
1176         return 0;
1177 }
1178
1179 void e2fsck_pass1_run(e2fsck_t ctx)
1180 {
1181         int     i;
1182         __u64   max_sizes;
1183         ext2_filsys fs = ctx->fs;
1184         ext2_ino_t      ino = 0;
1185         struct ext2_inode *inode = NULL;
1186         ext2_inode_scan scan = NULL;
1187         char            *block_buf = NULL;
1188 #ifdef RESOURCE_TRACK
1189         struct resource_track   rtrack;
1190 #endif
1191         unsigned char   frag, fsize;
1192         struct          problem_context pctx;
1193         struct          scan_callback_struct scan_struct;
1194         struct ext2_super_block *sb = ctx->fs->super;
1195         const char      *old_op;
1196         const char      *eop_next_inode = _("getting next inode from scan");
1197         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1198         int             low_dtime_check = 1;
1199         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1200         unsigned int    bufsize;
1201         int             failed_csum = 0;
1202         ext2_ino_t      ino_threshold = 0;
1203         dgrp_t          ra_group = 0;
1204         struct ea_quota ea_ibody_quota;
1205
1206         init_resource_track(&rtrack, ctx->fs->io);
1207         clear_problem_context(&pctx);
1208
1209         /* If we can do readahead, figure out how many groups to pull in. */
1210         if (!e2fsck_can_readahead(ctx->fs))
1211                 ctx->readahead_kb = 0;
1212         else if (ctx->readahead_kb == ~0ULL)
1213                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1214         pass1_readahead(ctx, &ra_group, &ino_threshold);
1215
1216         if (!(ctx->options & E2F_OPT_PREEN))
1217                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1218
1219         if (ext2fs_has_feature_dir_index(fs->super) &&
1220             !(ctx->options & E2F_OPT_NO)) {
1221                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1222                         ctx->dirs_to_hash = 0;
1223         }
1224
1225 #ifdef MTRACE
1226         mtrace_print("Pass 1");
1227 #endif
1228
1229 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1230
1231         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1232                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1233                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1234                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1235                 max_sizes = (max_sizes * (1UL << i));
1236                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1237         }
1238 #undef EXT2_BPP
1239
1240         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1241         extent_fs = ext2fs_has_feature_extents(sb);
1242         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1243         casefold_fs = ext2fs_has_feature_casefold(sb);
1244
1245         /*
1246          * Allocate bitmaps structures
1247          */
1248         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1249                                                     EXT2FS_BMAP64_RBTREE,
1250                                                     "inode_used_map",
1251                                                     &ctx->inode_used_map);
1252         if (pctx.errcode) {
1253                 pctx.num = 1;
1254                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1255                 ctx->flags |= E2F_FLAG_ABORT;
1256                 return;
1257         }
1258         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1259                         _("directory inode map"),
1260                         EXT2FS_BMAP64_AUTODIR,
1261                         "inode_dir_map", &ctx->inode_dir_map);
1262         if (pctx.errcode) {
1263                 pctx.num = 2;
1264                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1265                 ctx->flags |= E2F_FLAG_ABORT;
1266                 return;
1267         }
1268         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1269                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1270                         "inode_reg_map", &ctx->inode_reg_map);
1271         if (pctx.errcode) {
1272                 pctx.num = 6;
1273                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1274                 ctx->flags |= E2F_FLAG_ABORT;
1275                 return;
1276         }
1277         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1278                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1279                         "block_found_map", &ctx->block_found_map);
1280         if (pctx.errcode) {
1281                 pctx.num = 1;
1282                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1283                 ctx->flags |= E2F_FLAG_ABORT;
1284                 return;
1285         }
1286         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1287                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1288                         "block_metadata_map", &ctx->block_metadata_map);
1289         if (pctx.errcode) {
1290                 pctx.num = 1;
1291                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1292                 ctx->flags |= E2F_FLAG_ABORT;
1293                 return;
1294         }
1295         if (casefold_fs) {
1296                 pctx.errcode =
1297                         e2fsck_allocate_inode_bitmap(fs,
1298                                                      _("inode casefold map"),
1299                                                      EXT2FS_BMAP64_RBTREE,
1300                                                      "inode_casefold_map",
1301                                                      &ctx->inode_casefold_map);
1302                 if (pctx.errcode) {
1303                         pctx.num = 1;
1304                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1305                         ctx->flags |= E2F_FLAG_ABORT;
1306                         return;
1307                 }
1308         }
1309         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1310                                            &ctx->inode_link_info);
1311         if (pctx.errcode) {
1312                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1313                 ctx->flags |= E2F_FLAG_ABORT;
1314                 return;
1315         }
1316         bufsize = inode_size;
1317         if (bufsize < sizeof(struct ext2_inode_large))
1318                 bufsize = sizeof(struct ext2_inode_large);
1319         inode = (struct ext2_inode *)
1320                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1321
1322         inodes_to_process = (struct process_inode_block *)
1323                 e2fsck_allocate_memory(ctx,
1324                                        (ctx->process_inode_size *
1325                                         sizeof(struct process_inode_block)),
1326                                        "array of inodes to process");
1327         process_inode_count = 0;
1328
1329         pctx.errcode = ext2fs_init_dblist(fs, 0);
1330         if (pctx.errcode) {
1331                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1332                 ctx->flags |= E2F_FLAG_ABORT;
1333                 goto endit;
1334         }
1335
1336         /*
1337          * If the last orphan field is set, clear it, since the pass1
1338          * processing will automatically find and clear the orphans.
1339          * In the future, we may want to try using the last_orphan
1340          * linked list ourselves, but for now, we clear it so that the
1341          * ext3 mount code won't get confused.
1342          */
1343         if (!(ctx->options & E2F_OPT_READONLY)) {
1344                 if (fs->super->s_last_orphan) {
1345                         fs->super->s_last_orphan = 0;
1346                         ext2fs_mark_super_dirty(fs);
1347                 }
1348         }
1349
1350         mark_table_blocks(ctx);
1351         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1352                                                 &ctx->block_found_map);
1353         if (pctx.errcode) {
1354                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1355                 ctx->flags |= E2F_FLAG_ABORT;
1356                 goto endit;
1357         }
1358         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1359                                                     "block iterate buffer");
1360         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1361                 e2fsck_use_inode_shortcuts(ctx, 1);
1362         e2fsck_intercept_block_allocations(ctx);
1363         old_op = ehandler_operation(_("opening inode scan"));
1364         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1365                                               &scan);
1366         ehandler_operation(old_op);
1367         if (pctx.errcode) {
1368                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1369                 ctx->flags |= E2F_FLAG_ABORT;
1370                 goto endit;
1371         }
1372         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1373                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1374         ctx->stashed_inode = inode;
1375         scan_struct.ctx = ctx;
1376         scan_struct.block_buf = block_buf;
1377         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1378         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1379                                               ctx->fs->group_desc_count)))
1380                 goto endit;
1381         if ((fs->super->s_wtime &&
1382              fs->super->s_wtime < fs->super->s_inodes_count) ||
1383             (fs->super->s_mtime &&
1384              fs->super->s_mtime < fs->super->s_inodes_count) ||
1385             (fs->super->s_mkfs_time &&
1386              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1387                 low_dtime_check = 0;
1388
1389         if (ext2fs_has_feature_mmp(fs->super) &&
1390             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1391             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1392                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1393                                           fs->super->s_mmp_block);
1394
1395         /* Set up ctx->lost_and_found if possible */
1396         (void) e2fsck_get_lost_and_found(ctx, 0);
1397
1398 #ifdef HAVE_PTHREAD
1399         if (ctx->global_ctx) {
1400                 if (ctx->options & E2F_OPT_DEBUG &&
1401                     ctx->options & E2F_OPT_MULTITHREAD)
1402                         fprintf(stderr, "thread %d jumping to group %d\n",
1403                                         ctx->thread_info.et_thread_index,
1404                                         ctx->thread_info.et_group_start);
1405                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1406                                         ctx->thread_info.et_group_start);
1407                 if (pctx.errcode) {
1408                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1409                         ctx->flags |= E2F_FLAG_ABORT;
1410                         goto endit;
1411                 }
1412         }
1413 #endif
1414
1415         while (1) {
1416                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1417                         if (e2fsck_mmp_update(fs))
1418                                 fatal_error(ctx, 0);
1419                 }
1420                 old_op = ehandler_operation(eop_next_inode);
1421                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1422                                                           inode, inode_size);
1423                 if (ino > ino_threshold)
1424                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1425                 ehandler_operation(old_op);
1426                 if (e2fsck_should_abort(ctx))
1427                         goto endit;
1428                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1429                         /*
1430                          * If badblocks says badblocks is bad, offer to clear
1431                          * the list, update the in-core bb list, and restart
1432                          * the inode scan.
1433                          */
1434                         if (ino == EXT2_BAD_INO &&
1435                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1436                                         &pctx)) {
1437                                 errcode_t err;
1438
1439                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1440                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1441                                 ctx->fs->badblocks = NULL;
1442                                 err = ext2fs_read_bb_inode(ctx->fs,
1443                                                         &ctx->fs->badblocks);
1444                                 if (err) {
1445                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1446                                                     &pctx);
1447                                         ctx->flags |= E2F_FLAG_ABORT;
1448                                 } else
1449                                         ctx->flags |= E2F_FLAG_RESTART;
1450                                 goto endit;
1451                         }
1452                         if (!ctx->inode_bb_map)
1453                                 alloc_bb_map(ctx);
1454                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1455                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1456                         continue;
1457                 }
1458                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1459                         break;
1460                 if (pctx.errcode &&
1461                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1462                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1463                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1464                         ctx->flags |= E2F_FLAG_ABORT;
1465                         goto endit;
1466                 }
1467                 if (!ino)
1468                         break;
1469                 pctx.ino = ino;
1470                 pctx.inode = inode;
1471                 ctx->stashed_ino = ino;
1472
1473                 /* Clear trashed inode? */
1474                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1475                     inode->i_links_count > 0 &&
1476                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1477                         pctx.errcode = 0;
1478                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1479                 }
1480                 failed_csum = pctx.errcode != 0;
1481
1482                 /*
1483                  * Check for inodes who might have been part of the
1484                  * orphaned list linked list.  They should have gotten
1485                  * dealt with by now, unless the list had somehow been
1486                  * corrupted.
1487                  *
1488                  * FIXME: In the future, inodes which are still in use
1489                  * (and which are therefore) pending truncation should
1490                  * be handled specially.  Right now we just clear the
1491                  * dtime field, and the normal e2fsck handling of
1492                  * inodes where i_size and the inode blocks are
1493                  * inconsistent is to fix i_size, instead of releasing
1494                  * the extra blocks.  This won't catch the inodes that
1495                  * was at the end of the orphan list, but it's better
1496                  * than nothing.  The right answer is that there
1497                  * shouldn't be any bugs in the orphan list handling.  :-)
1498                  */
1499                 if (inode->i_dtime && low_dtime_check &&
1500                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1501                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1502                                 inode->i_dtime = inode->i_links_count ?
1503                                         0 : ctx->now;
1504                                 e2fsck_write_inode(ctx, ino, inode,
1505                                                    "pass1");
1506                                 failed_csum = 0;
1507                         }
1508                 }
1509
1510                 if (inode->i_links_count) {
1511                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1512                                            ino, inode->i_links_count);
1513                         if (pctx.errcode) {
1514                                 pctx.num = inode->i_links_count;
1515                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1516                                 ctx->flags |= E2F_FLAG_ABORT;
1517                                 goto endit;
1518                         }
1519                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1520                            !quota_inum_is_reserved(fs, ino)) {
1521                         if (!inode->i_dtime && inode->i_mode) {
1522                                 if (fix_problem(ctx,
1523                                             PR_1_ZERO_DTIME, &pctx)) {
1524                                         inode->i_dtime = ctx->now;
1525                                         e2fsck_write_inode(ctx, ino, inode,
1526                                                            "pass1");
1527                                         failed_csum = 0;
1528                                 }
1529                         }
1530                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1531                         continue;
1532                 }
1533
1534                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1535                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1536                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1537                      (!casefold_fs &&
1538                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1539                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1540                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1541                 }
1542
1543                 /* Conflicting inlinedata/extents inode flags? */
1544                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1545                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1546                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1547                                                                inode_size,
1548                                                                &pctx);
1549                         if (res < 0) {
1550                                 /* skip FINISH_INODE_LOOP */
1551                                 continue;
1552                         }
1553                 }
1554
1555                 /* Test for incorrect inline_data flags settings. */
1556                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1557                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1558                         size_t size = 0;
1559
1560                         pctx.errcode = get_inline_data_ea_size(fs, ino, inode,
1561                                                                &size);
1562                         if (!pctx.errcode &&
1563                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1564                                 ext2fs_set_feature_inline_data(sb);
1565                                 ext2fs_mark_super_dirty(fs);
1566                                 inlinedata_fs = 1;
1567                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1568                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1569                                 /* skip FINISH_INODE_LOOP */
1570                                 continue;
1571                         }
1572                 }
1573
1574                 /* Test for inline data flag but no attr */
1575                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1576                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1577                         size_t size = 0;
1578                         errcode_t err;
1579                         int flags;
1580
1581                         flags = fs->flags;
1582                         if (failed_csum)
1583                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1584                         err = get_inline_data_ea_size(fs, ino, inode, &size);
1585                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1586                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1587
1588                         switch (err) {
1589                         case 0:
1590                                 /* Everything is awesome... */
1591                                 break;
1592                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1593                         case EXT2_ET_BAD_EA_HASH:
1594                         case EXT2_ET_BAD_EA_HEADER:
1595                         case EXT2_ET_EA_BAD_NAME_LEN:
1596                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1597                         case EXT2_ET_EA_KEY_NOT_FOUND:
1598                         case EXT2_ET_EA_NO_SPACE:
1599                         case EXT2_ET_MISSING_EA_FEATURE:
1600                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1601                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1602                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1603                         case EXT2_ET_NO_INLINE_DATA:
1604                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1605                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1606                         case EXT2_ET_EA_INODE_CORRUPTED:
1607                                 /* broken EA or no system.data EA; truncate */
1608                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1609                                                 &pctx)) {
1610                                         err = ext2fs_inode_size_set(fs, inode, 0);
1611                                         if (err) {
1612                                                 pctx.errcode = err;
1613                                                 ctx->flags |= E2F_FLAG_ABORT;
1614                                                 goto endit;
1615                                         }
1616                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1617                                         memset(&inode->i_block, 0,
1618                                                sizeof(inode->i_block));
1619                                         e2fsck_write_inode(ctx, ino, inode,
1620                                                            "pass1");
1621                                         failed_csum = 0;
1622                                 }
1623                                 break;
1624                         default:
1625                                 /* Some other kind of non-xattr error? */
1626                                 pctx.errcode = err;
1627                                 ctx->flags |= E2F_FLAG_ABORT;
1628                                 goto endit;
1629                         }
1630                 }
1631
1632                 /*
1633                  * Test for incorrect extent flag settings.
1634                  *
1635                  * On big-endian machines we must be careful:
1636                  * When the inode is read, the i_block array is not swapped
1637                  * if the extent flag is set.  Therefore if we are testing
1638                  * for or fixing a wrongly-set flag, we must potentially
1639                  * (un)swap before testing, or after fixing.
1640                  */
1641
1642                 /*
1643                  * In this case the extents flag was set when read, so
1644                  * extent_header_verify is ok.  If the inode is cleared,
1645                  * no need to swap... so no extra swapping here.
1646                  */
1647                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1648                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1649                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1650                         if ((ext2fs_extent_header_verify(inode->i_block,
1651                                                  sizeof(inode->i_block)) == 0) &&
1652                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1653                                 ext2fs_set_feature_extents(sb);
1654                                 ext2fs_mark_super_dirty(fs);
1655                                 extent_fs = 1;
1656                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1657                         clear_inode:
1658                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1659                                 if (ino == EXT2_BAD_INO)
1660                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1661                                                                  ino);
1662                                 /* skip FINISH_INODE_LOOP */
1663                                 continue;
1664                         }
1665                 }
1666
1667                 /*
1668                  * For big-endian machines:
1669                  * If the inode didn't have the extents flag set when it
1670                  * was read, then the i_blocks array was swapped.  To test
1671                  * as an extents header, we must swap it back first.
1672                  * IF we then set the extents flag, the entire i_block
1673                  * array must be un/re-swapped to make it proper extents data.
1674                  */
1675                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1676                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1677                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1678                     (LINUX_S_ISREG(inode->i_mode) ||
1679                      LINUX_S_ISDIR(inode->i_mode))) {
1680                         void *ehp;
1681 #ifdef WORDS_BIGENDIAN
1682                         __u32 tmp_block[EXT2_N_BLOCKS];
1683
1684                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1685                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1686                         ehp = tmp_block;
1687 #else
1688                         ehp = inode->i_block;
1689 #endif
1690                         if ((ext2fs_extent_header_verify(ehp,
1691                                          sizeof(inode->i_block)) == 0) &&
1692                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1693                                 inode->i_flags |= EXT4_EXTENTS_FL;
1694 #ifdef WORDS_BIGENDIAN
1695                                 memcpy(inode->i_block, tmp_block,
1696                                        sizeof(inode->i_block));
1697 #endif
1698                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1699                                 failed_csum = 0;
1700                         }
1701                 }
1702
1703                 if (ino == EXT2_BAD_INO) {
1704                         struct process_block_struct pb;
1705
1706                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1707                              inode->i_gid || inode->i_links_count ||
1708                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1709                              inode->i_file_acl) &&
1710                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1711                                 memset(inode, 0, sizeof(struct ext2_inode));
1712                                 e2fsck_write_inode(ctx, ino, inode,
1713                                                    "clear bad inode");
1714                                 failed_csum = 0;
1715                         }
1716
1717                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1718                                                           &pb.fs_meta_blocks);
1719                         if (pctx.errcode) {
1720                                 pctx.num = 4;
1721                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1722                                 ctx->flags |= E2F_FLAG_ABORT;
1723                                 goto endit;
1724                         }
1725                         pb.ino = EXT2_BAD_INO;
1726                         pb.num_blocks = pb.last_block = 0;
1727                         pb.last_db_block = -1;
1728                         pb.num_illegal_blocks = 0;
1729                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1730                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1731                         pb.inode = inode;
1732                         pb.pctx = &pctx;
1733                         pb.ctx = ctx;
1734                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1735                                      block_buf, process_bad_block, &pb);
1736                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1737                         if (pctx.errcode) {
1738                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1739                                 ctx->flags |= E2F_FLAG_ABORT;
1740                                 goto endit;
1741                         }
1742                         if (pb.bbcheck)
1743                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1744                                 ctx->flags |= E2F_FLAG_ABORT;
1745                                 goto endit;
1746                         }
1747                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1748                         clear_problem_context(&pctx);
1749                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1750                         continue;
1751                 } else if (ino == EXT2_ROOT_INO) {
1752                         /*
1753                          * Make sure the root inode is a directory; if
1754                          * not, offer to clear it.  It will be
1755                          * regenerated in pass #3.
1756                          */
1757                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1758                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1759                                         goto clear_inode;
1760                         }
1761                         /*
1762                          * If dtime is set, offer to clear it.  mke2fs
1763                          * version 0.2b created filesystems with the
1764                          * dtime field set for the root and lost+found
1765                          * directories.  We won't worry about
1766                          * /lost+found, since that can be regenerated
1767                          * easily.  But we will fix the root directory
1768                          * as a special case.
1769                          */
1770                         if (inode->i_dtime && inode->i_links_count) {
1771                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1772                                         inode->i_dtime = 0;
1773                                         e2fsck_write_inode(ctx, ino, inode,
1774                                                            "pass1");
1775                                         failed_csum = 0;
1776                                 }
1777                         }
1778                 } else if (ino == EXT2_JOURNAL_INO) {
1779                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1780                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1781                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1782                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1783                                                 &pctx)) {
1784                                         inode->i_mode = LINUX_S_IFREG;
1785                                         e2fsck_write_inode(ctx, ino, inode,
1786                                                            "pass1");
1787                                         failed_csum = 0;
1788                                 }
1789                                 check_blocks(ctx, &pctx, block_buf, NULL);
1790                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1791                                 continue;
1792                         }
1793                         if ((inode->i_links_count ||
1794                              inode->i_blocks || inode->i_block[0]) &&
1795                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1796                                         &pctx)) {
1797                                 memset(inode, 0, inode_size);
1798                                 ext2fs_icount_store(ctx->inode_link_info,
1799                                                     ino, 0);
1800                                 e2fsck_write_inode_full(ctx, ino, inode,
1801                                                         inode_size, "pass1");
1802                                 failed_csum = 0;
1803                         }
1804                 } else if (quota_inum_is_reserved(fs, ino)) {
1805                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1806                         if (ext2fs_has_feature_quota(fs->super) &&
1807                             quota_inum_is_super(fs->super, ino)) {
1808                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1809                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1810                                                         &pctx)) {
1811                                         inode->i_mode = LINUX_S_IFREG;
1812                                         e2fsck_write_inode(ctx, ino, inode,
1813                                                         "pass1");
1814                                         failed_csum = 0;
1815                                 }
1816                                 check_blocks(ctx, &pctx, block_buf, NULL);
1817                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1818                                 continue;
1819                         }
1820                         if ((inode->i_links_count ||
1821                              inode->i_blocks || inode->i_block[0]) &&
1822                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1823                                         &pctx)) {
1824                                 memset(inode, 0, inode_size);
1825                                 ext2fs_icount_store(ctx->inode_link_info,
1826                                                     ino, 0);
1827                                 e2fsck_write_inode_full(ctx, ino, inode,
1828                                                         inode_size, "pass1");
1829                                 failed_csum = 0;
1830                         }
1831                 } else if (ino == fs->super->s_orphan_file_inum) {
1832                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1833                         if (ext2fs_has_feature_orphan_file(fs->super)) {
1834                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1835                                     fix_problem(ctx, PR_1_ORPHAN_FILE_BAD_MODE,
1836                                                 &pctx)) {
1837                                         inode->i_mode = LINUX_S_IFREG;
1838                                         e2fsck_write_inode(ctx, ino, inode,
1839                                                            "pass1");
1840                                         failed_csum = 0;
1841                                 }
1842                                 check_blocks(ctx, &pctx, block_buf, NULL);
1843                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1844                                 continue;
1845                         }
1846                         if ((inode->i_links_count ||
1847                              inode->i_blocks || inode->i_block[0]) &&
1848                             fix_problem(ctx, PR_1_ORPHAN_FILE_NOT_CLEAR,
1849                                         &pctx)) {
1850                                 memset(inode, 0, inode_size);
1851                                 ext2fs_icount_store(ctx->inode_link_info, ino,
1852                                                     0);
1853                                 e2fsck_write_inode_full(ctx, ino, inode,
1854                                                         inode_size, "pass1");
1855                                 failed_csum = 0;
1856                         }
1857                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1858                         problem_t problem = 0;
1859
1860                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1861                         if (ino == EXT2_BOOT_LOADER_INO) {
1862                                 if (LINUX_S_ISDIR(inode->i_mode))
1863                                         problem = PR_1_RESERVED_BAD_MODE;
1864                         } else if (ino == EXT2_RESIZE_INO) {
1865                                 if (inode->i_mode &&
1866                                     !LINUX_S_ISREG(inode->i_mode))
1867                                         problem = PR_1_RESERVED_BAD_MODE;
1868                         } else {
1869                                 if (inode->i_mode != 0)
1870                                         problem = PR_1_RESERVED_BAD_MODE;
1871                         }
1872                         if (problem) {
1873                                 if (fix_problem(ctx, problem, &pctx)) {
1874                                         inode->i_mode = 0;
1875                                         e2fsck_write_inode(ctx, ino, inode,
1876                                                            "pass1");
1877                                         failed_csum = 0;
1878                                 }
1879                         }
1880                         check_blocks(ctx, &pctx, block_buf, NULL);
1881                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1882                         continue;
1883                 }
1884
1885                 if (!inode->i_links_count) {
1886                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1887                         continue;
1888                 }
1889                 /*
1890                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1891                  * deleted files.  Oops.
1892                  *
1893                  * Since all new ext2 implementations get this right,
1894                  * we now assume that the case of non-zero
1895                  * i_links_count and non-zero dtime means that we
1896                  * should keep the file, not delete it.
1897                  *
1898                  */
1899                 if (inode->i_dtime) {
1900                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1901                                 inode->i_dtime = 0;
1902                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1903                                 failed_csum = 0;
1904                         }
1905                 }
1906
1907                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1908                 switch (fs->super->s_creator_os) {
1909                     case EXT2_OS_HURD:
1910                         frag = inode->osd2.hurd2.h_i_frag;
1911                         fsize = inode->osd2.hurd2.h_i_fsize;
1912                         break;
1913                     default:
1914                         frag = fsize = 0;
1915                 }
1916
1917                 if (inode->i_faddr || frag || fsize ||
1918                     (!ext2fs_has_feature_largedir(fs->super) &&
1919                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1920                         mark_inode_bad(ctx, ino);
1921                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1922                     !ext2fs_has_feature_64bit(fs->super) &&
1923                     inode->osd2.linux2.l_i_file_acl_high != 0)
1924                         mark_inode_bad(ctx, ino);
1925                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1926                     !ext2fs_has_feature_huge_file(fs->super) &&
1927                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1928                         mark_inode_bad(ctx, ino);
1929                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1930                         if (imagic_fs) {
1931                                 if (!ctx->inode_imagic_map)
1932                                         alloc_imagic_map(ctx);
1933                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1934                                                          ino);
1935                         } else {
1936                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1937                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1938                                         e2fsck_write_inode(ctx, ino,
1939                                                            inode, "pass1");
1940                                         failed_csum = 0;
1941                                 }
1942                         }
1943                 }
1944
1945                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1946                 check_is_really_dir(ctx, &pctx, block_buf);
1947
1948                 /*
1949                  * ext2fs_inode_has_valid_blocks2 does not actually look
1950                  * at i_block[] values, so not endian-sensitive here.
1951                  */
1952                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1953                     LINUX_S_ISLNK(inode->i_mode) &&
1954                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1955                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1956                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1957                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1958                         failed_csum = 0;
1959                 }
1960
1961                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1962                     add_encrypted_file(ctx, &pctx) < 0)
1963                         goto clear_inode;
1964
1965                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1966                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1967
1968                 if (LINUX_S_ISDIR(inode->i_mode)) {
1969                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1970                         e2fsck_add_dir_info(ctx, ino, 0);
1971                         ctx->fs_directory_count++;
1972                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1973                                 add_casefolded_dir(ctx, ino);
1974                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1975                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1976                         ctx->fs_regular_count++;
1977                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1978                            e2fsck_pass1_check_device_inode(fs, inode)) {
1979                         check_extents_inlinedata(ctx, &pctx);
1980                         check_immutable(ctx, &pctx);
1981                         check_size(ctx, &pctx);
1982                         ctx->fs_chardev_count++;
1983                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1984                            e2fsck_pass1_check_device_inode(fs, inode)) {
1985                         check_extents_inlinedata(ctx, &pctx);
1986                         check_immutable(ctx, &pctx);
1987                         check_size(ctx, &pctx);
1988                         ctx->fs_blockdev_count++;
1989                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1990                            e2fsck_pass1_check_symlink(fs, ino, inode,
1991                                                       block_buf)) {
1992                         check_immutable(ctx, &pctx);
1993                         ctx->fs_symlinks_count++;
1994                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1995                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1996                                 continue;
1997                         } else if (ext2fs_is_fast_symlink(inode)) {
1998                                 ctx->fs_fast_symlinks_count++;
1999                                 check_blocks(ctx, &pctx, block_buf,
2000                                              &ea_ibody_quota);
2001                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2002                                 continue;
2003                         }
2004                 }
2005                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
2006                          e2fsck_pass1_check_device_inode(fs, inode)) {
2007                         check_extents_inlinedata(ctx, &pctx);
2008                         check_immutable(ctx, &pctx);
2009                         check_size(ctx, &pctx);
2010                         ctx->fs_fifo_count++;
2011                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
2012                            e2fsck_pass1_check_device_inode(fs, inode)) {
2013                         check_extents_inlinedata(ctx, &pctx);
2014                         check_immutable(ctx, &pctx);
2015                         check_size(ctx, &pctx);
2016                         ctx->fs_sockets_count++;
2017                 } else
2018                         mark_inode_bad(ctx, ino);
2019                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2020                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
2021                         if (inode->i_block[EXT2_IND_BLOCK])
2022                                 ctx->fs_ind_count++;
2023                         if (inode->i_block[EXT2_DIND_BLOCK])
2024                                 ctx->fs_dind_count++;
2025                         if (inode->i_block[EXT2_TIND_BLOCK])
2026                                 ctx->fs_tind_count++;
2027                 }
2028                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2029                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2030                     (inode->i_block[EXT2_IND_BLOCK] ||
2031                      inode->i_block[EXT2_DIND_BLOCK] ||
2032                      inode->i_block[EXT2_TIND_BLOCK] ||
2033                      ext2fs_file_acl_block(fs, inode))) {
2034                         struct process_inode_block *itp;
2035
2036                         itp = &inodes_to_process[process_inode_count];
2037                         itp->ino = ino;
2038                         itp->ea_ibody_quota = ea_ibody_quota;
2039                         if (inode_size < sizeof(struct ext2_inode_large))
2040                                 memcpy(&itp->inode, inode, inode_size);
2041                         else
2042                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2043                         process_inode_count++;
2044                 } else
2045                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2046
2047                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2048
2049                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2050                         goto endit;
2051
2052                 if (process_inode_count >= ctx->process_inode_size) {
2053                         process_inodes(ctx, block_buf);
2054
2055                         if (e2fsck_should_abort(ctx))
2056                                 goto endit;
2057                 }
2058         }
2059         process_inodes(ctx, block_buf);
2060         ext2fs_close_inode_scan(scan);
2061         scan = NULL;
2062
2063         reserve_block_for_root_repair(ctx);
2064         reserve_block_for_lnf_repair(ctx);
2065
2066         /*
2067          * If any extended attribute blocks' reference counts need to
2068          * be adjusted, either up (ctx->refcount_extra), or down
2069          * (ctx->refcount), then fix them.
2070          */
2071         if (ctx->refcount) {
2072                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2073                 ea_refcount_free(ctx->refcount);
2074                 ctx->refcount = 0;
2075         }
2076         if (ctx->refcount_extra) {
2077                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2078                                         block_buf, +1);
2079                 ea_refcount_free(ctx->refcount_extra);
2080                 ctx->refcount_extra = 0;
2081         }
2082
2083         if (ctx->ea_block_quota_blocks) {
2084                 ea_refcount_free(ctx->ea_block_quota_blocks);
2085                 ctx->ea_block_quota_blocks = 0;
2086         }
2087
2088         if (ctx->ea_block_quota_inodes) {
2089                 ea_refcount_free(ctx->ea_block_quota_inodes);
2090                 ctx->ea_block_quota_inodes = 0;
2091         }
2092
2093         if (ctx->invalid_bitmaps)
2094                 handle_fs_bad_blocks(ctx);
2095
2096         /* We don't need the block_ea_map any more */
2097         if (ctx->block_ea_map) {
2098                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2099                 ctx->block_ea_map = 0;
2100         }
2101
2102         /* We don't need the encryption policy => ID map any more */
2103         destroy_encryption_policy_map(ctx);
2104
2105         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2106                 clear_problem_context(&pctx);
2107                 pctx.errcode = ext2fs_create_resize_inode(fs);
2108                 if (pctx.errcode) {
2109                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2110                                          &pctx)) {
2111                                 ctx->flags |= E2F_FLAG_ABORT;
2112                                 goto endit;
2113                         }
2114                         pctx.errcode = 0;
2115                 }
2116                 if (!pctx.errcode) {
2117                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2118                                           "recreate inode");
2119                         inode->i_mtime = ctx->now;
2120                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2121                                            "recreate inode");
2122                 }
2123                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2124         }
2125
2126         if (ctx->flags & E2F_FLAG_RESTART) {
2127                 /*
2128                  * Only the master copy of the superblock and block
2129                  * group descriptors are going to be written during a
2130                  * restart, so set the superblock to be used to be the
2131                  * master superblock.
2132                  */
2133                 ctx->use_superblock = 0;
2134                 goto endit;
2135         }
2136
2137         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2138                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2139                         ext2fs_set_feature_largedir(fs->super);
2140                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2141                         ext2fs_mark_super_dirty(fs);
2142                 }
2143                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2144                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2145                         ext2fs_update_dynamic_rev(fs);
2146                         ext2fs_mark_super_dirty(fs);
2147                 }
2148         }
2149
2150         if (ctx->block_dup_map) {
2151                 if (ctx->options & E2F_OPT_PREEN) {
2152                         clear_problem_context(&pctx);
2153                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2154                 }
2155                 e2fsck_pass1_dupblocks(ctx, block_buf);
2156         }
2157         ctx->flags |= E2F_FLAG_ALLOC_OK;
2158 endit:
2159         e2fsck_use_inode_shortcuts(ctx, 0);
2160         ext2fs_free_mem(&inodes_to_process);
2161         inodes_to_process = 0;
2162
2163         if (scan)
2164                 ext2fs_close_inode_scan(scan);
2165         if (block_buf)
2166                 ext2fs_free_mem(&block_buf);
2167         if (inode)
2168                 ext2fs_free_mem(&inode);
2169
2170         /*
2171          * The l+f inode may have been cleared, so zap it now and
2172          * later passes will recalculate it if necessary
2173          */
2174         ctx->lost_and_found = 0;
2175
2176         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2177                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2178         else
2179                 ctx->invalid_bitmaps++;
2180 }
2181
2182 #ifdef HAVE_PTHREAD
2183 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2184                                           ext2fs_generic_bitmap *dest)
2185 {
2186         errcode_t ret;
2187
2188         ret = ext2fs_copy_bitmap(*src, dest);
2189         if (ret)
2190                 return ret;
2191
2192         (*dest)->fs = fs;
2193         ext2fs_free_generic_bmap(*src);
2194         *src = NULL;
2195
2196         return 0;
2197 }
2198
2199
2200 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2201                                       ext2_filsys src)
2202 {
2203         errcode_t       retval;
2204
2205         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2206         if (dest->dblist)
2207                 dest->dblist->fs = dest;
2208         if (src->block_map) {
2209                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2210                                                   &dest->block_map);
2211                 if (retval)
2212                         return retval;
2213         }
2214         if (src->inode_map) {
2215                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2216                                                   &dest->inode_map);
2217                 if (retval)
2218                         return retval;
2219         }
2220
2221         if (src->badblocks) {
2222                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2223                 if (retval)
2224                         return retval;
2225         }
2226
2227         /* disable it for now */
2228         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2229         retval = ext2fs_open_channel(dest, src_context->io_options,
2230                                      src_context->io_manager,
2231                                      src_context->openfs_flags,
2232                                      src->io->block_size);
2233         if (retval)
2234                 return retval;
2235
2236         /* Block size might not be default */
2237         io_channel_set_blksize(dest->io, src->io->block_size);
2238         ehandler_init(dest->io);
2239
2240         assert(dest->io->magic == src->io->magic);
2241         assert(dest->io->manager == src->io->manager);
2242         assert(strcmp(dest->io->name, src->io->name) == 0);
2243         assert(dest->io->block_size == src->io->block_size);
2244         assert(dest->io->read_error == src->io->read_error);
2245         assert(dest->io->write_error == src->io->write_error);
2246         assert(dest->io->refcount == src->io->refcount);
2247         assert(dest->io->flags == src->io->flags);
2248         assert(dest->io->app_data == dest);
2249         assert(src->io->app_data == src);
2250         assert(dest->io->align == src->io->align);
2251
2252         /* The data should be written to disk immediately */
2253         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2254         /* icache will be rebuilt if needed, so do not copy from @src */
2255         src->icache = NULL;
2256         return 0;
2257 }
2258
2259 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2260 {
2261         struct ext2_inode_cache *icache = dest->icache;
2262         errcode_t retval = 0;
2263         io_channel dest_io;
2264         io_channel dest_image_io;
2265
2266         dest_io = dest->io;
2267         dest_image_io = dest->image_io;
2268
2269         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2270         dest->io = dest_io;
2271         dest->image_io = dest_image_io;
2272         dest->icache = icache;
2273         if (dest->dblist)
2274                 dest->dblist->fs = dest;
2275         if (src->inode_map) {
2276                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2277                                                   &dest->inode_map);
2278                 if (retval)
2279                         return retval;
2280         }
2281         if (src->block_map) {
2282                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2283                                                   &dest->block_map);
2284                 if (retval)
2285                         return retval;
2286         }
2287
2288         if (src->icache) {
2289                 ext2fs_free_inode_cache(src->icache);
2290                 src->icache = NULL;
2291         }
2292
2293         if (src->badblocks) {
2294                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2295
2296                 ext2fs_badblocks_list_free(src->badblocks);
2297                 src->badblocks = NULL;
2298         }
2299
2300         io_channel_close(src->io);
2301         return retval;
2302 }
2303
2304 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2305                                              int thread_index, int num_threads)
2306 {
2307         errcode_t               retval;
2308         e2fsck_t                thread_context;
2309         ext2_filsys             thread_fs;
2310         ext2_filsys             global_fs = global_ctx->fs;
2311         struct e2fsck_thread    *tinfo;
2312         dgrp_t                  average_group;
2313
2314         assert(global_ctx->inode_used_map == NULL);
2315         assert(global_ctx->inode_dir_map == NULL);
2316         assert(global_ctx->inode_bb_map == NULL);
2317         assert(global_ctx->inode_imagic_map == NULL);
2318         assert(global_ctx->inode_reg_map == NULL);
2319         assert(global_ctx->inodes_to_rebuild == NULL);
2320
2321         assert(global_ctx->block_found_map == NULL);
2322         assert(global_ctx->block_dup_map == NULL);
2323         assert(global_ctx->block_ea_map == NULL);
2324         assert(global_ctx->block_metadata_map == NULL);
2325         assert(global_ctx->fs->dblist == NULL);
2326
2327         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2328         if (retval) {
2329                 com_err(global_ctx->program_name, retval, "while allocating memory");
2330                 return retval;
2331         }
2332         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2333         thread_context->global_ctx = global_ctx;
2334
2335         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2336         if (retval) {
2337                 com_err(global_ctx->program_name, retval, "while allocating memory");
2338                 goto out_context;
2339         }
2340
2341         io_channel_flush_cleanup(global_fs->io);
2342         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2343         if (retval) {
2344                 com_err(global_ctx->program_name, retval, "while copying fs");
2345                 goto out_fs;
2346         }
2347         thread_fs->priv_data = thread_context;
2348
2349         thread_context->thread_info.et_thread_index = thread_index;
2350         set_up_logging(thread_context);
2351
2352         /*
2353          * Distribute work to multiple threads:
2354          * Each thread work on fs->group_desc_count / nthread groups.
2355          */
2356         tinfo = &thread_context->thread_info;
2357         average_group = thread_fs->group_desc_count / num_threads;
2358         if (average_group == 0)
2359                 average_group = 1;
2360         tinfo->et_group_start = average_group * thread_index;
2361         if (thread_index == num_threads - 1)
2362                 tinfo->et_group_end = thread_fs->group_desc_count;
2363         else
2364                 tinfo->et_group_end = average_group * (thread_index + 1);
2365         tinfo->et_group_next = tinfo->et_group_start;
2366
2367         thread_context->fs = thread_fs;
2368         *thread_ctx = thread_context;
2369         return 0;
2370 out_fs:
2371         ext2fs_free_mem(&thread_fs);
2372 out_context:
2373         ext2fs_free_mem(&thread_context);
2374         return retval;
2375 }
2376
2377 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2378 {
2379         errcode_t        retval;
2380         int              flags = global_ctx->flags;
2381         ext2_filsys      thread_fs = thread_ctx->fs;
2382         ext2_filsys      global_fs = global_ctx->fs;
2383         FILE            *global_logf = global_ctx->logf;
2384         FILE            *global_problem_logf = global_ctx->problem_logf;
2385 #ifdef HAVE_SETJMP_H
2386         jmp_buf          old_jmp;
2387
2388         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2389 #endif
2390         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2391 #ifdef HAVE_SETJMP_H
2392         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2393 #endif
2394         /* Keep the global singal flags*/
2395         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2396                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2397
2398         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2399         if (retval) {
2400                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2401                 return retval;
2402         }
2403         global_fs->priv_data = global_ctx;
2404         global_ctx->fs = global_fs;
2405         global_ctx->logf = global_logf;
2406         global_ctx->problem_logf = global_problem_logf;
2407
2408         if (thread_ctx->inode_used_map) {
2409                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2410                                         &thread_ctx->inode_used_map,
2411                                         &global_ctx->inode_used_map);
2412                 if (retval)
2413                         return retval;
2414         }
2415         if (thread_ctx->inode_bad_map) {
2416                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2417                                         &thread_ctx->inode_bad_map,
2418                                         &global_ctx->inode_bad_map);
2419                 if (retval)
2420                         return retval;
2421         }
2422         if (thread_ctx->inode_dir_map) {
2423                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2424                                         &thread_ctx->inode_dir_map,
2425                                         &global_ctx->inode_dir_map);
2426                 if (retval)
2427                         return retval;
2428         }
2429         if (thread_ctx->inode_bb_map) {
2430                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2431                                         &thread_ctx->inode_bb_map,
2432                                         &global_ctx->inode_bb_map);
2433                 if (retval)
2434                         return retval;
2435         }
2436         if (thread_ctx->inode_imagic_map) {
2437                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2438                                         &thread_ctx->inode_imagic_map,
2439                                         &global_ctx->inode_imagic_map);
2440                 if (retval)
2441                         return retval;
2442         }
2443         if (thread_ctx->inode_reg_map) {
2444                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2445                                         &thread_ctx->inode_reg_map,
2446                                         &global_ctx->inode_reg_map);
2447                 if (retval)
2448                         return retval;
2449         }
2450         if (thread_ctx->inodes_to_rebuild) {
2451                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2452                                         &thread_ctx->inodes_to_rebuild,
2453                                         &global_ctx->inodes_to_rebuild);
2454                 if (retval)
2455                         return retval;
2456         }
2457         if (thread_ctx->block_found_map) {
2458                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2459                                         &thread_ctx->block_found_map,
2460                                         &global_ctx->block_found_map);
2461                 if (retval)
2462                         return retval;
2463         }
2464         if (thread_ctx->block_dup_map) {
2465                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2466                                         &thread_ctx->block_dup_map,
2467                                         &global_ctx->block_dup_map);
2468                 if (retval)
2469                         return retval;
2470         }
2471         if (thread_ctx->block_ea_map) {
2472                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2473                                         &thread_ctx->block_ea_map,
2474                                         &global_ctx->block_ea_map);
2475                 if (retval)
2476                         return retval;
2477         }
2478         if (thread_ctx->block_metadata_map) {
2479                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2480                                         &thread_ctx->block_metadata_map,
2481                                         &global_ctx->block_metadata_map);
2482                 if (retval)
2483                         return retval;
2484         }
2485
2486         return 0;
2487 }
2488
2489 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2490 {
2491         errcode_t       retval;
2492
2493         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2494         ext2fs_free_mem(&thread_ctx->fs);
2495         if (thread_ctx->logf)
2496                 fclose(thread_ctx->logf);
2497         if (thread_ctx->problem_logf) {
2498                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2499                 fclose(thread_ctx->problem_logf);
2500         }
2501         ext2fs_free_mem(&thread_ctx);
2502
2503         return retval;
2504 }
2505
2506 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2507                                       int num_threads, e2fsck_t global_ctx)
2508 {
2509         errcode_t                        rc;
2510         errcode_t                        ret = 0;
2511         int                              i;
2512         struct e2fsck_thread_info       *pinfo;
2513
2514         for (i = 0; i < num_threads; i++) {
2515                 pinfo = &infos[i];
2516
2517                 if (!pinfo->eti_started)
2518                         continue;
2519
2520                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2521                 if (rc) {
2522                         com_err(global_ctx->program_name, rc,
2523                                 _("while joining thread\n"));
2524                         if (ret == 0)
2525                                 ret = rc;
2526                 }
2527                 e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2528         }
2529         free(infos);
2530
2531         return ret;
2532 }
2533
2534 static void *e2fsck_pass1_thread(void *arg)
2535 {
2536         struct e2fsck_thread_info       *info = arg;
2537         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2538
2539 #ifdef HAVE_SETJMP_H
2540         /*
2541          * When fatal_error() happens, jump to here. The thread
2542          * context's flags will be saved, but its abort_loc will
2543          * be overwritten by original jump buffer for the later
2544          * tests.
2545          */
2546         if (setjmp(thread_ctx->abort_loc)) {
2547                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2548                 goto out;
2549         }
2550         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2551 #endif
2552
2553         e2fsck_pass1_run(thread_ctx);
2554
2555 out:
2556         return NULL;
2557 }
2558
2559 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2560                                       int num_threads, e2fsck_t global_ctx)
2561 {
2562         struct e2fsck_thread_info       *infos;
2563         pthread_attr_t                   attr;
2564         errcode_t                        retval;
2565         errcode_t                        ret;
2566         struct e2fsck_thread_info       *tmp_pinfo;
2567         int                              i;
2568         e2fsck_t                         thread_ctx;
2569
2570         retval = pthread_attr_init(&attr);
2571         if (retval) {
2572                 com_err(global_ctx->program_name, retval,
2573                         _("while setting pthread attribute\n"));
2574                 return retval;
2575         }
2576
2577         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2578         if (infos == NULL) {
2579                 retval = -ENOMEM;
2580                 com_err(global_ctx->program_name, retval,
2581                         _("while allocating memory for threads\n"));
2582                 pthread_attr_destroy(&attr);
2583                 return retval;
2584         }
2585
2586         for (i = 0; i < num_threads; i++) {
2587                 tmp_pinfo = &infos[i];
2588                 tmp_pinfo->eti_thread_index = i;
2589                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
2590                                                      i, num_threads);
2591                 if (retval) {
2592                         com_err(global_ctx->program_name, retval,
2593                                 _("while preparing pass1 thread\n"));
2594                         break;
2595                 }
2596                 tmp_pinfo->eti_thread_ctx = thread_ctx;
2597
2598                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
2599                                         &e2fsck_pass1_thread, tmp_pinfo);
2600                 if (retval) {
2601                         com_err(global_ctx->program_name, retval,
2602                                 _("while creating thread\n"));
2603                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
2604                         break;
2605                 }
2606
2607                 tmp_pinfo->eti_started = 1;
2608         }
2609
2610         /* destroy the thread attribute object, since it is no longer needed */
2611         ret = pthread_attr_destroy(&attr);
2612         if (ret) {
2613                 com_err(global_ctx->program_name, ret,
2614                         _("while destroying thread attribute\n"));
2615                 if (retval == 0)
2616                         retval = ret;
2617         }
2618
2619         if (retval) {
2620                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2621                 return retval;
2622         }
2623         *pinfo = infos;
2624         return 0;
2625 }
2626
2627 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
2628 {
2629         struct e2fsck_thread_info       *infos = NULL;
2630         int                              num_threads = 1;
2631         errcode_t                        retval;
2632
2633         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
2634         if (retval) {
2635                 com_err(global_ctx->program_name, retval,
2636                         _("while starting pass1 threads\n"));
2637                 goto out_abort;
2638         }
2639
2640         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2641         if (retval) {
2642                 com_err(global_ctx->program_name, retval,
2643                         _("while joining pass1 threads\n"));
2644                 goto out_abort;
2645         }
2646         return;
2647 out_abort:
2648         global_ctx->flags |= E2F_FLAG_ABORT;
2649         return;
2650 }
2651 #endif
2652
2653 void e2fsck_pass1(e2fsck_t ctx)
2654 {
2655
2656 #ifdef HAVE_PTHREAD
2657         e2fsck_pass1_multithread(ctx);
2658 #else
2659         e2fsck_pass1_run(ctx);
2660 #endif
2661 }
2662
2663 #undef FINISH_INODE_LOOP
2664
2665 /*
2666  * When the inode_scan routines call this callback at the end of the
2667  * glock group, call process_inodes.
2668  */
2669 static errcode_t scan_callback(ext2_filsys fs,
2670                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2671                                dgrp_t group, void * priv_data)
2672 {
2673         struct scan_callback_struct *scan_struct;
2674         e2fsck_t ctx;
2675         struct e2fsck_thread *tinfo;
2676
2677         scan_struct = (struct scan_callback_struct *) priv_data;
2678         ctx = scan_struct->ctx;
2679
2680         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2681
2682         if (ctx->progress)
2683                 if ((ctx->progress)(ctx, 1, group+1,
2684                                     ctx->fs->group_desc_count))
2685                         return EXT2_ET_CANCEL_REQUESTED;
2686
2687 #ifdef HAVE_PTHREAD
2688         if (ctx->global_ctx) {
2689                 tinfo = &ctx->thread_info;
2690                 tinfo->et_group_next++;
2691                 if (tinfo->et_group_next >= tinfo->et_group_end)
2692                         return EXT2_ET_SCAN_FINISHED;
2693         }
2694 #endif
2695
2696         return 0;
2697 }
2698
2699 /*
2700  * Process the inodes in the "inodes to process" list.
2701  */
2702 static void process_inodes(e2fsck_t ctx, char *block_buf)
2703 {
2704         int                     i;
2705         struct ext2_inode       *old_stashed_inode;
2706         ext2_ino_t              old_stashed_ino;
2707         const char              *old_operation;
2708         char                    buf[80];
2709         struct problem_context  pctx;
2710
2711 #if 0
2712         printf("begin process_inodes: ");
2713 #endif
2714         if (process_inode_count == 0)
2715                 return;
2716         old_operation = ehandler_operation(0);
2717         old_stashed_inode = ctx->stashed_inode;
2718         old_stashed_ino = ctx->stashed_ino;
2719         qsort(inodes_to_process, process_inode_count,
2720                       sizeof(struct process_inode_block), process_inode_cmp);
2721         clear_problem_context(&pctx);
2722         for (i=0; i < process_inode_count; i++) {
2723                 pctx.inode = ctx->stashed_inode =
2724                         (struct ext2_inode *) &inodes_to_process[i].inode;
2725                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2726
2727 #if 0
2728                 printf("%u ", pctx.ino);
2729 #endif
2730                 sprintf(buf, _("reading indirect blocks of inode %u"),
2731                         pctx.ino);
2732                 ehandler_operation(buf);
2733                 check_blocks(ctx, &pctx, block_buf,
2734                              &inodes_to_process[i].ea_ibody_quota);
2735                 if (e2fsck_should_abort(ctx))
2736                         break;
2737         }
2738         ctx->stashed_inode = old_stashed_inode;
2739         ctx->stashed_ino = old_stashed_ino;
2740         process_inode_count = 0;
2741 #if 0
2742         printf("end process inodes\n");
2743 #endif
2744         ehandler_operation(old_operation);
2745 }
2746
2747 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2748 {
2749         const struct process_inode_block *ib_a =
2750                 (const struct process_inode_block *) a;
2751         const struct process_inode_block *ib_b =
2752                 (const struct process_inode_block *) b;
2753         int     ret;
2754
2755         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2756                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2757         if (ret == 0)
2758                 /*
2759                  * We only call process_inodes() for non-extent
2760                  * inodes, so it's OK to pass NULL to
2761                  * ext2fs_file_acl_block() here.
2762                  */
2763                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2764                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2765         if (ret == 0)
2766                 ret = ib_a->ino - ib_b->ino;
2767         return ret;
2768 }
2769
2770 /*
2771  * Mark an inode as being bad in some what
2772  */
2773 static void mark_inode_bad(e2fsck_t ctx, ext2_ino_t ino)
2774 {
2775         struct          problem_context pctx;
2776
2777         if (!ctx->inode_bad_map) {
2778                 clear_problem_context(&pctx);
2779
2780                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2781                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2782                                 "inode_bad_map", &ctx->inode_bad_map);
2783                 if (pctx.errcode) {
2784                         pctx.num = 3;
2785                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2786                         /* Should never get here */
2787                         ctx->flags |= E2F_FLAG_ABORT;
2788                         return;
2789                 }
2790         }
2791         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2792 }
2793
2794 static void add_casefolded_dir(e2fsck_t ctx, ext2_ino_t ino)
2795 {
2796         struct          problem_context pctx;
2797
2798         if (!ctx->casefolded_dirs) {
2799                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2800                 if (pctx.errcode)
2801                         goto error;
2802         }
2803         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2804         if (pctx.errcode == 0)
2805                 return;
2806 error:
2807         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2808         /* Should never get here */
2809         ctx->flags |= E2F_FLAG_ABORT;
2810 }
2811
2812 /*
2813  * This procedure will allocate the inode "bb" (badblock) map table
2814  */
2815 static void alloc_bb_map(e2fsck_t ctx)
2816 {
2817         struct          problem_context pctx;
2818
2819         clear_problem_context(&pctx);
2820         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2821                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2822                         "inode_bb_map", &ctx->inode_bb_map);
2823         if (pctx.errcode) {
2824                 pctx.num = 4;
2825                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2826                 /* Should never get here */
2827                 ctx->flags |= E2F_FLAG_ABORT;
2828                 return;
2829         }
2830 }
2831
2832 /*
2833  * This procedure will allocate the inode imagic table
2834  */
2835 static void alloc_imagic_map(e2fsck_t ctx)
2836 {
2837         struct          problem_context pctx;
2838
2839         clear_problem_context(&pctx);
2840         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2841                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2842                         "inode_imagic_map", &ctx->inode_imagic_map);
2843         if (pctx.errcode) {
2844                 pctx.num = 5;
2845                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2846                 /* Should never get here */
2847                 ctx->flags |= E2F_FLAG_ABORT;
2848                 return;
2849         }
2850 }
2851
2852 /*
2853  * Marks a block as in use, setting the dup_map if it's been set
2854  * already.  Called by process_block and process_bad_block.
2855  *
2856  * WARNING: Assumes checks have already been done to make sure block
2857  * is valid.  This is true in both process_block and process_bad_block.
2858  */
2859 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2860 {
2861         struct          problem_context pctx;
2862
2863         clear_problem_context(&pctx);
2864
2865         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2866                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2867                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2868                         return;
2869                 }
2870                 if (!ctx->block_dup_map) {
2871                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2872                                         _("multiply claimed block map"),
2873                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2874                                         &ctx->block_dup_map);
2875                         if (pctx.errcode) {
2876                                 pctx.num = 3;
2877                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2878                                             &pctx);
2879                                 /* Should never get here */
2880                                 ctx->flags |= E2F_FLAG_ABORT;
2881                                 return;
2882                         }
2883                 }
2884                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2885         } else {
2886                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2887         }
2888 }
2889
2890 /*
2891  * When cluster size is greater than one block, it is caller's responsibility
2892  * to make sure block parameter starts at a cluster boundary.
2893  */
2894 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2895                                       unsigned int num)
2896 {
2897         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2898                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2899         else {
2900                 unsigned int i;
2901
2902                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2903                         mark_block_used(ctx, block + i);
2904         }
2905 }
2906
2907 /*
2908  * Adjust the extended attribute block's reference counts at the end
2909  * of pass 1, either by subtracting out references for EA blocks that
2910  * are still referenced in ctx->refcount, or by adding references for
2911  * EA blocks that had extra references as accounted for in
2912  * ctx->refcount_extra.
2913  */
2914 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2915                                     char *block_buf, int adjust_sign)
2916 {
2917         struct ext2_ext_attr_header     *header;
2918         struct problem_context          pctx;
2919         ext2_filsys                     fs = ctx->fs;
2920         blk64_t                         blk;
2921         __u32                           should_be;
2922         ea_value_t                      count;
2923
2924         clear_problem_context(&pctx);
2925
2926         ea_refcount_intr_begin(refcount);
2927         while (1) {
2928                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2929                         break;
2930                 pctx.blk = blk;
2931                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2932                                                      pctx.ino);
2933                 if (pctx.errcode) {
2934                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2935                         return;
2936                 }
2937                 header = (struct ext2_ext_attr_header *) block_buf;
2938                 pctx.blkcount = header->h_refcount;
2939                 should_be = header->h_refcount + adjust_sign * (int)count;
2940                 pctx.num = should_be;
2941                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2942                         header->h_refcount = should_be;
2943                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2944                                                              block_buf,
2945                                                              pctx.ino);
2946                         if (pctx.errcode) {
2947                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2948                                             &pctx);
2949                                 continue;
2950                         }
2951                 }
2952         }
2953 }
2954
2955 /*
2956  * Handle processing the extended attribute blocks
2957  */
2958 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2959                            char *block_buf, struct ea_quota *ea_block_quota)
2960 {
2961         ext2_filsys fs = ctx->fs;
2962         ext2_ino_t      ino = pctx->ino;
2963         struct ext2_inode *inode = pctx->inode;
2964         blk64_t         blk;
2965         char *          end;
2966         struct ext2_ext_attr_header *header;
2967         struct ext2_ext_attr_entry *first, *entry;
2968         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2969         __u64           quota_inodes = 0;
2970         region_t        region = 0;
2971         int             failed_csum = 0;
2972
2973         ea_block_quota->blocks = 0;
2974         ea_block_quota->inodes = 0;
2975
2976         blk = ext2fs_file_acl_block(fs, inode);
2977         if (blk == 0)
2978                 return 0;
2979
2980         /*
2981          * If the Extended attribute flag isn't set, then a non-zero
2982          * file acl means that the inode is corrupted.
2983          *
2984          * Or if the extended attribute block is an invalid block,
2985          * then the inode is also corrupted.
2986          */
2987         if (!ext2fs_has_feature_xattr(fs->super) ||
2988             (blk < fs->super->s_first_data_block) ||
2989             (blk >= ext2fs_blocks_count(fs->super))) {
2990                 mark_inode_bad(ctx, ino);
2991                 return 0;
2992         }
2993
2994         /* If ea bitmap hasn't been allocated, create it */
2995         if (!ctx->block_ea_map) {
2996                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2997                                         _("ext attr block map"),
2998                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2999                                         &ctx->block_ea_map);
3000                 if (pctx->errcode) {
3001                         pctx->num = 2;
3002                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
3003                         ctx->flags |= E2F_FLAG_ABORT;
3004                         return 0;
3005                 }
3006         }
3007
3008         /* Create the EA refcount structure if necessary */
3009         if (!ctx->refcount) {
3010                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
3011                 if (pctx->errcode) {
3012                         pctx->num = 1;
3013                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3014                         ctx->flags |= E2F_FLAG_ABORT;
3015                         return 0;
3016                 }
3017         }
3018
3019 #if 0
3020         /* Debugging text */
3021         printf("Inode %u has EA block %u\n", ino, blk);
3022 #endif
3023
3024         /* Have we seen this EA block before? */
3025         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3026                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3027                 ea_block_quota->inodes = 0;
3028
3029                 if (ctx->ea_block_quota_blocks) {
3030                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3031                                           &quota_blocks);
3032                         if (quota_blocks)
3033                                 ea_block_quota->blocks = quota_blocks;
3034                 }
3035
3036                 if (ctx->ea_block_quota_inodes)
3037                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3038                                           &ea_block_quota->inodes);
3039
3040                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3041                         return 1;
3042                 /* Ooops, this EA was referenced more than it stated */
3043                 if (!ctx->refcount_extra) {
3044                         pctx->errcode = ea_refcount_create(0,
3045                                            &ctx->refcount_extra);
3046                         if (pctx->errcode) {
3047                                 pctx->num = 2;
3048                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3049                                 ctx->flags |= E2F_FLAG_ABORT;
3050                                 return 0;
3051                         }
3052                 }
3053                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3054                 return 1;
3055         }
3056
3057         /*
3058          * OK, we haven't seen this EA block yet.  So we need to
3059          * validate it
3060          */
3061         pctx->blk = blk;
3062         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3063         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3064                 pctx->errcode = 0;
3065                 failed_csum = 1;
3066         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3067                 pctx->errcode = 0;
3068
3069         if (pctx->errcode &&
3070             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3071                 pctx->errcode = 0;
3072                 goto clear_extattr;
3073         }
3074         header = (struct ext2_ext_attr_header *) block_buf;
3075         pctx->blk = ext2fs_file_acl_block(fs, inode);
3076         if (((ctx->ext_attr_ver == 1) &&
3077              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3078             ((ctx->ext_attr_ver == 2) &&
3079              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3080                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3081                         goto clear_extattr;
3082         }
3083
3084         if (header->h_blocks != 1) {
3085                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3086                         goto clear_extattr;
3087         }
3088
3089         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3090                 goto clear_extattr;
3091
3092         region = region_create(0, fs->blocksize);
3093         if (!region) {
3094                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3095                 ctx->flags |= E2F_FLAG_ABORT;
3096                 return 0;
3097         }
3098         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3099                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3100                         goto clear_extattr;
3101         }
3102
3103         first = (struct ext2_ext_attr_entry *)(header+1);
3104         end = block_buf + fs->blocksize;
3105         entry = first;
3106         while ((char *)entry < end && *(__u32 *)entry) {
3107                 __u32 hash;
3108
3109                 if (region_allocate(region, (char *)entry - (char *)header,
3110                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3111                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3112                                 goto clear_extattr;
3113                         break;
3114                 }
3115                 if ((ctx->ext_attr_ver == 1 &&
3116                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3117                     (ctx->ext_attr_ver == 2 &&
3118                      entry->e_name_index == 0)) {
3119                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3120                                 goto clear_extattr;
3121                         break;
3122                 }
3123                 if (entry->e_value_inum == 0) {
3124                         if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3125                             (entry->e_value_offs + entry->e_value_size >
3126                              fs->blocksize)) {
3127                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3128                                         goto clear_extattr;
3129                                 break;
3130                         }
3131                         if (entry->e_value_size &&
3132                             region_allocate(region, entry->e_value_offs,
3133                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3134                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3135                                                 pctx))
3136                                         goto clear_extattr;
3137                         }
3138
3139                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3140                                                           entry->e_value_offs);
3141                         if (entry->e_hash != hash)
3142                                 hash = ext2fs_ext_attr_hash_entry_signed(entry,
3143                                         block_buf + entry->e_value_offs);
3144
3145                         if (entry->e_hash != hash) {
3146                                 pctx->num = entry->e_hash;
3147                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3148                                         goto clear_extattr;
3149                                 entry->e_hash = hash;
3150                         }
3151                 } else {
3152                         problem_t problem;
3153                         blk64_t entry_quota_blocks;
3154
3155                         problem = check_large_ea_inode(ctx, entry, pctx,
3156                                                        &entry_quota_blocks);
3157                         if (problem && fix_problem(ctx, problem, pctx))
3158                                 goto clear_extattr;
3159
3160                         quota_blocks += entry_quota_blocks;
3161                         quota_inodes++;
3162                 }
3163
3164                 entry = EXT2_EXT_ATTR_NEXT(entry);
3165         }
3166         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3167                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3168                         goto clear_extattr;
3169         }
3170         region_free(region);
3171
3172         /*
3173          * We only get here if there was no other errors that were fixed.
3174          * If there was a checksum fail, ask to correct it.
3175          */
3176         if (failed_csum &&
3177             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3178                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3179                                                        pctx->ino);
3180                 if (pctx->errcode)
3181                         return 0;
3182         }
3183
3184         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3185                 if (!ctx->ea_block_quota_blocks) {
3186                         pctx->errcode = ea_refcount_create(0,
3187                                                 &ctx->ea_block_quota_blocks);
3188                         if (pctx->errcode) {
3189                                 pctx->num = 3;
3190                                 goto refcount_fail;
3191                         }
3192                 }
3193                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3194                                   quota_blocks);
3195         }
3196
3197         if (quota_inodes) {
3198                 if (!ctx->ea_block_quota_inodes) {
3199                         pctx->errcode = ea_refcount_create(0,
3200                                                 &ctx->ea_block_quota_inodes);
3201                         if (pctx->errcode) {
3202                                 pctx->num = 4;
3203 refcount_fail:
3204                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3205                                 ctx->flags |= E2F_FLAG_ABORT;
3206                                 return 0;
3207                         }
3208                 }
3209
3210                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3211                                   quota_inodes);
3212         }
3213         ea_block_quota->blocks = quota_blocks;
3214         ea_block_quota->inodes = quota_inodes;
3215
3216         inc_ea_inode_refs(ctx, pctx, first, end);
3217         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3218         mark_block_used(ctx, blk);
3219         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3220         return 1;
3221
3222 clear_extattr:
3223         if (region)
3224                 region_free(region);
3225         ext2fs_file_acl_block_set(fs, inode, 0);
3226         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3227         return 0;
3228 }
3229
3230 /* Returns 1 if bad htree, 0 if OK */
3231 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3232                         ext2_ino_t ino, struct ext2_inode *inode,
3233                         char *block_buf)
3234 {
3235         struct ext2_dx_root_info        *root;
3236         ext2_filsys                     fs = ctx->fs;
3237         errcode_t                       retval;
3238         blk64_t                         blk;
3239
3240         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3241              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3242             (!ext2fs_has_feature_dir_index(fs->super) &&
3243              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3244                 return 1;
3245
3246         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3247
3248         if ((pctx->errcode) ||
3249             (blk == 0) ||
3250             (blk < fs->super->s_first_data_block) ||
3251             (blk >= ext2fs_blocks_count(fs->super))) {
3252                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3253                         return 1;
3254                 else
3255                         return 0;
3256         }
3257
3258         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3259         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3260                 return 1;
3261
3262         /* XXX should check that beginning matches a directory */
3263         root = (struct ext2_dx_root_info *) (block_buf + 24);
3264
3265         if ((root->reserved_zero || root->info_length < 8) &&
3266             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3267                 return 1;
3268
3269         pctx->num = root->hash_version;
3270         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3271             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3272             (root->hash_version != EXT2_HASH_TEA) &&
3273             (root->hash_version != EXT2_HASH_SIPHASH) &&
3274             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3275                 return 1;
3276
3277         if (ext4_hash_in_dirent(inode)) {
3278                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3279                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3280                         return 1;
3281         } else {
3282                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3283                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3284                         return 1;
3285         }
3286
3287         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3288             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3289                 return 1;
3290
3291         pctx->num = root->indirect_levels;
3292         /* if htree level is clearly too high, consider it to be broken */
3293         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3294             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3295                 return 1;
3296
3297         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3298         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3299             !ext2fs_has_feature_largedir(fs->super)) {
3300                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3301                 unsigned idx_pb = 1 << (blockbits - 3);
3302
3303                 /* compare inode size/blocks vs. max-sized 2-level htree */
3304                 if (EXT2_I_SIZE(pctx->inode) <
3305                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3306                     pctx->inode->i_blocks <
3307                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3308                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3309                         return 1;
3310         }
3311
3312         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3313             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3314                 ctx->large_dirs++;
3315
3316         return 0;
3317 }
3318
3319 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3320                         struct ext2_inode *inode, int restart_flag,
3321                         const char *source)
3322 {
3323         inode->i_flags = 0;
3324         inode->i_links_count = 0;
3325         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3326         inode->i_dtime = ctx->now;
3327
3328         /*
3329          * If a special inode has such rotten block mappings that we
3330          * want to clear the whole inode, be sure to actually zap
3331          * the block maps because i_links_count isn't checked for
3332          * special inodes, and we'll end up right back here the next
3333          * time we run fsck.
3334          */
3335         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3336                 memset(inode->i_block, 0, sizeof(inode->i_block));
3337
3338         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3339         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3340         if (ctx->inode_reg_map)
3341                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3342         if (ctx->inode_bad_map)
3343                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3344
3345         /*
3346          * If the inode was partially accounted for before processing
3347          * was aborted, we need to restart the pass 1 scan.
3348          */
3349         ctx->flags |= restart_flag;
3350
3351         if (ino == EXT2_BAD_INO)
3352                 memset(inode, 0, sizeof(struct ext2_inode));
3353
3354         e2fsck_write_inode(ctx, ino, inode, source);
3355 }
3356
3357 /*
3358  * Use the multiple-blocks reclamation code to fix alignment problems in
3359  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3360  * physical cluster, and we want the block offsets within that cluster to
3361  * line up.
3362  */
3363 static int has_unaligned_cluster_map(e2fsck_t ctx,
3364                                      blk64_t last_pblk, blk64_t last_lblk,
3365                                      blk64_t pblk, blk64_t lblk)
3366 {
3367         blk64_t cluster_mask;
3368
3369         if (!ctx->fs->cluster_ratio_bits)
3370                 return 0;
3371         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3372
3373         /*
3374          * If the block in the logical cluster doesn't align with the block in
3375          * the physical cluster...
3376          */
3377         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3378                 return 1;
3379
3380         /*
3381          * If we cross a physical cluster boundary within a logical cluster...
3382          */
3383         if (last_pblk && (lblk & cluster_mask) != 0 &&
3384             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3385             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3386                 return 1;
3387
3388         return 0;
3389 }
3390
3391 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3392                              struct process_block_struct *pb,
3393                              blk64_t start_block, blk64_t end_block,
3394                              blk64_t eof_block,
3395                              ext2_extent_handle_t ehandle,
3396                              int try_repairs)
3397 {
3398         struct ext2fs_extent    extent;
3399         blk64_t                 blk, last_lblk;
3400         unsigned int            i, n;
3401         int                     is_dir, is_leaf;
3402         problem_t               problem;
3403         struct ext2_extent_info info;
3404         int                     failed_csum = 0;
3405
3406         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3407                 failed_csum = 1;
3408
3409         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3410         if (pctx->errcode)
3411                 return;
3412         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3413             !pb->eti.force_rebuild &&
3414             info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
3415                 struct extent_tree_level *etl;
3416
3417                 etl = pb->eti.ext_info + info.curr_level;
3418                 etl->num_extents += info.num_entries;
3419                 etl->max_extents += info.max_entries;
3420                 /*
3421                  * Implementation wart: Splitting extent blocks when appending
3422                  * will leave the old block with one free entry.  Therefore
3423                  * unless the node is totally full, pretend that a non-root
3424                  * extent block can hold one fewer entry than it actually does,
3425                  * so that we don't repeatedly rebuild the extent tree.
3426                  */
3427                 if (info.curr_level && info.num_entries < info.max_entries)
3428                         etl->max_extents--;
3429         }
3430
3431         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3432                                           &extent);
3433         while ((pctx->errcode == 0 ||
3434                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3435                info.num_entries-- > 0) {
3436                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3437                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3438                 last_lblk = extent.e_lblk + extent.e_len - 1;
3439
3440                 problem = 0;
3441                 pctx->blk = extent.e_pblk;
3442                 pctx->blk2 = extent.e_lblk;
3443                 pctx->num = extent.e_len;
3444                 pctx->blkcount = extent.e_lblk + extent.e_len;
3445
3446                 if (extent.e_pblk == 0 ||
3447                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3448                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3449                         problem = PR_1_EXTENT_BAD_START_BLK;
3450                 else if (extent.e_lblk < start_block)
3451                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3452                 else if ((end_block && last_lblk > end_block) &&
3453                          !(last_lblk > eof_block &&
3454                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3455                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3456                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3457                 else if (is_leaf && extent.e_len == 0)
3458                         problem = PR_1_EXTENT_LENGTH_ZERO;
3459                 else if (is_leaf &&
3460                          (extent.e_pblk + extent.e_len) >
3461                          ext2fs_blocks_count(ctx->fs->super))
3462                         problem = PR_1_EXTENT_ENDS_BEYOND;
3463                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3464                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3465                          ((extent.e_lblk + extent.e_len) >
3466                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3467                         problem = PR_1_TOOBIG_DIR;
3468
3469                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3470 #if 0
3471                         printf("extent_region(ino=%u, expect=%llu, "
3472                                "lblk=%llu, len=%u)\n", pb->ino,
3473                                (unsigned long long) pb->next_lblock,
3474                                (unsigned long long) extent.e_lblk,
3475                                extent.e_len);
3476 #endif
3477                         if (extent.e_lblk < pb->next_lblock)
3478                                 problem = PR_1_EXTENT_COLLISION;
3479                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3480                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3481                 }
3482
3483                 /*
3484                  * Uninitialized blocks in a directory?  Clear the flag and
3485                  * we'll interpret the blocks later.
3486                  */
3487                 if (try_repairs && is_dir && problem == 0 &&
3488                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3489                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3490                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3491                         pb->inode_modified = 1;
3492                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3493                                                               &extent);
3494                         if (pctx->errcode)
3495                                 return;
3496                         failed_csum = 0;
3497                 }
3498 #ifdef CONFIG_DEVELOPER_FEATURES
3499                 if (try_repairs && !is_dir && problem == 0 &&
3500                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3501                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3502                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3503                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3504                         pb->inode_modified = 1;
3505                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3506                                                               &extent);
3507                         if (pctx->errcode)
3508                                 return;
3509                         failed_csum = 0;
3510                 }
3511 #endif
3512                 if (try_repairs && problem) {
3513 report_problem:
3514                         if (fix_problem(ctx, problem, pctx)) {
3515                                 if (ctx->invalid_bitmaps) {
3516                                         /*
3517                                          * If fsck knows the bitmaps are bad,
3518                                          * skip to the next extent and
3519                                          * try to clear this extent again
3520                                          * after fixing the bitmaps, by
3521                                          * restarting fsck.
3522                                          */
3523                                         pctx->errcode = ext2fs_extent_get(
3524                                                           ehandle,
3525                                                           EXT2_EXTENT_NEXT_SIB,
3526                                                           &extent);
3527                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3528                                         if (pctx->errcode ==
3529                                                     EXT2_ET_NO_CURRENT_NODE) {
3530                                                 pctx->errcode = 0;
3531                                                 break;
3532                                         }
3533                                         continue;
3534                                 }
3535                                 e2fsck_read_bitmaps(ctx);
3536                                 pb->inode_modified = 1;
3537                                 pctx->errcode =
3538                                         ext2fs_extent_delete(ehandle, 0);
3539                                 if (pctx->errcode) {
3540                                         pctx->str = "ext2fs_extent_delete";
3541                                         return;
3542                                 }
3543                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3544                                 if (pctx->errcode &&
3545                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3546                                         pctx->str = "ext2fs_extent_fix_parents";
3547                                         return;
3548                                 }
3549                                 pctx->errcode = ext2fs_extent_get(ehandle,
3550                                                                   EXT2_EXTENT_CURRENT,
3551                                                                   &extent);
3552                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3553                                         pctx->errcode = 0;
3554                                         break;
3555                                 }
3556                                 failed_csum = 0;
3557                                 continue;
3558                         }
3559                         goto next;
3560                 }
3561
3562                 if (!is_leaf) {
3563                         blk64_t lblk = extent.e_lblk;
3564                         int next_try_repairs = 1;
3565
3566                         blk = extent.e_pblk;
3567
3568                         /*
3569                          * If this lower extent block collides with critical
3570                          * metadata, don't try to repair the damage.  Pass 1b
3571                          * will reallocate the block; then we can try again.
3572                          */
3573                         if (pb->ino != EXT2_RESIZE_INO &&
3574                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3575                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3576                                                       extent.e_pblk)) {
3577                                 next_try_repairs = 0;
3578                                 pctx->blk = blk;
3579                                 fix_problem(ctx,
3580                                             PR_1_CRITICAL_METADATA_COLLISION,
3581                                             pctx);
3582                                 if ((ctx->options & E2F_OPT_NO) == 0)
3583                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3584                         }
3585                         pctx->errcode = ext2fs_extent_get(ehandle,
3586                                                   EXT2_EXTENT_DOWN, &extent);
3587                         if (pctx->errcode &&
3588                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3589                                 pctx->str = "EXT2_EXTENT_DOWN";
3590                                 problem = PR_1_EXTENT_HEADER_INVALID;
3591                                 if (!next_try_repairs)
3592                                         return;
3593                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3594                                         goto report_problem;
3595                                 return;
3596                         }
3597                         /* The next extent should match this index's logical start */
3598                         if (extent.e_lblk != lblk) {
3599                                 struct ext2_extent_info e_info;
3600
3601                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3602                                                                        &e_info);
3603                                 if (pctx->errcode) {
3604                                         pctx->str = "ext2fs_extent_get_info";
3605                                         return;
3606                                 }
3607                                 pctx->blk = lblk;
3608                                 pctx->blk2 = extent.e_lblk;
3609                                 pctx->num = e_info.curr_level - 1;
3610                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3611                                 if (fix_problem(ctx, problem, pctx)) {
3612                                         pb->inode_modified = 1;
3613                                         pctx->errcode =
3614                                                 ext2fs_extent_fix_parents(ehandle);
3615                                         if (pctx->errcode) {
3616                                                 pctx->str = "ext2fs_extent_fix_parents";
3617                                                 return;
3618                                         }
3619                                 }
3620                         }
3621                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3622                                          last_lblk, eof_block, ehandle,
3623                                          next_try_repairs);
3624                         if (pctx->errcode)
3625                                 return;
3626                         pctx->errcode = ext2fs_extent_get(ehandle,
3627                                                   EXT2_EXTENT_UP, &extent);
3628                         if (pctx->errcode) {
3629                                 pctx->str = "EXT2_EXTENT_UP";
3630                                 return;
3631                         }
3632                         mark_block_used(ctx, blk);
3633                         pb->num_blocks++;
3634                         goto next;
3635                 }
3636
3637                 if ((pb->previous_block != 0) &&
3638                     (pb->previous_block+1 != extent.e_pblk)) {
3639                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3640                                 char type = '?';
3641
3642                                 if (pb->is_dir)
3643                                         type = 'd';
3644                                 else if (pb->is_reg)
3645                                         type = 'f';
3646
3647                                 printf(("%6lu(%c): expecting %6lu "
3648                                         "actual extent "
3649                                         "phys %6lu log %lu len %lu\n"),
3650                                        (unsigned long) pctx->ino, type,
3651                                        (unsigned long) pb->previous_block+1,
3652                                        (unsigned long) extent.e_pblk,
3653                                        (unsigned long) extent.e_lblk,
3654                                        (unsigned long) extent.e_len);
3655                         }
3656                         pb->fragmented = 1;
3657                 }
3658                 /*
3659                  * If we notice a gap in the logical block mappings of an
3660                  * extent-mapped directory, offer to close the hole by
3661                  * moving the logical block down, otherwise we'll go mad in
3662                  * pass 3 allocating empty directory blocks to fill the hole.
3663                  */
3664                 if (try_repairs && is_dir &&
3665                     pb->last_block + 1 < extent.e_lblk) {
3666                         blk64_t new_lblk;
3667
3668                         new_lblk = pb->last_block + 1;
3669                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3670                                 new_lblk = ((new_lblk +
3671                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3672                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3673                                            (extent.e_pblk &
3674                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3675                         pctx->blk = extent.e_lblk;
3676                         pctx->blk2 = new_lblk;
3677                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3678                                 extent.e_lblk = new_lblk;
3679                                 pb->inode_modified = 1;
3680                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3681                                                                 0, &extent);
3682                                 if (pctx->errcode) {
3683                                         pctx->errcode = 0;
3684                                         goto alloc_later;
3685                                 }
3686                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3687                                 if (pctx->errcode)
3688                                         goto failed_add_dir_block;
3689                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3690                                                                 extent.e_lblk);
3691                                 if (pctx->errcode)
3692                                         goto failed_add_dir_block;
3693                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3694                                 failed_csum = 0;
3695                         }
3696                 }
3697 alloc_later:
3698                 if (is_dir) {
3699                         while (++pb->last_db_block <
3700                                (e2_blkcnt_t) extent.e_lblk) {
3701                                 pctx->errcode = ext2fs_add_dir_block2(
3702                                                         ctx->fs->dblist,
3703                                                         pb->ino, 0,
3704                                                         pb->last_db_block);
3705                                 if (pctx->errcode) {
3706                                         pctx->blk = 0;
3707                                         pctx->num = pb->last_db_block;
3708                                         goto failed_add_dir_block;
3709                                 }
3710                         }
3711
3712                         for (i = 0; i < extent.e_len; i++) {
3713                                 pctx->errcode = ext2fs_add_dir_block2(
3714                                                         ctx->fs->dblist,
3715                                                         pctx->ino,
3716                                                         extent.e_pblk + i,
3717                                                         extent.e_lblk + i);
3718                                 if (pctx->errcode) {
3719                                         pctx->blk = extent.e_pblk + i;
3720                                         pctx->num = extent.e_lblk + i;
3721                                 failed_add_dir_block:
3722                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3723                                         /* Should never get here */
3724                                         ctx->flags |= E2F_FLAG_ABORT;
3725                                         return;
3726                                 }
3727                         }
3728                         if (extent.e_len > 0)
3729                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3730                 }
3731                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3732                                               pb->last_block,
3733                                               extent.e_pblk,
3734                                               extent.e_lblk)) {
3735                         for (i = 0; i < extent.e_len; i++) {
3736                                 pctx->blk = extent.e_lblk + i;
3737                                 pctx->blk2 = extent.e_pblk + i;
3738                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3739                                 mark_block_used(ctx, extent.e_pblk + i);
3740                                 mark_block_used(ctx, extent.e_pblk + i);
3741                         }
3742                 }
3743
3744                 /*
3745                  * Check whether first cluster got marked in previous iteration.
3746                  */
3747                 if (ctx->fs->cluster_ratio_bits &&
3748                     pb->previous_block &&
3749                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3750                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3751                         /* Set blk to the beginning of next cluster. */
3752                         blk = EXT2FS_C2B(
3753                                 ctx->fs,
3754                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3755                 else
3756                         /* Set blk to the beginning of current cluster. */
3757                         blk = EXT2FS_C2B(ctx->fs,
3758                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3759
3760                 if (blk < extent.e_pblk + extent.e_len) {
3761                         mark_blocks_used(ctx, blk,
3762                                          extent.e_pblk + extent.e_len - blk);
3763                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3764                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3765                         pb->num_blocks += n;
3766                 }
3767                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3768                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3769                 start_block = pb->last_block = last_lblk;
3770                 if (is_leaf && !is_dir &&
3771                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3772                         pb->last_init_lblock = last_lblk;
3773         next:
3774                 pctx->errcode = ext2fs_extent_get(ehandle,
3775                                                   EXT2_EXTENT_NEXT_SIB,
3776                                                   &extent);
3777         }
3778
3779         /* Failed csum but passes checks?  Ask to fix checksum. */
3780         if (failed_csum &&
3781             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3782                 pb->inode_modified = 1;
3783                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3784                 if (pctx->errcode)
3785                         return;
3786         }
3787
3788         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3789                 pctx->errcode = 0;
3790 }
3791
3792 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3793                                  struct process_block_struct *pb)
3794 {
3795         struct ext2_extent_info info;
3796         struct ext2_inode       *inode = pctx->inode;
3797         ext2_extent_handle_t    ehandle;
3798         ext2_filsys             fs = ctx->fs;
3799         ext2_ino_t              ino = pctx->ino;
3800         errcode_t               retval;
3801         blk64_t                 eof_lblk;
3802         struct ext3_extent_header       *eh;
3803
3804         /* Check for a proper extent header... */
3805         eh = (struct ext3_extent_header *) &inode->i_block[0];
3806         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3807         if (retval) {
3808                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3809                         e2fsck_clear_inode(ctx, ino, inode, 0,
3810                                            "check_blocks_extents");
3811                 pctx->errcode = 0;
3812                 return;
3813         }
3814
3815         /* ...since this function doesn't fail if i_block is zeroed. */
3816         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3817         if (pctx->errcode) {
3818                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3819                         e2fsck_clear_inode(ctx, ino, inode, 0,
3820                                            "check_blocks_extents");
3821                 pctx->errcode = 0;
3822                 return;
3823         }
3824
3825         retval = ext2fs_extent_get_info(ehandle, &info);
3826         if (retval == 0) {
3827                 int max_depth = info.max_depth;
3828
3829                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3830                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3831                 ctx->extent_depth_count[max_depth]++;
3832         }
3833
3834         /* Check maximum extent depth */
3835         pctx->blk = info.max_depth;
3836         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3837         if (pctx->blk2 < pctx->blk &&
3838             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3839                 pb->eti.force_rebuild = 1;
3840
3841         /* Can we collect extent tree level stats? */
3842         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3843         if (pctx->blk2 > pctx->blk)
3844                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3845         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3846         pb->eti.ino = pb->ino;
3847
3848         pb->next_lblock = 0;
3849
3850         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3851                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3852         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3853         if (pctx->errcode &&
3854             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3855                 pb->num_blocks = 0;
3856                 inode->i_blocks = 0;
3857                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3858                                    "check_blocks_extents");
3859                 pctx->errcode = 0;
3860         }
3861         ext2fs_extent_free(ehandle);
3862
3863         /* Rebuild unless it's a dir and we're rehashing it */
3864         if (LINUX_S_ISDIR(inode->i_mode) &&
3865             e2fsck_dir_will_be_rehashed(ctx, ino))
3866                 return;
3867
3868         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3869                 e2fsck_rebuild_extents_later(ctx, ino);
3870         else
3871                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3872 }
3873
3874 /*
3875  * In fact we don't need to check blocks for an inode with inline data
3876  * because this inode doesn't have any blocks.  In this function all
3877  * we need to do is add this inode into dblist when it is a directory.
3878  */
3879 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3880                                      struct process_block_struct *pb)
3881 {
3882         int     flags;
3883         size_t  inline_data_size = 0;
3884
3885         if (!pb->is_dir) {
3886                 pctx->errcode = 0;
3887                 return;
3888         }
3889
3890         /* Process the dirents in i_block[] as the "first" block. */
3891         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3892         if (pctx->errcode)
3893                 goto err;
3894
3895         /* Process the dirents in the EA as a "second" block. */
3896         flags = ctx->fs->flags;
3897         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3898         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3899                                                 &inline_data_size);
3900         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3901                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3902         if (pctx->errcode) {
3903                 pctx->errcode = 0;
3904                 return;
3905         }
3906
3907         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3908                 return;
3909
3910         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3911         if (pctx->errcode)
3912                 goto err;
3913
3914         return;
3915 err:
3916         pctx->blk = 0;
3917         pctx->num = 0;
3918         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3919         ctx->flags |= E2F_FLAG_ABORT;
3920 }
3921
3922 /*
3923  * This subroutine is called on each inode to account for all of the
3924  * blocks used by that inode.
3925  */
3926 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3927                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3928 {
3929         ext2_filsys fs = ctx->fs;
3930         struct process_block_struct pb;
3931         ext2_ino_t      ino = pctx->ino;
3932         struct ext2_inode *inode = pctx->inode;
3933         unsigned        bad_size = 0;
3934         int             dirty_inode = 0;
3935         int             extent_fs;
3936         int             inlinedata_fs;
3937         __u64           size;
3938         struct ea_quota ea_block_quota;
3939
3940         pb.ino = ino;
3941         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3942                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3943         pb.last_block = ~0;
3944         pb.last_init_lblock = -1;
3945         pb.last_db_block = -1;
3946         pb.num_illegal_blocks = 0;
3947         pb.suppress = 0; pb.clear = 0;
3948         pb.fragmented = 0;
3949         pb.compressed = 0;
3950         pb.previous_block = 0;
3951         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3952         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3953         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3954         pb.inode = inode;
3955         pb.pctx = pctx;
3956         pb.ctx = ctx;
3957         pb.inode_modified = 0;
3958         pb.eti.force_rebuild = 0;
3959         pctx->ino = ino;
3960         pctx->errcode = 0;
3961
3962         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3963         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3964
3965         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3966                 if (e2fsck_should_abort(ctx))
3967                         goto out;
3968                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3969         }
3970
3971         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3972                 check_blocks_inline_data(ctx, pctx, &pb);
3973         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3974                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3975                         check_blocks_extents(ctx, pctx, &pb);
3976                 else {
3977                         int flags;
3978                         /*
3979                          * If we've modified the inode, write it out before
3980                          * iterate() tries to use it.
3981                          */
3982                         if (dirty_inode) {
3983                                 e2fsck_write_inode(ctx, ino, inode,
3984                                                    "check_blocks");
3985                                 dirty_inode = 0;
3986                         }
3987                         flags = fs->flags;
3988                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3989                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3990                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3991                                                 block_buf, process_block, &pb);
3992                         /*
3993                          * We do not have uninitialized extents in non extent
3994                          * files.
3995                          */
3996                         pb.last_init_lblock = pb.last_block;
3997                         /*
3998                          * If iterate() changed a block mapping, we have to
3999                          * re-read the inode.  If we decide to clear the
4000                          * inode after clearing some stuff, we'll re-write the
4001                          * bad mappings into the inode!
4002                          */
4003                         if (pb.inode_modified)
4004                                 e2fsck_read_inode(ctx, ino, inode,
4005                                                   "check_blocks");
4006                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4007                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4008
4009                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
4010 #ifdef DEBUG
4011                                 printf("bmap rebuild ino=%d\n", ino);
4012 #endif
4013                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
4014                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
4015                                         e2fsck_rebuild_extents_later(ctx, ino);
4016                         }
4017                 }
4018         }
4019         end_problem_latch(ctx, PR_LATCH_BLOCK);
4020         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4021         if (e2fsck_should_abort(ctx))
4022                 goto out;
4023         if (pctx->errcode)
4024                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4025
4026         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4027                 if (LINUX_S_ISDIR(inode->i_mode))
4028                         ctx->fs_fragmented_dir++;
4029                 else
4030                         ctx->fs_fragmented++;
4031         }
4032
4033         if (pb.clear) {
4034                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4035                                    "check_blocks");
4036                 return;
4037         }
4038
4039         if (inode->i_flags & EXT2_INDEX_FL) {
4040                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4041                         inode->i_flags &= ~EXT2_INDEX_FL;
4042                         dirty_inode++;
4043                 } else {
4044                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4045                 }
4046         }
4047
4048         if (!pb.num_blocks && pb.is_dir &&
4049             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4050                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4051                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4052                         ctx->fs_directory_count--;
4053                         return;
4054                 }
4055         }
4056
4057         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4058             ino != fs->super->s_orphan_file_inum &&
4059             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4060             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4061                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4062                                ino,
4063                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4064                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4065                                   ino, (ea_ibody_quota ?
4066                                         ea_ibody_quota->inodes : 0) +
4067                                                 ea_block_quota.inodes + 1);
4068         }
4069
4070         if (!ext2fs_has_feature_huge_file(fs->super) ||
4071             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4072                 pb.num_blocks *= (fs->blocksize / 512);
4073         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4074 #if 0
4075         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4076                ino, inode->i_size, (unsigned long long) pb.last_block,
4077                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4078                (unsigned long long) pb.num_blocks);
4079 #endif
4080         size = EXT2_I_SIZE(inode);
4081         if (pb.is_dir) {
4082                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4083                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4084                         int flags;
4085                         size_t sz = 0;
4086                         errcode_t err;
4087
4088                         flags = ctx->fs->flags;
4089                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4090                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4091                                                       &sz);
4092                         ctx->fs->flags = (flags &
4093                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4094                                          (ctx->fs->flags &
4095                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4096                         if (err || sz != size) {
4097                                 bad_size = 7;
4098                                 pctx->num = sz;
4099                         }
4100                 } else if (size & (fs->blocksize - 1))
4101                         bad_size = 5;
4102                 else if (nblock > (pb.last_block + 1))
4103                         bad_size = 1;
4104                 else if (nblock < (pb.last_block + 1)) {
4105                         if (((pb.last_block + 1) - nblock) >
4106                             fs->super->s_prealloc_dir_blocks)
4107                                 bad_size = 2;
4108                 }
4109         } else {
4110                 if ((pb.last_init_lblock >= 0) &&
4111                     /* Do not allow initialized allocated blocks past i_size*/
4112                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4113                     !(inode->i_flags & EXT4_VERITY_FL))
4114                         bad_size = 3;
4115                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4116                          size > ext2_max_sizes[fs->super->s_log_block_size])
4117                         /* too big for a direct/indirect-mapped file */
4118                         bad_size = 4;
4119                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4120                          size >
4121                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4122                         /* too big for an extent-based file - 32bit ee_block */
4123                         bad_size = 6;
4124         }
4125         /* i_size for symlinks is checked elsewhere */
4126         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4127                 /* Did inline_data set pctx->num earlier? */
4128                 if (bad_size != 7)
4129                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4130                 pctx->group = bad_size;
4131                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4132                         ext2fs_inode_size_set(fs, inode, pctx->num);
4133                         if (EXT2_I_SIZE(inode) == 0 &&
4134                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4135                                 memset(inode->i_block, 0,
4136                                        sizeof(inode->i_block));
4137                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4138                         }
4139                         dirty_inode++;
4140                 }
4141                 pctx->num = 0;
4142         }
4143         if (LINUX_S_ISREG(inode->i_mode) &&
4144             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4145                 ctx->large_files++;
4146         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4147             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4148              (ext2fs_has_feature_huge_file(fs->super) &&
4149               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4150               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4151                 pctx->num = pb.num_blocks;
4152                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4153                         inode->i_blocks = pb.num_blocks;
4154                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4155                         dirty_inode++;
4156                 }
4157                 pctx->num = 0;
4158         }
4159
4160         /*
4161          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4162          * a block mapped file, so rebuild it as an extent file.  We can skip
4163          * symlinks because they're never rewritten.
4164          */
4165         if (ext2fs_has_feature_bigalloc(fs->super) &&
4166             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4167             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4168             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4169             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4170             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4171                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4172                 if (pctx->errcode)
4173                         goto out;
4174         }
4175
4176         if (ctx->dirs_to_hash && pb.is_dir &&
4177             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4178             !(inode->i_flags & EXT2_INDEX_FL) &&
4179             ((inode->i_size / fs->blocksize) >= 3))
4180                 e2fsck_rehash_dir_later(ctx, ino);
4181
4182 out:
4183         if (dirty_inode)
4184                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4185 }
4186
4187 #if 0
4188 /*
4189  * Helper function called by process block when an illegal block is
4190  * found.  It returns a description about why the block is illegal
4191  */
4192 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4193 {
4194         blk64_t super;
4195         int     i;
4196         static char     problem[80];
4197
4198         super = fs->super->s_first_data_block;
4199         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4200         if (block < super) {
4201                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4202                 return(problem);
4203         } else if (block >= ext2fs_blocks_count(fs->super)) {
4204                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4205                 return(problem);
4206         }
4207         for (i = 0; i < fs->group_desc_count; i++) {
4208                 if (block == super) {
4209                         sprintf(problem, "is the superblock in group %d", i);
4210                         break;
4211                 }
4212                 if (block > super &&
4213                     block <= (super + fs->desc_blocks)) {
4214                         sprintf(problem, "is in the group descriptors "
4215                                 "of group %d", i);
4216                         break;
4217                 }
4218                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4219                         sprintf(problem, "is the block bitmap of group %d", i);
4220                         break;
4221                 }
4222                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4223                         sprintf(problem, "is the inode bitmap of group %d", i);
4224                         break;
4225                 }
4226                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4227                     (block < ext2fs_inode_table_loc(fs, i)
4228                      + fs->inode_blocks_per_group)) {
4229                         sprintf(problem, "is in the inode table of group %d",
4230                                 i);
4231                         break;
4232                 }
4233                 super += fs->super->s_blocks_per_group;
4234         }
4235         return(problem);
4236 }
4237 #endif
4238
4239 /*
4240  * This is a helper function for check_blocks().
4241  */
4242 static int process_block(ext2_filsys fs,
4243                   blk64_t       *block_nr,
4244                   e2_blkcnt_t blockcnt,
4245                   blk64_t ref_block EXT2FS_ATTR((unused)),
4246                   int ref_offset EXT2FS_ATTR((unused)),
4247                   void *priv_data)
4248 {
4249         struct process_block_struct *p;
4250         struct problem_context *pctx;
4251         blk64_t blk = *block_nr;
4252         int     ret_code = 0;
4253         problem_t       problem = 0;
4254         e2fsck_t        ctx;
4255
4256         p = (struct process_block_struct *) priv_data;
4257         pctx = p->pctx;
4258         ctx = p->ctx;
4259
4260         /*
4261          * For a directory, add logical block zero for processing even if it's
4262          * not mapped or we'll be perennially stuck with broken "." and ".."
4263          * entries.
4264          */
4265         if (p->is_dir && blockcnt == 0 && blk == 0) {
4266                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4267                 if (pctx->errcode) {
4268                         pctx->blk = blk;
4269                         pctx->num = blockcnt;
4270                         goto failed_add_dir_block;
4271                 }
4272                 p->last_db_block++;
4273         }
4274
4275         if (blk == 0)
4276                 return 0;
4277
4278 #if 0
4279         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4280                blockcnt);
4281 #endif
4282
4283         /*
4284          * Simplistic fragmentation check.  We merely require that the
4285          * file be contiguous.  (Which can never be true for really
4286          * big files that are greater than a block group.)
4287          */
4288         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4289                 if (p->previous_block+1 != blk) {
4290                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4291                                 char type = '?';
4292
4293                                 if (p->is_dir)
4294                                         type = 'd';
4295                                 else if (p->is_reg)
4296                                         type = 'f';
4297
4298                                 printf(_("%6lu(%c): expecting %6lu "
4299                                          "got phys %6lu (blkcnt %lld)\n"),
4300                                        (unsigned long) pctx->ino, type,
4301                                        (unsigned long) p->previous_block+1,
4302                                        (unsigned long) blk,
4303                                        (long long) blockcnt);
4304                         }
4305                         p->fragmented = 1;
4306                 }
4307         }
4308
4309         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4310             !pctx->inode->i_size_high &&
4311             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4312                 problem = PR_1_TOOBIG_DIR;
4313         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4314                 problem = PR_1_TOOBIG_DIR;
4315         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4316                 problem = PR_1_TOOBIG_REG;
4317         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4318                 problem = PR_1_TOOBIG_SYMLINK;
4319
4320         if (blk < fs->super->s_first_data_block ||
4321             blk >= ext2fs_blocks_count(fs->super))
4322                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4323
4324         /*
4325          * If this IND/DIND/TIND block is squatting atop some critical metadata
4326          * (group descriptors, superblock, bitmap, inode table), any write to
4327          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4328          * fix that and restart fsck.
4329          */
4330         if (blockcnt < 0 &&
4331             p->ino != EXT2_RESIZE_INO &&
4332             blk < ctx->fs->super->s_blocks_count &&
4333             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4334                 pctx->blk = blk;
4335                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4336                 if ((ctx->options & E2F_OPT_NO) == 0)
4337                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4338         }
4339
4340         if (problem) {
4341                 p->num_illegal_blocks++;
4342                 /*
4343                  * A bit of subterfuge here -- we're trying to fix a block
4344                  * mapping, but the IND/DIND/TIND block could have collided
4345                  * with some critical metadata.  So, fix the in-core mapping so
4346                  * iterate won't go insane, but return 0 instead of
4347                  * BLOCK_CHANGED so that it won't write the remapping out to
4348                  * our multiply linked block.
4349                  *
4350                  * Even if we previously determined that an *IND block
4351                  * conflicts with critical metadata, we must still try to
4352                  * iterate the *IND block as if it is an *IND block to find and
4353                  * mark the blocks it points to.  Better to be overly cautious
4354                  * with the used_blocks map so that we don't move the *IND
4355                  * block to a block that's really in use!
4356                  */
4357                 if (p->ino != EXT2_RESIZE_INO &&
4358                     ref_block != 0 &&
4359                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4360                                               ref_block)) {
4361                         *block_nr = 0;
4362                         return 0;
4363                 }
4364                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4365                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4366                                 p->clear = 1;
4367                                 return BLOCK_ABORT;
4368                         }
4369                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4370                                 p->suppress = 1;
4371                                 set_latch_flags(PR_LATCH_BLOCK,
4372                                                 PRL_SUPPRESS, 0);
4373                         }
4374                 }
4375                 pctx->blk = blk;
4376                 pctx->blkcount = blockcnt;
4377                 if (fix_problem(ctx, problem, pctx)) {
4378                         blk = *block_nr = 0;
4379                         ret_code = BLOCK_CHANGED;
4380                         p->inode_modified = 1;
4381                         /*
4382                          * If the directory block is too big and is beyond the
4383                          * end of the FS, don't bother trying to add it for
4384                          * processing -- the kernel would never have created a
4385                          * directory this large, and we risk an ENOMEM abort.
4386                          * In any case, the toobig handler for extent-based
4387                          * directories also doesn't feed toobig blocks to
4388                          * pass 2.
4389                          */
4390                         if (problem == PR_1_TOOBIG_DIR)
4391                                 return ret_code;
4392                         goto mark_dir;
4393                 } else
4394                         return 0;
4395         }
4396
4397         if (p->ino == EXT2_RESIZE_INO) {
4398                 /*
4399                  * The resize inode has already be sanity checked
4400                  * during pass #0 (the superblock checks).  All we
4401                  * have to do is mark the double indirect block as
4402                  * being in use; all of the other blocks are handled
4403                  * by mark_table_blocks()).
4404                  */
4405                 if (blockcnt == BLOCK_COUNT_DIND)
4406                         mark_block_used(ctx, blk);
4407                 p->num_blocks++;
4408         } else if (!(ctx->fs->cluster_ratio_bits &&
4409                      p->previous_block &&
4410                      (EXT2FS_B2C(ctx->fs, blk) ==
4411                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4412                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4413                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4414                 mark_block_used(ctx, blk);
4415                 p->num_blocks++;
4416         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4417                                              p->last_block, blk, blockcnt)) {
4418                 pctx->blk = blockcnt;
4419                 pctx->blk2 = blk;
4420                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4421                 mark_block_used(ctx, blk);
4422                 mark_block_used(ctx, blk);
4423         }
4424         if (blockcnt >= 0)
4425                 p->last_block = blockcnt;
4426         p->previous_block = blk;
4427 mark_dir:
4428         if (p->is_dir && (blockcnt >= 0)) {
4429                 while (++p->last_db_block < blockcnt) {
4430                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4431                                                               p->ino, 0,
4432                                                               p->last_db_block);
4433                         if (pctx->errcode) {
4434                                 pctx->blk = 0;
4435                                 pctx->num = p->last_db_block;
4436                                 goto failed_add_dir_block;
4437                         }
4438                 }
4439                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4440                                                       blk, blockcnt);
4441                 if (pctx->errcode) {
4442                         pctx->blk = blk;
4443                         pctx->num = blockcnt;
4444                 failed_add_dir_block:
4445                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4446                         /* Should never get here */
4447                         ctx->flags |= E2F_FLAG_ABORT;
4448                         return BLOCK_ABORT;
4449                 }
4450         }
4451         return ret_code;
4452 }
4453
4454 static int process_bad_block(ext2_filsys fs,
4455                       blk64_t *block_nr,
4456                       e2_blkcnt_t blockcnt,
4457                       blk64_t ref_block EXT2FS_ATTR((unused)),
4458                       int ref_offset EXT2FS_ATTR((unused)),
4459                       void *priv_data)
4460 {
4461         struct process_block_struct *p;
4462         blk64_t         blk = *block_nr;
4463         blk64_t         first_block;
4464         dgrp_t          i;
4465         struct problem_context *pctx;
4466         e2fsck_t        ctx;
4467
4468         if (!blk)
4469                 return 0;
4470
4471         p = (struct process_block_struct *) priv_data;
4472         ctx = p->ctx;
4473         pctx = p->pctx;
4474
4475         pctx->ino = EXT2_BAD_INO;
4476         pctx->blk = blk;
4477         pctx->blkcount = blockcnt;
4478
4479         if ((blk < fs->super->s_first_data_block) ||
4480             (blk >= ext2fs_blocks_count(fs->super))) {
4481                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4482                         *block_nr = 0;
4483                         return BLOCK_CHANGED;
4484                 } else
4485                         return 0;
4486         }
4487
4488         if (blockcnt < 0) {
4489                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4490                         p->bbcheck = 1;
4491                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4492                                 *block_nr = 0;
4493                                 return BLOCK_CHANGED;
4494                         }
4495                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4496                                                     blk)) {
4497                         p->bbcheck = 1;
4498                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4499                                         pctx)) {
4500                                 *block_nr = 0;
4501                                 return BLOCK_CHANGED;
4502                         }
4503                         if (e2fsck_should_abort(ctx))
4504                                 return BLOCK_ABORT;
4505                 } else
4506                         mark_block_used(ctx, blk);
4507                 return 0;
4508         }
4509 #if 0
4510         printf ("DEBUG: Marking %u as bad.\n", blk);
4511 #endif
4512         ctx->fs_badblocks_count++;
4513         /*
4514          * If the block is not used, then mark it as used and return.
4515          * If it is already marked as found, this must mean that
4516          * there's an overlap between the filesystem table blocks
4517          * (bitmaps and inode table) and the bad block list.
4518          */
4519         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4520                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4521                 return 0;
4522         }
4523         /*
4524          * Try to find the where the filesystem block was used...
4525          */
4526         first_block = fs->super->s_first_data_block;
4527
4528         for (i = 0; i < fs->group_desc_count; i++ ) {
4529                 pctx->group = i;
4530                 pctx->blk = blk;
4531                 if (!ext2fs_bg_has_super(fs, i))
4532                         goto skip_super;
4533                 if (blk == first_block) {
4534                         if (i == 0) {
4535                                 if (fix_problem(ctx,
4536                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4537                                                 pctx)) {
4538                                         *block_nr = 0;
4539                                         return BLOCK_CHANGED;
4540                                 }
4541                                 return 0;
4542                         }
4543                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4544                         return 0;
4545                 }
4546                 if ((blk > first_block) &&
4547                     (blk <= first_block + fs->desc_blocks)) {
4548                         if (i == 0) {
4549                                 pctx->blk = *block_nr;
4550                                 if (fix_problem(ctx,
4551                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4552                                         *block_nr = 0;
4553                                         return BLOCK_CHANGED;
4554                                 }
4555                                 return 0;
4556                         }
4557                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4558                         return 0;
4559                 }
4560         skip_super:
4561                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4562                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4563                                 ctx->invalid_block_bitmap_flag[i]++;
4564                                 ctx->invalid_bitmaps++;
4565                         }
4566                         return 0;
4567                 }
4568                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4569                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4570                                 ctx->invalid_inode_bitmap_flag[i]++;
4571                                 ctx->invalid_bitmaps++;
4572                         }
4573                         return 0;
4574                 }
4575                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4576                     (blk < (ext2fs_inode_table_loc(fs, i) +
4577                             fs->inode_blocks_per_group))) {
4578                         /*
4579                          * If there are bad blocks in the inode table,
4580                          * the inode scan code will try to do
4581                          * something reasonable automatically.
4582                          */
4583                         return 0;
4584                 }
4585                 first_block += fs->super->s_blocks_per_group;
4586         }
4587         /*
4588          * If we've gotten to this point, then the only
4589          * possibility is that the bad block inode meta data
4590          * is using a bad block.
4591          */
4592         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4593             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4594             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4595                 p->bbcheck = 1;
4596                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4597                         *block_nr = 0;
4598                         return BLOCK_CHANGED;
4599                 }
4600                 if (e2fsck_should_abort(ctx))
4601                         return BLOCK_ABORT;
4602                 return 0;
4603         }
4604
4605         pctx->group = -1;
4606
4607         /* Warn user that the block wasn't claimed */
4608         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4609
4610         return 0;
4611 }
4612
4613 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4614                             const char *name, int num, blk64_t *new_block)
4615 {
4616         ext2_filsys fs = ctx->fs;
4617         dgrp_t          last_grp;
4618         blk64_t         old_block = *new_block;
4619         blk64_t         last_block;
4620         dgrp_t          flexbg;
4621         unsigned        flexbg_size;
4622         int             i, is_flexbg;
4623         char            *buf;
4624         struct problem_context  pctx;
4625
4626         clear_problem_context(&pctx);
4627
4628         pctx.group = group;
4629         pctx.blk = old_block;
4630         pctx.str = name;
4631
4632         /*
4633          * For flex_bg filesystems, first try to allocate the metadata
4634          * within the flex_bg, and if that fails then try finding the
4635          * space anywhere in the filesystem.
4636          */
4637         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4638         if (is_flexbg) {
4639                 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
4640                 flexbg = group / flexbg_size;
4641                 first_block = ext2fs_group_first_block2(fs,
4642                                                         flexbg_size * flexbg);
4643                 last_grp = group | (flexbg_size - 1);
4644                 if (last_grp >= fs->group_desc_count)
4645                         last_grp = fs->group_desc_count - 1;
4646                 last_block = ext2fs_group_last_block2(fs, last_grp);
4647         } else
4648                 last_block = ext2fs_group_last_block2(fs, group);
4649         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4650                                                num, ctx->block_found_map,
4651                                                new_block);
4652         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4653                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4654                                 fs->super->s_first_data_block,
4655                                 ext2fs_blocks_count(fs->super),
4656                                 num, ctx->block_found_map, new_block);
4657         if (pctx.errcode) {
4658                 pctx.num = num;
4659                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4660                 ext2fs_unmark_valid(fs);
4661                 ctx->flags |= E2F_FLAG_ABORT;
4662                 return;
4663         }
4664         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4665         if (pctx.errcode) {
4666                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4667                 ext2fs_unmark_valid(fs);
4668                 ctx->flags |= E2F_FLAG_ABORT;
4669                 return;
4670         }
4671         ext2fs_mark_super_dirty(fs);
4672         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4673         pctx.blk2 = *new_block;
4674         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4675                           PR_1_RELOC_TO), &pctx);
4676         pctx.blk2 = 0;
4677         for (i = 0; i < num; i++) {
4678                 pctx.blk = i;
4679                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4680                 if (old_block) {
4681                         pctx.errcode = io_channel_read_blk64(fs->io,
4682                                    old_block + i, 1, buf);
4683                         if (pctx.errcode)
4684                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4685                         pctx.blk = (*new_block) + i;
4686                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4687                                                               1, buf);
4688                 } else {
4689                         pctx.blk = (*new_block) + i;
4690                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4691                                                            NULL, NULL);
4692                 }
4693
4694                 if (pctx.errcode)
4695                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4696         }
4697         ext2fs_free_mem(&buf);
4698 }
4699
4700 /*
4701  * This routine gets called at the end of pass 1 if bad blocks are
4702  * detected in the superblock, group descriptors, inode_bitmaps, or
4703  * block bitmaps.  At this point, all of the blocks have been mapped
4704  * out, so we can try to allocate new block(s) to replace the bad
4705  * blocks.
4706  */
4707 static void handle_fs_bad_blocks(e2fsck_t ctx)
4708 {
4709         ext2_filsys fs = ctx->fs;
4710         dgrp_t          i;
4711         blk64_t         first_block;
4712         blk64_t         new_blk;
4713
4714         for (i = 0; i < fs->group_desc_count; i++) {
4715                 first_block = ext2fs_group_first_block2(fs, i);
4716
4717                 if (ctx->invalid_block_bitmap_flag[i]) {
4718                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4719                         new_table_block(ctx, first_block, i, _("block bitmap"),
4720                                         1, &new_blk);
4721                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4722                 }
4723                 if (ctx->invalid_inode_bitmap_flag[i]) {
4724                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4725                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4726                                         1, &new_blk);
4727                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4728                 }
4729                 if (ctx->invalid_inode_table_flag[i]) {
4730                         new_blk = ext2fs_inode_table_loc(fs, i);
4731                         new_table_block(ctx, first_block, i, _("inode table"),
4732                                         fs->inode_blocks_per_group,
4733                                         &new_blk);
4734                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4735                         ctx->flags |= E2F_FLAG_RESTART;
4736                 }
4737         }
4738         ctx->invalid_bitmaps = 0;
4739 }
4740
4741 /*
4742  * This routine marks all blocks which are used by the superblock,
4743  * group descriptors, inode bitmaps, and block bitmaps.
4744  */
4745 static void mark_table_blocks(e2fsck_t ctx)
4746 {
4747         ext2_filsys fs = ctx->fs;
4748         blk64_t b;
4749         dgrp_t  i;
4750         unsigned int    j;
4751         struct problem_context pctx;
4752
4753         clear_problem_context(&pctx);
4754
4755         for (i = 0; i < fs->group_desc_count; i++) {
4756                 pctx.group = i;
4757
4758                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4759                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4760
4761                 /*
4762                  * Mark the blocks used for the inode table
4763                  */
4764                 if (ext2fs_inode_table_loc(fs, i)) {
4765                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4766                              j < fs->inode_blocks_per_group;
4767                              j++, b++) {
4768                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4769                                                              b)) {
4770                                         pctx.blk = b;
4771                                         if (!ctx->invalid_inode_table_flag[i] &&
4772                                             fix_problem(ctx,
4773                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4774                                                 ctx->invalid_inode_table_flag[i]++;
4775                                                 ctx->invalid_bitmaps++;
4776                                         }
4777                                 } else {
4778                                     ext2fs_mark_block_bitmap2(
4779                                                 ctx->block_found_map, b);
4780                                     ext2fs_mark_block_bitmap2(
4781                                                 ctx->block_metadata_map, b);
4782                                 }
4783                         }
4784                 }
4785
4786                 /*
4787                  * Mark block used for the block bitmap
4788                  */
4789                 if (ext2fs_block_bitmap_loc(fs, i)) {
4790                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4791                                      ext2fs_block_bitmap_loc(fs, i))) {
4792                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4793                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4794                                         ctx->invalid_block_bitmap_flag[i]++;
4795                                         ctx->invalid_bitmaps++;
4796                                 }
4797                         } else {
4798                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4799                                      ext2fs_block_bitmap_loc(fs, i));
4800                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4801                                      ext2fs_block_bitmap_loc(fs, i));
4802                         }
4803                 }
4804                 /*
4805                  * Mark block used for the inode bitmap
4806                  */
4807                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4808                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4809                                      ext2fs_inode_bitmap_loc(fs, i))) {
4810                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4811                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4812                                         ctx->invalid_inode_bitmap_flag[i]++;
4813                                         ctx->invalid_bitmaps++;
4814                                 }
4815                         } else {
4816                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4817                                      ext2fs_inode_bitmap_loc(fs, i));
4818                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4819                                      ext2fs_inode_bitmap_loc(fs, i));
4820                         }
4821                 }
4822         }
4823 }
4824
4825 /*
4826  * These subroutines short circuits ext2fs_get_blocks and
4827  * ext2fs_check_directory; we use them since we already have the inode
4828  * structure, so there's no point in letting the ext2fs library read
4829  * the inode again.
4830  */
4831 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4832                                   blk_t *blocks)
4833 {
4834         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4835         int     i;
4836
4837         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4838                 return EXT2_ET_CALLBACK_NOTHANDLED;
4839
4840         for (i=0; i < EXT2_N_BLOCKS; i++)
4841                 blocks[i] = ctx->stashed_inode->i_block[i];
4842         return 0;
4843 }
4844
4845 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4846                                   struct ext2_inode *inode)
4847 {
4848         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4849
4850         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4851                 return EXT2_ET_CALLBACK_NOTHANDLED;
4852         *inode = *ctx->stashed_inode;
4853         return 0;
4854 }
4855
4856 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4857                             struct ext2_inode *inode)
4858 {
4859         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4860
4861         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4862                 (inode != ctx->stashed_inode))
4863                 *ctx->stashed_inode = *inode;
4864         return EXT2_ET_CALLBACK_NOTHANDLED;
4865 }
4866
4867 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4868 {
4869         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4870
4871         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4872                 return EXT2_ET_CALLBACK_NOTHANDLED;
4873
4874         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4875                 return EXT2_ET_NO_DIRECTORY;
4876         return 0;
4877 }
4878
4879 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4880                                         blk64_t *ret)
4881 {
4882         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4883         errcode_t       retval;
4884         blk64_t         new_block;
4885
4886         if (ctx->block_found_map) {
4887                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4888                                            &new_block);
4889                 if (retval)
4890                         return retval;
4891                 if (fs->block_map) {
4892                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4893                         ext2fs_mark_bb_dirty(fs);
4894                 }
4895         } else {
4896                 if (!fs->block_map) {
4897                         retval = ext2fs_read_block_bitmap(fs);
4898                         if (retval)
4899                                 return retval;
4900                 }
4901
4902                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4903                 if (retval)
4904                         return retval;
4905         }
4906
4907         *ret = new_block;
4908         return (0);
4909 }
4910
4911 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4912                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4913 {
4914         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4915         errcode_t       retval;
4916
4917         if (ctx->block_found_map)
4918                 return ext2fs_new_range(fs, flags, goal, len,
4919                                         ctx->block_found_map, pblk, plen);
4920
4921         if (!fs->block_map) {
4922                 retval = ext2fs_read_block_bitmap(fs);
4923                 if (retval)
4924                         return retval;
4925         }
4926
4927         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4928                                 pblk, plen);
4929 }
4930
4931 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4932 {
4933         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4934
4935         /* Never free a critical metadata block */
4936         if (ctx->block_found_map &&
4937             ctx->block_metadata_map &&
4938             inuse < 0 &&
4939             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4940                 return;
4941
4942         if (ctx->block_found_map) {
4943                 if (inuse > 0)
4944                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4945                 else
4946                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4947         }
4948 }
4949
4950 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4951                                            blk_t num, int inuse)
4952 {
4953         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4954
4955         /* Never free a critical metadata block */
4956         if (ctx->block_found_map &&
4957             ctx->block_metadata_map &&
4958             inuse < 0 &&
4959             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4960                 return;
4961
4962         if (ctx->block_found_map) {
4963                 if (inuse > 0)
4964                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4965                                                         blk, num);
4966                 else
4967                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4968                                                         blk, num);
4969         }
4970 }
4971
4972 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4973 {
4974         ext2_filsys fs = ctx->fs;
4975
4976         if (use_shortcuts) {
4977                 fs->get_blocks = pass1_get_blocks;
4978                 fs->check_directory = pass1_check_directory;
4979                 fs->read_inode = pass1_read_inode;
4980                 fs->write_inode = pass1_write_inode;
4981                 ctx->stashed_ino = 0;
4982         } else {
4983                 fs->get_blocks = 0;
4984                 fs->check_directory = 0;
4985                 fs->read_inode = 0;
4986                 fs->write_inode = 0;
4987         }
4988 }
4989
4990 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4991 {
4992         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4993         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4994                                                 e2fsck_block_alloc_stats, 0);
4995         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4996         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4997                                         e2fsck_block_alloc_stats_range, NULL);
4998 }