e2fsck/pass1.c

   1 /*
   2  * pass1.c -- pass #1 of e2fsck: sequential scan of the inode table
   3  *
   4  * Copyright (C) 1993, 1994, 1995, 1996, 1997 Theodore Ts'o.
   5  *
   6  * %Begin-Header%
   7  * This file may be redistributed under the terms of the GNU Public
   8  * License.
   9  * %End-Header%
  10  *
  11  * Pass 1 of e2fsck iterates over all the inodes in the filesystems,
  12  * and applies the following tests to each inode:
  13  *
  14  *      - The mode field of the inode must be legal.
  15  *      - The size and block count fields of the inode are correct.
  16  *      - A data block must not be used by another inode
  17  *
  18  * Pass 1 also gathers the collects the following information:
  19  *
  20  *      - A bitmap of which inodes are in use.          (inode_used_map)
  21  *      - A bitmap of which inodes are directories.     (inode_dir_map)
  22  *      - A bitmap of which inodes are regular files.   (inode_reg_map)
  23  *      - A bitmap of which inodes have bad fields.     (inode_bad_map)
  24  *      - A bitmap of which inodes are in bad blocks.   (inode_bb_map)
  25  *      - A bitmap of which inodes are imagic inodes.   (inode_imagic_map)
  26  *      - A bitmap of which inodes are casefolded.      (inode_casefold_map)
  27  *      - A bitmap of which blocks are in use.          (block_found_map)
  28  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
  29  *      - The data blocks of the directory inodes.      (dir_map)
  30  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
  31  *      - The encryption policy ID of each encrypted inode. (encrypted_files)
  32  *
  33  * Pass 1 is designed to stash away enough information so that the
  34  * other passes should not need to read in the inode information
  35  * during the normal course of a filesystem check.  (Although if an
  36  * inconsistency is detected, other passes may need to read in an
  37  * inode to fix it.)
  38  *
  39  * Note that pass 1B will be invoked if there are any duplicate blocks
  40  * found.
  41  */
  42
  43 #define _GNU_SOURCE 1 /* get strnlen() */
  44 #include "config.h"
  45 #include <string.h>
  46 #include <time.h>
  47 #ifdef HAVE_ERRNO_H
  48 #include <errno.h>
  49 #endif
  50 #include <assert.h>
  51 #ifdef HAVE_PTHREAD
  52 #include <pthread.h>
  53 #endif
  54
  55 #include "e2fsck.h"
  56 #include <ext2fs/ext2_ext_attr.h>
  57 /* todo remove this finally */
  58 #include <ext2fs/ext2fsP.h>
  59 #include <e2p/e2p.h>
  60
  61 #include "problem.h"
  62
  63 #ifdef NO_INLINE_FUNCS
  64 #define _INLINE_
  65 #else
  66 #define _INLINE_ inline
  67 #endif
  68
  69 #undef DEBUG
  70
  71 struct ea_quota {
  72         blk64_t blocks;
  73         __u64 inodes;
  74 };
  75
  76 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  77                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  78                          int ref_offset, void *priv_data);
  79 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  80                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  81                              int ref_offset, void *priv_data);
  82 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  83                          char *block_buf,
  84                          const struct ea_quota *ea_ibody_quota);
  85 static void mark_table_blocks(e2fsck_t ctx);
  86 static void alloc_bb_map(e2fsck_t ctx);
  87 static void alloc_imagic_map(e2fsck_t ctx);
  88 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
  89 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino);
  90 static void handle_fs_bad_blocks(e2fsck_t ctx);
  91 static 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;
 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_entry2(ctx->fs, entry, NULL, &hash);
 353         if (retval) {
 354                 com_err("check_large_ea_inode", retval,
 355                         _("while hashing entry with e_value_inum = %u"),
 356                         entry->e_value_inum);
 357                 fatal_error(ctx, 0);
 358         }
 359
 360         if (hash == entry->e_hash) {
 361                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 362                                                      entry->e_value_size);
 363         } else {
 364                 /* This might be an old Lustre-style ea_inode reference. */
 365                 if (inode.i_mtime == pctx->ino &&
 366                     inode.i_generation == pctx->inode->i_generation) {
 367                         *quota_blocks = 0;
 368                 } else {
 369                         /* If target inode is also missing EA_INODE flag,
 370                          * this is likely to be a bad reference.
 371                          */
 372                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 373                                 pctx->num = entry->e_value_inum;
 374                                 return PR_1_ATTR_VALUE_EA_INODE;
 375                         } else {
 376                                 pctx->num = entry->e_hash;
 377                                 return PR_1_ATTR_HASH;
 378                         }
 379                 }
 380         }
 381
 382         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 383                 pctx->num = entry->e_value_inum;
 384                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 385                         inode.i_flags |= EXT4_EA_INODE_FL;
 386                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 387                                            &inode);
 388                 } else {
 389                         return PR_1_ATTR_NO_EA_INODE_FL;
 390                 }
 391         }
 392         return 0;
 393 }
 394
 395 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 396                               struct ext2_ext_attr_entry *first, void *end)
 397 {
 398         struct ext2_ext_attr_entry *entry = first;
 399         struct ext2_ext_attr_entry *np = EXT2_EXT_ATTR_NEXT(entry);
 400
 401         while ((void *) entry < end && (void *) np < end &&
 402                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 403                 if (!entry->e_value_inum)
 404                         goto next;
 405                 if (!ctx->ea_inode_refs) {
 406                         pctx->errcode = ea_refcount_create(0,
 407                                                            &ctx->ea_inode_refs);
 408                         if (pctx->errcode) {
 409                                 pctx->num = 4;
 410                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 411                                 ctx->flags |= E2F_FLAG_ABORT;
 412                                 return;
 413                         }
 414                 }
 415                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 416                                       0);
 417         next:
 418                 entry = np;
 419                 np = EXT2_EXT_ATTR_NEXT(entry);
 420         }
 421 }
 422
 423 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 424                               struct ea_quota *ea_ibody_quota)
 425 {
 426         struct ext2_super_block *sb = ctx->fs->super;
 427         struct ext2_inode_large *inode;
 428         struct ext2_ext_attr_entry *entry;
 429         char *start, *header, *end;
 430         unsigned int storage_size, remain;
 431         problem_t problem = 0;
 432         region_t region = 0;
 433
 434         ea_ibody_quota->blocks = 0;
 435         ea_ibody_quota->inodes = 0;
 436
 437         inode = (struct ext2_inode_large *) pctx->inode;
 438         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 439                 inode->i_extra_isize;
 440         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 441                  inode->i_extra_isize;
 442         end = header + storage_size;
 443         start = header + sizeof(__u32);
 444         entry = (struct ext2_ext_attr_entry *) start;
 445
 446         /* scan all entry's headers first */
 447
 448         /* take finish entry 0UL into account */
 449         remain = storage_size - sizeof(__u32);
 450
 451         region = region_create(0, storage_size);
 452         if (!region) {
 453                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 454                 problem = 0;
 455                 ctx->flags |= E2F_FLAG_ABORT;
 456                 return;
 457         }
 458         if (region_allocate(region, 0, sizeof(__u32))) {
 459                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 460                 goto fix;
 461         }
 462
 463         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 464                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 465                 __u32 hash;
 466
 467                 if (region_allocate(region, (char *)entry - (char *)header,
 468                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 469                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 470                         goto fix;
 471                 }
 472
 473                 /* header eats this space */
 474                 remain -= sizeof(struct ext2_ext_attr_entry);
 475
 476                 /* is attribute name valid? */
 477                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 478                         pctx->num = entry->e_name_len;
 479                         problem = PR_1_ATTR_NAME_LEN;
 480                         goto fix;
 481                 }
 482
 483                 /* attribute len eats this space */
 484                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 485
 486                 if (entry->e_value_inum == 0) {
 487                         /* check value size */
 488                         if (entry->e_value_size > remain) {
 489                                 pctx->num = entry->e_value_size;
 490                                 problem = PR_1_ATTR_VALUE_SIZE;
 491                                 goto fix;
 492                         }
 493
 494                         if (entry->e_value_size &&
 495                             region_allocate(region,
 496                                             sizeof(__u32) + entry->e_value_offs,
 497                                             EXT2_EXT_ATTR_SIZE(
 498                                                 entry->e_value_size))) {
 499                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 500                                 goto fix;
 501                         }
 502
 503                         hash = ext2fs_ext_attr_hash_entry(entry,
 504                                                           start + entry->e_value_offs);
 505
 506                         /* e_hash may be 0 in older inode's ea */
 507                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 508                                 pctx->num = entry->e_hash;
 509                                 problem = PR_1_ATTR_HASH;
 510                                 goto fix;
 511                         }
 512                 } else {
 513                         blk64_t quota_blocks;
 514
 515                         problem = check_large_ea_inode(ctx, entry, pctx,
 516                                                        &quota_blocks);
 517                         if (problem != 0)
 518                                 goto fix;
 519
 520                         ea_ibody_quota->blocks += quota_blocks;
 521                         ea_ibody_quota->inodes++;
 522                 }
 523
 524                 /* If EA value is stored in external inode then it does not
 525                  * consume space here */
 526                 if (entry->e_value_inum == 0)
 527                         remain -= entry->e_value_size;
 528
 529                 entry = EXT2_EXT_ATTR_NEXT(entry);
 530         }
 531
 532         if (region_allocate(region, (char *)entry - (char *)header,
 533                             sizeof(__u32))) {
 534                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 535                 goto fix;
 536         }
 537 fix:
 538         if (region)
 539                 region_free(region);
 540         /*
 541          * it seems like a corruption. it's very unlikely we could repair
 542          * EA(s) in automatic fashion -bzzz
 543          */
 544         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 545                 inc_ea_inode_refs(ctx, pctx,
 546                                   (struct ext2_ext_attr_entry *)start, end);
 547                 return;
 548         }
 549
 550         /* simply remove all possible EA(s) */
 551         *((__u32 *)header) = 0UL;
 552         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 553                                 EXT2_INODE_SIZE(sb), "pass1");
 554         ea_ibody_quota->blocks = 0;
 555         ea_ibody_quota->inodes = 0;
 556 }
 557
 558 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 559         return (xtime & (1U << 31)) != 0 &&
 560                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 561 }
 562
 563 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 564         check_inode_extra_negative_epoch(inode->i_##xtime, \
 565                                          inode->i_##xtime##_extra)
 566
 567 /* When today's date is earlier than 2242, we assume that atimes,
 568  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 569  * actually pre-1970 dates mis-encoded.
 570  */
 571 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 572
 573 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 574                                     struct ea_quota *ea_ibody_quota)
 575 {
 576         struct ext2_super_block *sb = ctx->fs->super;
 577         struct ext2_inode_large *inode;
 578         __u32 *eamagic;
 579         int min, max;
 580
 581         ea_ibody_quota->blocks = 0;
 582         ea_ibody_quota->inodes = 0;
 583
 584         inode = (struct ext2_inode_large *) pctx->inode;
 585         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 586                 /* this isn't large inode. so, nothing to check */
 587                 return;
 588         }
 589
 590 #if 0
 591         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 592                         inode->i_extra_isize);
 593 #endif
 594         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 595         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 596         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 597         /*
 598          * For now we will allow i_extra_isize to be 0, but really
 599          * implementations should never allow i_extra_isize to be 0
 600          */
 601         if (inode->i_extra_isize &&
 602             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 603              inode->i_extra_isize & 3)) {
 604                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 605                         return;
 606                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 607                         inode->i_extra_isize = sb->s_want_extra_isize;
 608                 else
 609                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 610                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 611                                         EXT2_INODE_SIZE(sb), "pass1");
 612         }
 613
 614         /* check if there is no place for an EA header */
 615         if (inode->i_extra_isize >= max - sizeof(__u32))
 616                 return;
 617
 618         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 619                         inode->i_extra_isize);
 620         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 621                 /* it seems inode has an extended attribute(s) in body */
 622                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 623         }
 624
 625         /*
 626          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 627          * the old, invalid format, repair it.
 628          */
 629         if (((sizeof(time_t) <= 4) ||
 630              (((sizeof(time_t) > 4) &&
 631                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 632             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 634              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 635              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 636
 637                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 638                         return;
 639
 640                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 641                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 642                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 643                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 644                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 645                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 646                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 647                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 648                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 649                                         EXT2_INODE_SIZE(sb), "pass1");
 650         }
 651
 652 }
 653
 654 /*
 655  * Check to see if the inode might really be a directory, despite i_mode
 656  *
 657  * This is a lot of complexity for something for which I'm not really
 658  * convinced happens frequently in the wild.  If for any reason this
 659  * causes any problems, take this code out.
 660  * [tytso:20070331.0827EDT]
 661  */
 662 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 663                                 char *buf)
 664 {
 665         struct ext2_inode *inode = pctx->inode;
 666         struct ext2_dir_entry   *dirent;
 667         errcode_t               retval;
 668         blk64_t                 blk;
 669         unsigned int            i, rec_len, not_device = 0;
 670         int                     extent_fs;
 671         int                     inlinedata_fs;
 672
 673         /*
 674          * If the mode looks OK, we believe it.  If the first block in
 675          * the i_block array is 0, this cannot be a directory. If the
 676          * inode is extent-mapped, it is still the case that the latter
 677          * cannot be 0 - the magic number in the extent header would make
 678          * it nonzero.
 679          */
 680         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 681             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 682                 return;
 683
 684         /*
 685          * Check the block numbers in the i_block array for validity:
 686          * zero blocks are skipped (but the first one cannot be zero -
 687          * see above), other blocks are checked against the first and
 688          * max data blocks (from the the superblock) and against the
 689          * block bitmap. Any invalid block found means this cannot be
 690          * a directory.
 691          *
 692          * If there are non-zero blocks past the fourth entry, then
 693          * this cannot be a device file: we remember that for the next
 694          * check.
 695          *
 696          * For extent mapped files, we don't do any sanity checking:
 697          * just try to get the phys block of logical block 0 and run
 698          * with it.
 699          *
 700          * For inline data files, we just try to get the size of inline
 701          * data.  If it's true, we will treat it as a directory.
 702          */
 703
 704         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 705         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 706         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 707                 size_t size;
 708                 __u32 dotdot;
 709                 unsigned int rec_len2;
 710                 struct ext2_dir_entry de;
 711
 712                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 713                         return;
 714                 /*
 715                  * If the size isn't a multiple of 4, it's probably not a
 716                  * directory??
 717                  */
 718                 if (size & 3)
 719                         return;
 720                 /*
 721                  * If the first 10 bytes don't look like a directory entry,
 722                  * it's probably not a directory.
 723                  */
 724                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 725                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 726                        EXT2_DIR_REC_LEN(0));
 727                 dotdot = ext2fs_le32_to_cpu(dotdot);
 728                 de.inode = ext2fs_le32_to_cpu(de.inode);
 729                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 730                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 731                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 732                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 733                      dotdot != EXT2_ROOT_INO) ||
 734                     de.inode >= ctx->fs->super->s_inodes_count ||
 735                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 736                      de.inode != 0) ||
 737                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 738                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 739                         return;
 740                 /* device files never have a "system.data" entry */
 741                 goto isdir;
 742         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 743                 /* extent mapped */
 744                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 745                                  &blk))
 746                         return;
 747                 /* device files are never extent mapped */
 748                 not_device++;
 749         } else {
 750                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 751                         blk = inode->i_block[i];
 752                         if (!blk)
 753                                 continue;
 754                         if (i >= 4)
 755                                 not_device++;
 756
 757                         if (blk < ctx->fs->super->s_first_data_block ||
 758                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 759                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 760                                                            blk))
 761                                 return; /* Invalid block, can't be dir */
 762                 }
 763                 blk = inode->i_block[0];
 764         }
 765
 766         /*
 767          * If the mode says this is a device file and the i_links_count field
 768          * is sane and we have not ruled it out as a device file previously,
 769          * we declare it a device file, not a directory.
 770          */
 771         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 772             (inode->i_links_count == 1) && !not_device)
 773                 return;
 774
 775         /* read the first block */
 776         ehandler_operation(_("reading directory block"));
 777         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 778         ehandler_operation(0);
 779         if (retval)
 780                 return;
 781
 782         dirent = (struct ext2_dir_entry *) buf;
 783         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 784         if (retval)
 785                 return;
 786         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 787             (dirent->name[0] != '.') ||
 788             (dirent->inode != pctx->ino) ||
 789             (rec_len < 12) ||
 790             (rec_len % 4) ||
 791             (rec_len >= ctx->fs->blocksize - 12))
 792                 return;
 793
 794         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 795         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 796         if (retval)
 797                 return;
 798         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 799             (dirent->name[0] != '.') ||
 800             (dirent->name[1] != '.') ||
 801             (rec_len < 12) ||
 802             (rec_len % 4))
 803                 return;
 804
 805 isdir:
 806         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 807                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 808                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 809                                         EXT2_INODE_SIZE(ctx->fs->super),
 810                                         "check_is_really_dir");
 811         }
 812 }
 813
 814 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 815                                      int flags, ext2_icount_t hint,
 816                                      ext2_icount_t *ret)
 817 {
 818         unsigned int            threshold;
 819         unsigned int            save_type;
 820         ext2_ino_t              num_dirs;
 821         errcode_t               retval;
 822         char                    *tdb_dir;
 823         int                     enable;
 824
 825         *ret = 0;
 826
 827         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 828                            &tdb_dir);
 829         profile_get_uint(ctx->profile, "scratch_files",
 830                          "numdirs_threshold", 0, 0, &threshold);
 831         profile_get_boolean(ctx->profile, "scratch_files",
 832                             "icount", 0, 1, &enable);
 833
 834         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 835         if (retval)
 836                 num_dirs = 1024;        /* Guess */
 837
 838         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 839             (!threshold || num_dirs > threshold)) {
 840                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 841                                                   flags, ret);
 842                 if (retval == 0)
 843                         return 0;
 844         }
 845         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 846                                &save_type);
 847         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 848                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 849         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 850         ctx->fs->default_bitmap_type = save_type;
 851         return retval;
 852 }
 853
 854 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 855                                             e2fsck_t ctx,
 856                                             struct problem_context *pctx)
 857 {
 858         errcode_t retval;
 859         struct ext2_inode_large inode;
 860
 861         /*
 862          * Reread inode.  If we don't see checksum error, then this inode
 863          * has been fixed elsewhere.
 864          */
 865         ctx->stashed_ino = 0;
 866         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 867                                         sizeof(inode));
 868         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 869                 return retval;
 870         if (!retval)
 871                 return 0;
 872
 873         /*
 874          * Checksum still doesn't match.  That implies that the inode passes
 875          * all the sanity checks, so maybe the checksum is simply corrupt.
 876          * See if the user will go for fixing that.
 877          */
 878         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 879                 return 0;
 880
 881         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 882                                          sizeof(inode));
 883         return retval;
 884 }
 885
 886 static void reserve_block_for_root_repair(e2fsck_t ctx)
 887 {
 888         blk64_t         blk = 0;
 889         errcode_t       err;
 890         ext2_filsys     fs = ctx->fs;
 891
 892         ctx->root_repair_block = 0;
 893         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 894                 return;
 895
 896         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 897         if (err)
 898                 return;
 899         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 900         ctx->root_repair_block = blk;
 901 }
 902
 903 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 904 {
 905         blk64_t         blk = 0;
 906         errcode_t       err;
 907         ext2_filsys     fs = ctx->fs;
 908         static const char name[] = "lost+found";
 909         ext2_ino_t      ino;
 910
 911         ctx->lnf_repair_block = 0;
 912         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 913                 return;
 914
 915         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 916         if (err)
 917                 return;
 918         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 919         ctx->lnf_repair_block = blk;
 920 }
 921
 922 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 923                                          size_t *sz)
 924 {
 925         void *p;
 926         struct ext2_xattr_handle *handle;
 927         errcode_t retval;
 928
 929         retval = ext2fs_xattrs_open(fs, ino, &handle);
 930         if (retval)
 931                 return retval;
 932
 933         retval = ext2fs_xattrs_read(handle);
 934         if (retval)
 935                 goto err;
 936
 937         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 938         if (retval)
 939                 goto err;
 940         ext2fs_free_mem(&p);
 941 err:
 942         (void) ext2fs_xattrs_close(&handle);
 943         return retval;
 944 }
 945
 946 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 947                                     struct problem_context *pctx,
 948                                     int failed_csum)
 949 {
 950         if (!failed_csum)
 951                 return;
 952
 953         /*
 954          * If the inode failed the checksum and the user didn't
 955          * clear the inode, test the checksum again -- if it still
 956          * fails, ask the user if the checksum should be corrected.
 957          */
 958         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 959         if (pctx->errcode)
 960                 ctx->flags |= E2F_FLAG_ABORT;
 961 }
 962 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 963         do { \
 964                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 965                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 966                         return; \
 967         } while (0)
 968
 969 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 970 {
 971         __u32 x;
 972         int i;
 973
 974         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 975                 x = inode->i_block[i];
 976 #ifdef WORDS_BIGENDIAN
 977                 x = ext2fs_swab32(x);
 978 #endif
 979                 if (x >= ext2fs_blocks_count(fs->super))
 980                         return 0;
 981         }
 982
 983         return 1;
 984 }
 985
 986 /*
 987  * Figure out what to do with an inode that has both extents and inline data
 988  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 989  */
 990 static int fix_inline_data_extents_file(e2fsck_t ctx,
 991                                         ext2_ino_t ino,
 992                                         struct ext2_inode *inode,
 993                                         int inode_size,
 994                                         struct problem_context *pctx)
 995 {
 996         size_t max_inline_ea_size;
 997         ext2_filsys fs = ctx->fs;
 998         int dirty = 0;
 999
1000         /* Both feature flags not set?  Just run the regular checks */
1001         if (!ext2fs_has_feature_extents(fs->super) &&
1002             !ext2fs_has_feature_inline_data(fs->super))
1003                 return 0;
1004
1005         /* Clear both flags if it's a special file */
1006         if (LINUX_S_ISCHR(inode->i_mode) ||
1007             LINUX_S_ISBLK(inode->i_mode) ||
1008             LINUX_S_ISFIFO(inode->i_mode) ||
1009             LINUX_S_ISSOCK(inode->i_mode)) {
1010                 check_extents_inlinedata(ctx, pctx);
1011                 return 0;
1012         }
1013
1014         /* If it looks like an extent tree, try to clear inlinedata */
1015         if (ext2fs_extent_header_verify(inode->i_block,
1016                                  sizeof(inode->i_block)) == 0 &&
1017             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1018                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1019                 dirty = 1;
1020                 goto out;
1021         }
1022
1023         /* If it looks short enough to be inline data, try to clear extents */
1024         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1025                 max_inline_ea_size = inode_size -
1026                                      (EXT2_GOOD_OLD_INODE_SIZE +
1027                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1028         else
1029                 max_inline_ea_size = 0;
1030         if (EXT2_I_SIZE(inode) <
1031             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1032             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1033                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1034                 dirty = 1;
1035                 goto out;
1036         }
1037
1038         /*
1039          * Too big for inline data, but no evidence of extent tree -
1040          * maybe it's a block map file?  If the mappings all look valid?
1041          */
1042         if (could_be_block_map(fs, inode) &&
1043             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1044 #ifdef WORDS_BIGENDIAN
1045                 int i;
1046
1047                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1048                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1049 #endif
1050
1051                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1052                 dirty = 1;
1053                 goto out;
1054         }
1055
1056         /* Oh well, just clear the busted inode. */
1057         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1058                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1059                 return -1;
1060         }
1061
1062 out:
1063         if (dirty)
1064                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1065
1066         return 0;
1067 }
1068
1069 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1070 {
1071         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1072         dgrp_t start = *group, grp;
1073         blk64_t blocks_to_read = 0;
1074         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1075
1076         if (ctx->readahead_kb == 0)
1077                 goto out;
1078
1079         /* Keep iterating groups until we have enough to readahead */
1080         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1081         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1082                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1083                         continue;
1084                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1085                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1086                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1087                                         inodes_per_block;
1088                 if (blocks_to_read * ctx->fs->blocksize >
1089                     ctx->readahead_kb * 1024)
1090                         break;
1091         }
1092
1093         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1094                                grp - start + 1);
1095         if (err == EAGAIN) {
1096                 ctx->readahead_kb /= 2;
1097                 err = 0;
1098         }
1099
1100 out:
1101         if (err) {
1102                 /* Error; disable itable readahead */
1103                 *group = ctx->fs->group_desc_count;
1104                 *next_ino = ctx->fs->super->s_inodes_count;
1105         } else {
1106                 /*
1107                  * Don't do more readahead until we've reached the first inode
1108                  * of the last inode scan buffer block for the last group.
1109                  */
1110                 *group = grp + 1;
1111                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1112                                      ctx->inode_buffer_blocks :
1113                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1114                                     ctx->fs->blocksize /
1115                                     EXT2_INODE_SIZE(ctx->fs->super);
1116                 inodes_in_group--;
1117                 *next_ino = inodes_in_group -
1118                             (inodes_in_group % inodes_per_buffer) + 1 +
1119                             (grp * ctx->fs->super->s_inodes_per_group);
1120         }
1121 }
1122
1123 /*
1124  * Check if the passed ino is one of the used superblock quota inodes.
1125  *
1126  * Before the quota inodes were journaled, older superblock quota inodes
1127  * were just regular files in the filesystem and not reserved inodes.  This
1128  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1129  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1130  */
1131 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1132 {
1133         enum quota_type qtype;
1134
1135         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1136                 if (*quota_sb_inump(sb, qtype) == ino)
1137                         return 1;
1138
1139         return 0;
1140 }
1141
1142 /*
1143  * Check if the passed ino is one of the reserved quota inodes.
1144  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1145  * inodes.  These inodes may or may not be in use by the quota feature.
1146  */
1147 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1148 {
1149         enum quota_type qtype;
1150
1151         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1152                 if (quota_type2inum(qtype, fs->super) == ino)
1153                         return 1;
1154
1155         return 0;
1156 }
1157
1158 static int e2fsck_should_abort(e2fsck_t ctx)
1159 {
1160         e2fsck_t global_ctx;
1161
1162         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1163                 return 1;
1164
1165         if (ctx->global_ctx) {
1166                 global_ctx = ctx->global_ctx;
1167                 if (global_ctx->flags & E2F_FLAG_SIGNAL_MASK)
1168                         return 1;
1169         }
1170         return 0;
1171 }
1172
1173 void e2fsck_pass1_run(e2fsck_t ctx)
1174 {
1175         int     i;
1176         __u64   max_sizes;
1177         ext2_filsys fs = ctx->fs;
1178         ext2_ino_t      ino = 0;
1179         struct ext2_inode *inode = NULL;
1180         ext2_inode_scan scan = NULL;
1181         char            *block_buf = NULL;
1182 #ifdef RESOURCE_TRACK
1183         struct resource_track   rtrack;
1184 #endif
1185         unsigned char   frag, fsize;
1186         struct          problem_context pctx;
1187         struct          scan_callback_struct scan_struct;
1188         struct ext2_super_block *sb = ctx->fs->super;
1189         const char      *old_op;
1190         const char      *eop_next_inode = _("getting next inode from scan");
1191         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1192         int             low_dtime_check = 1;
1193         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1194         unsigned int    bufsize;
1195         int             failed_csum = 0;
1196         ext2_ino_t      ino_threshold = 0;
1197         dgrp_t          ra_group = 0;
1198         struct ea_quota ea_ibody_quota;
1199
1200         init_resource_track(&rtrack, ctx->fs->io);
1201         clear_problem_context(&pctx);
1202
1203         /* If we can do readahead, figure out how many groups to pull in. */
1204         if (!e2fsck_can_readahead(ctx->fs))
1205                 ctx->readahead_kb = 0;
1206         else if (ctx->readahead_kb == ~0ULL)
1207                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1208         pass1_readahead(ctx, &ra_group, &ino_threshold);
1209
1210         if (!(ctx->options & E2F_OPT_PREEN))
1211                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1212
1213         if (ext2fs_has_feature_dir_index(fs->super) &&
1214             !(ctx->options & E2F_OPT_NO)) {
1215                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1216                         ctx->dirs_to_hash = 0;
1217         }
1218
1219 #ifdef MTRACE
1220         mtrace_print("Pass 1");
1221 #endif
1222
1223 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1224
1225         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1226                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1227                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1228                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1229                 max_sizes = (max_sizes * (1UL << i));
1230                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1231         }
1232 #undef EXT2_BPP
1233
1234         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1235         extent_fs = ext2fs_has_feature_extents(sb);
1236         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1237         casefold_fs = ext2fs_has_feature_casefold(sb);
1238
1239         /*
1240          * Allocate bitmaps structures
1241          */
1242         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1243                                                     EXT2FS_BMAP64_RBTREE,
1244                                                     "inode_used_map",
1245                                                     &ctx->inode_used_map);
1246         if (pctx.errcode) {
1247                 pctx.num = 1;
1248                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1249                 ctx->flags |= E2F_FLAG_ABORT;
1250                 return;
1251         }
1252         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1253                         _("directory inode map"),
1254                         EXT2FS_BMAP64_AUTODIR,
1255                         "inode_dir_map", &ctx->inode_dir_map);
1256         if (pctx.errcode) {
1257                 pctx.num = 2;
1258                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1259                 ctx->flags |= E2F_FLAG_ABORT;
1260                 return;
1261         }
1262         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1263                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1264                         "inode_reg_map", &ctx->inode_reg_map);
1265         if (pctx.errcode) {
1266                 pctx.num = 6;
1267                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1268                 ctx->flags |= E2F_FLAG_ABORT;
1269                 return;
1270         }
1271         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1272                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1273                         "block_found_map", &ctx->block_found_map);
1274         if (pctx.errcode) {
1275                 pctx.num = 1;
1276                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1277                 ctx->flags |= E2F_FLAG_ABORT;
1278                 return;
1279         }
1280         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1281                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1282                         "block_metadata_map", &ctx->block_metadata_map);
1283         if (pctx.errcode) {
1284                 pctx.num = 1;
1285                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1286                 ctx->flags |= E2F_FLAG_ABORT;
1287                 return;
1288         }
1289         if (casefold_fs) {
1290                 pctx.errcode =
1291                         e2fsck_allocate_inode_bitmap(fs,
1292                                                      _("inode casefold map"),
1293                                                      EXT2FS_BMAP64_RBTREE,
1294                                                      "inode_casefold_map",
1295                                                      &ctx->inode_casefold_map);
1296                 if (pctx.errcode) {
1297                         pctx.num = 1;
1298                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1299                         ctx->flags |= E2F_FLAG_ABORT;
1300                         return;
1301                 }
1302         }
1303         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1304                                            &ctx->inode_link_info);
1305         if (pctx.errcode) {
1306                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1307                 ctx->flags |= E2F_FLAG_ABORT;
1308                 return;
1309         }
1310         bufsize = inode_size;
1311         if (bufsize < sizeof(struct ext2_inode_large))
1312                 bufsize = sizeof(struct ext2_inode_large);
1313         inode = (struct ext2_inode *)
1314                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1315
1316         inodes_to_process = (struct process_inode_block *)
1317                 e2fsck_allocate_memory(ctx,
1318                                        (ctx->process_inode_size *
1319                                         sizeof(struct process_inode_block)),
1320                                        "array of inodes to process");
1321         process_inode_count = 0;
1322
1323         pctx.errcode = ext2fs_init_dblist(fs, 0);
1324         if (pctx.errcode) {
1325                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1326                 ctx->flags |= E2F_FLAG_ABORT;
1327                 goto endit;
1328         }
1329
1330         /*
1331          * If the last orphan field is set, clear it, since the pass1
1332          * processing will automatically find and clear the orphans.
1333          * In the future, we may want to try using the last_orphan
1334          * linked list ourselves, but for now, we clear it so that the
1335          * ext3 mount code won't get confused.
1336          */
1337         if (!(ctx->options & E2F_OPT_READONLY)) {
1338                 if (fs->super->s_last_orphan) {
1339                         fs->super->s_last_orphan = 0;
1340                         ext2fs_mark_super_dirty(fs);
1341                 }
1342         }
1343
1344         mark_table_blocks(ctx);
1345         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1346                                                 &ctx->block_found_map);
1347         if (pctx.errcode) {
1348                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1349                 ctx->flags |= E2F_FLAG_ABORT;
1350                 goto endit;
1351         }
1352         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1353                                                     "block interate buffer");
1354         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1355                 e2fsck_use_inode_shortcuts(ctx, 1);
1356         e2fsck_intercept_block_allocations(ctx);
1357         old_op = ehandler_operation(_("opening inode scan"));
1358         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1359                                               &scan);
1360         ehandler_operation(old_op);
1361         if (pctx.errcode) {
1362                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1363                 ctx->flags |= E2F_FLAG_ABORT;
1364                 goto endit;
1365         }
1366         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1367                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1368         ctx->stashed_inode = inode;
1369         scan_struct.ctx = ctx;
1370         scan_struct.block_buf = block_buf;
1371         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1372         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1373                                               ctx->fs->group_desc_count)))
1374                 goto endit;
1375         if ((fs->super->s_wtime &&
1376              fs->super->s_wtime < fs->super->s_inodes_count) ||
1377             (fs->super->s_mtime &&
1378              fs->super->s_mtime < fs->super->s_inodes_count) ||
1379             (fs->super->s_mkfs_time &&
1380              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1381                 low_dtime_check = 0;
1382
1383         if (ext2fs_has_feature_mmp(fs->super) &&
1384             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1385             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1386                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1387                                           fs->super->s_mmp_block);
1388
1389         /* Set up ctx->lost_and_found if possible */
1390         (void) e2fsck_get_lost_and_found(ctx, 0);
1391
1392 #ifdef HAVE_PTHREAD
1393         if (ctx->global_ctx) {
1394                 if (ctx->options & E2F_OPT_DEBUG &&
1395                     ctx->options & E2F_OPT_MULTITHREAD)
1396                         fprintf(stderr, "thread %d jumping to group %d\n",
1397                                         ctx->thread_info.et_thread_index,
1398                                         ctx->thread_info.et_group_start);
1399                 pctx.errcode = ext2fs_inode_scan_goto_blockgroup(scan,
1400                                         ctx->thread_info.et_group_start);
1401                 if (pctx.errcode) {
1402                         fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1403                         ctx->flags |= E2F_FLAG_ABORT;
1404                         goto endit;
1405                 }
1406         }
1407 #endif
1408
1409         while (1) {
1410                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1411                         if (e2fsck_mmp_update(fs))
1412                                 fatal_error(ctx, 0);
1413                 }
1414                 old_op = ehandler_operation(eop_next_inode);
1415                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1416                                                           inode, inode_size);
1417                 if (ino > ino_threshold)
1418                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1419                 ehandler_operation(old_op);
1420                 if (e2fsck_should_abort(ctx))
1421                         goto endit;
1422                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1423                         /*
1424                          * If badblocks says badblocks is bad, offer to clear
1425                          * the list, update the in-core bb list, and restart
1426                          * the inode scan.
1427                          */
1428                         if (ino == EXT2_BAD_INO &&
1429                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1430                                         &pctx)) {
1431                                 errcode_t err;
1432
1433                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1434                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1435                                 ctx->fs->badblocks = NULL;
1436                                 err = ext2fs_read_bb_inode(ctx->fs,
1437                                                         &ctx->fs->badblocks);
1438                                 if (err) {
1439                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1440                                                     &pctx);
1441                                         ctx->flags |= E2F_FLAG_ABORT;
1442                                 } else
1443                                         ctx->flags |= E2F_FLAG_RESTART;
1444                                 goto endit;
1445                         }
1446                         if (!ctx->inode_bb_map)
1447                                 alloc_bb_map(ctx);
1448                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1449                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1450                         continue;
1451                 }
1452                 if (pctx.errcode == EXT2_ET_SCAN_FINISHED)
1453                         break;
1454                 if (pctx.errcode &&
1455                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1456                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1457                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1458                         ctx->flags |= E2F_FLAG_ABORT;
1459                         goto endit;
1460                 }
1461                 if (!ino)
1462                         break;
1463 #ifdef HAVE_PTHREAD
1464                 if (ctx->global_ctx)
1465                         ctx->thread_info.et_inode_number++;
1466 #endif
1467                 pctx.ino = ino;
1468                 pctx.inode = inode;
1469                 ctx->stashed_ino = ino;
1470
1471                 /* Clear trashed inode? */
1472                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1473                     inode->i_links_count > 0 &&
1474                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1475                         pctx.errcode = 0;
1476                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1477                 }
1478                 failed_csum = pctx.errcode != 0;
1479
1480                 /*
1481                  * Check for inodes who might have been part of the
1482                  * orphaned list linked list.  They should have gotten
1483                  * dealt with by now, unless the list had somehow been
1484                  * corrupted.
1485                  *
1486                  * FIXME: In the future, inodes which are still in use
1487                  * (and which are therefore) pending truncation should
1488                  * be handled specially.  Right now we just clear the
1489                  * dtime field, and the normal e2fsck handling of
1490                  * inodes where i_size and the inode blocks are
1491                  * inconsistent is to fix i_size, instead of releasing
1492                  * the extra blocks.  This won't catch the inodes that
1493                  * was at the end of the orphan list, but it's better
1494                  * than nothing.  The right answer is that there
1495                  * shouldn't be any bugs in the orphan list handling.  :-)
1496                  */
1497                 if (inode->i_dtime && low_dtime_check &&
1498                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1499                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1500                                 inode->i_dtime = inode->i_links_count ?
1501                                         0 : ctx->now;
1502                                 e2fsck_write_inode(ctx, ino, inode,
1503                                                    "pass1");
1504                                 failed_csum = 0;
1505                         }
1506                 }
1507
1508                 if (inode->i_links_count) {
1509                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1510                                            ino, inode->i_links_count);
1511                         if (pctx.errcode) {
1512                                 pctx.num = inode->i_links_count;
1513                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1514                                 ctx->flags |= E2F_FLAG_ABORT;
1515                                 goto endit;
1516                         }
1517                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1518                            !quota_inum_is_reserved(fs, ino)) {
1519                         if (!inode->i_dtime && inode->i_mode) {
1520                                 if (fix_problem(ctx,
1521                                             PR_1_ZERO_DTIME, &pctx)) {
1522                                         inode->i_dtime = ctx->now;
1523                                         e2fsck_write_inode(ctx, ino, inode,
1524                                                            "pass1");
1525                                         failed_csum = 0;
1526                                 }
1527                         }
1528                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1529                         continue;
1530                 }
1531
1532                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1533                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1534                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1535                      (!casefold_fs &&
1536                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1537                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1538                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1539                 }
1540
1541                 /* Conflicting inlinedata/extents inode flags? */
1542                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1543                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1544                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1545                                                                inode_size,
1546                                                                &pctx);
1547                         if (res < 0) {
1548                                 /* skip FINISH_INODE_LOOP */
1549                                 continue;
1550                         }
1551                 }
1552
1553                 /* Test for incorrect inline_data flags settings. */
1554                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1555                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1556                         size_t size = 0;
1557
1558                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1559                         if (!pctx.errcode &&
1560                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1561                                 ext2fs_set_feature_inline_data(sb);
1562                                 ext2fs_mark_super_dirty(fs);
1563                                 inlinedata_fs = 1;
1564                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1565                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1566                                 /* skip FINISH_INODE_LOOP */
1567                                 continue;
1568                         }
1569                 }
1570
1571                 /* Test for inline data flag but no attr */
1572                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1573                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1574                         size_t size = 0;
1575                         errcode_t err;
1576                         int flags;
1577
1578                         flags = fs->flags;
1579                         if (failed_csum)
1580                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1581                         err = get_inline_data_ea_size(fs, ino, &size);
1582                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1583                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1584
1585                         switch (err) {
1586                         case 0:
1587                                 /* Everything is awesome... */
1588                                 break;
1589                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1590                         case EXT2_ET_BAD_EA_HASH:
1591                         case EXT2_ET_BAD_EA_HEADER:
1592                         case EXT2_ET_EA_BAD_NAME_LEN:
1593                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1594                         case EXT2_ET_EA_KEY_NOT_FOUND:
1595                         case EXT2_ET_EA_NO_SPACE:
1596                         case EXT2_ET_MISSING_EA_FEATURE:
1597                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1598                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1599                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1600                         case EXT2_ET_NO_INLINE_DATA:
1601                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1602                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1603                         case EXT2_ET_EA_INODE_CORRUPTED:
1604                                 /* broken EA or no system.data EA; truncate */
1605                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1606                                                 &pctx)) {
1607                                         err = ext2fs_inode_size_set(fs, inode, 0);
1608                                         if (err) {
1609                                                 pctx.errcode = err;
1610                                                 ctx->flags |= E2F_FLAG_ABORT;
1611                                                 goto endit;
1612                                         }
1613                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1614                                         memset(&inode->i_block, 0,
1615                                                sizeof(inode->i_block));
1616                                         e2fsck_write_inode(ctx, ino, inode,
1617                                                            "pass1");
1618                                         failed_csum = 0;
1619                                 }
1620                                 break;
1621                         default:
1622                                 /* Some other kind of non-xattr error? */
1623                                 pctx.errcode = err;
1624                                 ctx->flags |= E2F_FLAG_ABORT;
1625                                 goto endit;
1626                         }
1627                 }
1628
1629                 /*
1630                  * Test for incorrect extent flag settings.
1631                  *
1632                  * On big-endian machines we must be careful:
1633                  * When the inode is read, the i_block array is not swapped
1634                  * if the extent flag is set.  Therefore if we are testing
1635                  * for or fixing a wrongly-set flag, we must potentially
1636                  * (un)swap before testing, or after fixing.
1637                  */
1638
1639                 /*
1640                  * In this case the extents flag was set when read, so
1641                  * extent_header_verify is ok.  If the inode is cleared,
1642                  * no need to swap... so no extra swapping here.
1643                  */
1644                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1645                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1646                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1647                         if ((ext2fs_extent_header_verify(inode->i_block,
1648                                                  sizeof(inode->i_block)) == 0) &&
1649                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1650                                 ext2fs_set_feature_extents(sb);
1651                                 ext2fs_mark_super_dirty(fs);
1652                                 extent_fs = 1;
1653                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1654                         clear_inode:
1655                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1656                                 if (ino == EXT2_BAD_INO)
1657                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1658                                                                  ino);
1659                                 /* skip FINISH_INODE_LOOP */
1660                                 continue;
1661                         }
1662                 }
1663
1664                 /*
1665                  * For big-endian machines:
1666                  * If the inode didn't have the extents flag set when it
1667                  * was read, then the i_blocks array was swapped.  To test
1668                  * as an extents header, we must swap it back first.
1669                  * IF we then set the extents flag, the entire i_block
1670                  * array must be un/re-swapped to make it proper extents data.
1671                  */
1672                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1673                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1674                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1675                     (LINUX_S_ISREG(inode->i_mode) ||
1676                      LINUX_S_ISDIR(inode->i_mode))) {
1677                         void *ehp;
1678 #ifdef WORDS_BIGENDIAN
1679                         __u32 tmp_block[EXT2_N_BLOCKS];
1680
1681                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1682                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1683                         ehp = tmp_block;
1684 #else
1685                         ehp = inode->i_block;
1686 #endif
1687                         if ((ext2fs_extent_header_verify(ehp,
1688                                          sizeof(inode->i_block)) == 0) &&
1689                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1690                                 inode->i_flags |= EXT4_EXTENTS_FL;
1691 #ifdef WORDS_BIGENDIAN
1692                                 memcpy(inode->i_block, tmp_block,
1693                                        sizeof(inode->i_block));
1694 #endif
1695                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1696                                 failed_csum = 0;
1697                         }
1698                 }
1699
1700                 if (ino == EXT2_BAD_INO) {
1701                         struct process_block_struct pb;
1702
1703                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1704                              inode->i_gid || inode->i_links_count ||
1705                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1706                              inode->i_file_acl) &&
1707                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1708                                 memset(inode, 0, sizeof(struct ext2_inode));
1709                                 e2fsck_write_inode(ctx, ino, inode,
1710                                                    "clear bad inode");
1711                                 failed_csum = 0;
1712                         }
1713
1714                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1715                                                           &pb.fs_meta_blocks);
1716                         if (pctx.errcode) {
1717                                 pctx.num = 4;
1718                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1719                                 ctx->flags |= E2F_FLAG_ABORT;
1720                                 goto endit;
1721                         }
1722                         pb.ino = EXT2_BAD_INO;
1723                         pb.num_blocks = pb.last_block = 0;
1724                         pb.last_db_block = -1;
1725                         pb.num_illegal_blocks = 0;
1726                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1727                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1728                         pb.inode = inode;
1729                         pb.pctx = &pctx;
1730                         pb.ctx = ctx;
1731                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1732                                      block_buf, process_bad_block, &pb);
1733                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1734                         if (pctx.errcode) {
1735                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1736                                 ctx->flags |= E2F_FLAG_ABORT;
1737                                 goto endit;
1738                         }
1739                         if (pb.bbcheck)
1740                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1741                                 ctx->flags |= E2F_FLAG_ABORT;
1742                                 goto endit;
1743                         }
1744                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1745                         clear_problem_context(&pctx);
1746                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1747                         continue;
1748                 } else if (ino == EXT2_ROOT_INO) {
1749                         /*
1750                          * Make sure the root inode is a directory; if
1751                          * not, offer to clear it.  It will be
1752                          * regenerated in pass #3.
1753                          */
1754                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1755                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1756                                         goto clear_inode;
1757                         }
1758                         /*
1759                          * If dtime is set, offer to clear it.  mke2fs
1760                          * version 0.2b created filesystems with the
1761                          * dtime field set for the root and lost+found
1762                          * directories.  We won't worry about
1763                          * /lost+found, since that can be regenerated
1764                          * easily.  But we will fix the root directory
1765                          * as a special case.
1766                          */
1767                         if (inode->i_dtime && inode->i_links_count) {
1768                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1769                                         inode->i_dtime = 0;
1770                                         e2fsck_write_inode(ctx, ino, inode,
1771                                                            "pass1");
1772                                         failed_csum = 0;
1773                                 }
1774                         }
1775                 } else if (ino == EXT2_JOURNAL_INO) {
1776                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1777                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1778                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1779                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1780                                                 &pctx)) {
1781                                         inode->i_mode = LINUX_S_IFREG;
1782                                         e2fsck_write_inode(ctx, ino, inode,
1783                                                            "pass1");
1784                                         failed_csum = 0;
1785                                 }
1786                                 check_blocks(ctx, &pctx, block_buf, NULL);
1787                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1788                                 continue;
1789                         }
1790                         if ((inode->i_links_count ||
1791                              inode->i_blocks || inode->i_block[0]) &&
1792                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1793                                         &pctx)) {
1794                                 memset(inode, 0, inode_size);
1795                                 ext2fs_icount_store(ctx->inode_link_info,
1796                                                     ino, 0);
1797                                 e2fsck_write_inode_full(ctx, ino, inode,
1798                                                         inode_size, "pass1");
1799                                 failed_csum = 0;
1800                         }
1801                 } else if (quota_inum_is_reserved(fs, ino)) {
1802                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1803                         if (ext2fs_has_feature_quota(fs->super) &&
1804                             quota_inum_is_super(fs->super, ino)) {
1805                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1806                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1807                                                         &pctx)) {
1808                                         inode->i_mode = LINUX_S_IFREG;
1809                                         e2fsck_write_inode(ctx, ino, inode,
1810                                                         "pass1");
1811                                         failed_csum = 0;
1812                                 }
1813                                 check_blocks(ctx, &pctx, block_buf, NULL);
1814                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1815                                 continue;
1816                         }
1817                         if ((inode->i_links_count ||
1818                              inode->i_blocks || inode->i_block[0]) &&
1819                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1820                                         &pctx)) {
1821                                 memset(inode, 0, inode_size);
1822                                 ext2fs_icount_store(ctx->inode_link_info,
1823                                                     ino, 0);
1824                                 e2fsck_write_inode_full(ctx, ino, inode,
1825                                                         inode_size, "pass1");
1826                                 failed_csum = 0;
1827                         }
1828                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1829                         problem_t problem = 0;
1830
1831                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1832                         if (ino == EXT2_BOOT_LOADER_INO) {
1833                                 if (LINUX_S_ISDIR(inode->i_mode))
1834                                         problem = PR_1_RESERVED_BAD_MODE;
1835                         } else if (ino == EXT2_RESIZE_INO) {
1836                                 if (inode->i_mode &&
1837                                     !LINUX_S_ISREG(inode->i_mode))
1838                                         problem = PR_1_RESERVED_BAD_MODE;
1839                         } else {
1840                                 if (inode->i_mode != 0)
1841                                         problem = PR_1_RESERVED_BAD_MODE;
1842                         }
1843                         if (problem) {
1844                                 if (fix_problem(ctx, problem, &pctx)) {
1845                                         inode->i_mode = 0;
1846                                         e2fsck_write_inode(ctx, ino, inode,
1847                                                            "pass1");
1848                                         failed_csum = 0;
1849                                 }
1850                         }
1851                         check_blocks(ctx, &pctx, block_buf, NULL);
1852                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1853                         continue;
1854                 }
1855
1856                 if (!inode->i_links_count) {
1857                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1858                         continue;
1859                 }
1860                 /*
1861                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1862                  * deleted files.  Oops.
1863                  *
1864                  * Since all new ext2 implementations get this right,
1865                  * we now assume that the case of non-zero
1866                  * i_links_count and non-zero dtime means that we
1867                  * should keep the file, not delete it.
1868                  *
1869                  */
1870                 if (inode->i_dtime) {
1871                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1872                                 inode->i_dtime = 0;
1873                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1874                                 failed_csum = 0;
1875                         }
1876                 }
1877
1878                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1879                 switch (fs->super->s_creator_os) {
1880                     case EXT2_OS_HURD:
1881                         frag = inode->osd2.hurd2.h_i_frag;
1882                         fsize = inode->osd2.hurd2.h_i_fsize;
1883                         break;
1884                     default:
1885                         frag = fsize = 0;
1886                 }
1887
1888                 if (inode->i_faddr || frag || fsize ||
1889                     (!ext2fs_has_feature_largedir(fs->super) &&
1890                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1891                         mark_inode_bad(ctx, ino);
1892                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1893                     !ext2fs_has_feature_64bit(fs->super) &&
1894                     inode->osd2.linux2.l_i_file_acl_high != 0)
1895                         mark_inode_bad(ctx, ino);
1896                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1897                     !ext2fs_has_feature_huge_file(fs->super) &&
1898                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1899                         mark_inode_bad(ctx, ino);
1900                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1901                         if (imagic_fs) {
1902                                 if (!ctx->inode_imagic_map)
1903                                         alloc_imagic_map(ctx);
1904                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1905                                                          ino);
1906                         } else {
1907                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1908                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1909                                         e2fsck_write_inode(ctx, ino,
1910                                                            inode, "pass1");
1911                                         failed_csum = 0;
1912                                 }
1913                         }
1914                 }
1915
1916                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1917                 check_is_really_dir(ctx, &pctx, block_buf);
1918
1919                 /*
1920                  * ext2fs_inode_has_valid_blocks2 does not actually look
1921                  * at i_block[] values, so not endian-sensitive here.
1922                  */
1923                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1924                     LINUX_S_ISLNK(inode->i_mode) &&
1925                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1926                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1927                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1928                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1929                         failed_csum = 0;
1930                 }
1931
1932                 if ((inode->i_flags & EXT4_ENCRYPT_FL) &&
1933                     add_encrypted_file(ctx, &pctx) < 0)
1934                         goto clear_inode;
1935
1936                 if (casefold_fs && inode->i_flags & EXT4_CASEFOLD_FL)
1937                         ext2fs_mark_inode_bitmap2(ctx->inode_casefold_map, ino);
1938
1939                 if (LINUX_S_ISDIR(inode->i_mode)) {
1940                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1941                         e2fsck_add_dir_info(ctx, ino, 0);
1942                         ctx->fs_directory_count++;
1943                         if (inode->i_flags & EXT4_CASEFOLD_FL)
1944                                 add_casefolded_dir(ctx, ino);
1945                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1946                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1947                         ctx->fs_regular_count++;
1948                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1949                            e2fsck_pass1_check_device_inode(fs, inode)) {
1950                         check_extents_inlinedata(ctx, &pctx);
1951                         check_immutable(ctx, &pctx);
1952                         check_size(ctx, &pctx);
1953                         ctx->fs_chardev_count++;
1954                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1955                            e2fsck_pass1_check_device_inode(fs, inode)) {
1956                         check_extents_inlinedata(ctx, &pctx);
1957                         check_immutable(ctx, &pctx);
1958                         check_size(ctx, &pctx);
1959                         ctx->fs_blockdev_count++;
1960                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1961                            e2fsck_pass1_check_symlink(fs, ino, inode,
1962                                                       block_buf)) {
1963                         check_immutable(ctx, &pctx);
1964                         ctx->fs_symlinks_count++;
1965                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1966                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1967                                 continue;
1968                         } else if (ext2fs_is_fast_symlink(inode)) {
1969                                 ctx->fs_fast_symlinks_count++;
1970                                 check_blocks(ctx, &pctx, block_buf,
1971                                              &ea_ibody_quota);
1972                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1973                                 continue;
1974                         }
1975                 }
1976                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1977                          e2fsck_pass1_check_device_inode(fs, inode)) {
1978                         check_extents_inlinedata(ctx, &pctx);
1979                         check_immutable(ctx, &pctx);
1980                         check_size(ctx, &pctx);
1981                         ctx->fs_fifo_count++;
1982                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1983                            e2fsck_pass1_check_device_inode(fs, inode)) {
1984                         check_extents_inlinedata(ctx, &pctx);
1985                         check_immutable(ctx, &pctx);
1986                         check_size(ctx, &pctx);
1987                         ctx->fs_sockets_count++;
1988                 } else
1989                         mark_inode_bad(ctx, ino);
1990                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1991                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1992                         if (inode->i_block[EXT2_IND_BLOCK])
1993                                 ctx->fs_ind_count++;
1994                         if (inode->i_block[EXT2_DIND_BLOCK])
1995                                 ctx->fs_dind_count++;
1996                         if (inode->i_block[EXT2_TIND_BLOCK])
1997                                 ctx->fs_tind_count++;
1998                 }
1999                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
2000                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
2001                     (inode->i_block[EXT2_IND_BLOCK] ||
2002                      inode->i_block[EXT2_DIND_BLOCK] ||
2003                      inode->i_block[EXT2_TIND_BLOCK] ||
2004                      ext2fs_file_acl_block(fs, inode))) {
2005                         struct process_inode_block *itp;
2006
2007                         itp = &inodes_to_process[process_inode_count];
2008                         itp->ino = ino;
2009                         itp->ea_ibody_quota = ea_ibody_quota;
2010                         if (inode_size < sizeof(struct ext2_inode_large))
2011                                 memcpy(&itp->inode, inode, inode_size);
2012                         else
2013                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
2014                         process_inode_count++;
2015                 } else
2016                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
2017
2018                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
2019
2020                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2021                         goto endit;
2022
2023                 if (process_inode_count >= ctx->process_inode_size) {
2024                         process_inodes(ctx, block_buf);
2025
2026                         if (e2fsck_should_abort(ctx))
2027                                 goto endit;
2028                 }
2029         }
2030         process_inodes(ctx, block_buf);
2031         ext2fs_close_inode_scan(scan);
2032         scan = NULL;
2033
2034         reserve_block_for_root_repair(ctx);
2035         reserve_block_for_lnf_repair(ctx);
2036
2037         /*
2038          * If any extended attribute blocks' reference counts need to
2039          * be adjusted, either up (ctx->refcount_extra), or down
2040          * (ctx->refcount), then fix them.
2041          */
2042         if (ctx->refcount) {
2043                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
2044                 ea_refcount_free(ctx->refcount);
2045                 ctx->refcount = 0;
2046         }
2047         if (ctx->refcount_extra) {
2048                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
2049                                         block_buf, +1);
2050                 ea_refcount_free(ctx->refcount_extra);
2051                 ctx->refcount_extra = 0;
2052         }
2053
2054         if (ctx->ea_block_quota_blocks) {
2055                 ea_refcount_free(ctx->ea_block_quota_blocks);
2056                 ctx->ea_block_quota_blocks = 0;
2057         }
2058
2059         if (ctx->ea_block_quota_inodes) {
2060                 ea_refcount_free(ctx->ea_block_quota_inodes);
2061                 ctx->ea_block_quota_inodes = 0;
2062         }
2063
2064         if (ctx->invalid_bitmaps)
2065                 handle_fs_bad_blocks(ctx);
2066
2067         /* We don't need the block_ea_map any more */
2068         if (ctx->block_ea_map) {
2069                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2070                 ctx->block_ea_map = 0;
2071         }
2072
2073         /* We don't need the encryption policy => ID map any more */
2074         destroy_encryption_policy_map(ctx);
2075
2076         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2077                 clear_problem_context(&pctx);
2078                 pctx.errcode = ext2fs_create_resize_inode(fs);
2079                 if (pctx.errcode) {
2080                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2081                                          &pctx)) {
2082                                 ctx->flags |= E2F_FLAG_ABORT;
2083                                 goto endit;
2084                         }
2085                         pctx.errcode = 0;
2086                 }
2087                 if (!pctx.errcode) {
2088                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2089                                           "recreate inode");
2090                         inode->i_mtime = ctx->now;
2091                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2092                                            "recreate inode");
2093                 }
2094                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2095         }
2096
2097         if (ctx->flags & E2F_FLAG_RESTART) {
2098                 /*
2099                  * Only the master copy of the superblock and block
2100                  * group descriptors are going to be written during a
2101                  * restart, so set the superblock to be used to be the
2102                  * master superblock.
2103                  */
2104                 ctx->use_superblock = 0;
2105                 goto endit;
2106         }
2107
2108         if (ctx->large_dirs && !ext2fs_has_feature_largedir(fs->super)) {
2109                 if (fix_problem(ctx, PR_2_FEATURE_LARGE_DIRS, &pctx)) {
2110                         ext2fs_set_feature_largedir(fs->super);
2111                         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
2112                         ext2fs_mark_super_dirty(fs);
2113                 }
2114                 if (fs->super->s_rev_level == EXT2_GOOD_OLD_REV &&
2115                     fix_problem(ctx, PR_1_FS_REV_LEVEL, &pctx)) {
2116                         ext2fs_update_dynamic_rev(fs);
2117                         ext2fs_mark_super_dirty(fs);
2118                 }
2119         }
2120
2121         if (ctx->block_dup_map) {
2122                 if (ctx->options & E2F_OPT_PREEN) {
2123                         clear_problem_context(&pctx);
2124                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2125                 }
2126                 e2fsck_pass1_dupblocks(ctx, block_buf);
2127         }
2128         ctx->flags |= E2F_FLAG_ALLOC_OK;
2129 endit:
2130         e2fsck_use_inode_shortcuts(ctx, 0);
2131         ext2fs_free_mem(&inodes_to_process);
2132         inodes_to_process = 0;
2133
2134         if (scan)
2135                 ext2fs_close_inode_scan(scan);
2136         if (block_buf)
2137                 ext2fs_free_mem(&block_buf);
2138         if (inode)
2139                 ext2fs_free_mem(&inode);
2140
2141         /*
2142          * The l+f inode may have been cleared, so zap it now and
2143          * later passes will recalculate it if necessary
2144          */
2145         ctx->lost_and_found = 0;
2146
2147         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2148                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2149         else
2150                 ctx->invalid_bitmaps++;
2151 }
2152
2153 #ifdef HAVE_PTHREAD
2154 static errcode_t e2fsck_pass1_copy_bitmap(ext2_filsys fs, ext2fs_generic_bitmap *src,
2155                                           ext2fs_generic_bitmap *dest)
2156 {
2157         errcode_t ret;
2158
2159         ret = ext2fs_copy_bitmap(*src, dest);
2160         if (ret)
2161                 return ret;
2162
2163         (*dest)->fs = fs;
2164         ext2fs_free_generic_bmap(*src);
2165         *src = NULL;
2166
2167         return 0;
2168 }
2169
2170
2171 static errcode_t e2fsck_pass1_copy_fs(ext2_filsys dest, e2fsck_t src_context,
2172                                       ext2_filsys src)
2173 {
2174         errcode_t       retval;
2175
2176         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2177         if (dest->dblist)
2178                 dest->dblist->fs = dest;
2179         if (src->block_map) {
2180                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2181                                                   &dest->block_map);
2182                 if (retval)
2183                         return retval;
2184         }
2185         if (src->inode_map) {
2186                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2187                                                   &dest->inode_map);
2188                 if (retval)
2189                         return retval;
2190         }
2191
2192         if (src->badblocks) {
2193                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2194                 if (retval)
2195                         return retval;
2196         }
2197
2198         /* disable it for now */
2199         src_context->openfs_flags &= ~EXT2_FLAG_EXCLUSIVE;
2200         retval = ext2fs_open_channel(dest, src_context->io_options,
2201                                      src_context->io_manager,
2202                                      src_context->openfs_flags,
2203                                      src->io->block_size);
2204         if (retval)
2205                 return retval;
2206
2207         /* Block size might not be default */
2208         io_channel_set_blksize(dest->io, src->io->block_size);
2209         ehandler_init(dest->io);
2210
2211         assert(dest->io->magic == src->io->magic);
2212         assert(dest->io->manager == src->io->manager);
2213         assert(strcmp(dest->io->name, src->io->name) == 0);
2214         assert(dest->io->block_size == src->io->block_size);
2215         assert(dest->io->read_error == src->io->read_error);
2216         assert(dest->io->write_error == src->io->write_error);
2217         assert(dest->io->refcount == src->io->refcount);
2218         assert(dest->io->flags == src->io->flags);
2219         assert(dest->io->app_data == dest);
2220         assert(src->io->app_data == src);
2221         assert(dest->io->align == src->io->align);
2222
2223         /* The data should be written to disk immediately */
2224         dest->io->flags |= CHANNEL_FLAGS_WRITETHROUGH;
2225         /* icache will be rebuilt if needed, so do not copy from @src */
2226         src->icache = NULL;
2227         return 0;
2228 }
2229
2230 static int e2fsck_pass1_merge_fs(ext2_filsys dest, ext2_filsys src)
2231 {
2232         struct ext2_inode_cache *icache = dest->icache;
2233         errcode_t retval = 0;
2234         io_channel dest_io;
2235         io_channel dest_image_io;
2236
2237         dest_io = dest->io;
2238         dest_image_io = dest->image_io;
2239
2240         memcpy(dest, src, sizeof(struct struct_ext2_filsys));
2241         dest->io = dest_io;
2242         dest->image_io = dest_image_io;
2243         dest->icache = icache;
2244         if (dest->dblist)
2245                 dest->dblist->fs = dest;
2246         if (src->inode_map) {
2247                 retval = e2fsck_pass1_copy_bitmap(dest, &src->inode_map,
2248                                                   &dest->inode_map);
2249                 if (retval)
2250                         return retval;
2251         }
2252         if (src->block_map) {
2253                 retval = e2fsck_pass1_copy_bitmap(dest, &src->block_map,
2254                                                   &dest->block_map);
2255                 if (retval)
2256                         return retval;
2257         }
2258
2259         if (src->icache) {
2260                 ext2fs_free_inode_cache(src->icache);
2261                 src->icache = NULL;
2262         }
2263
2264         if (src->badblocks) {
2265                 retval = ext2fs_badblocks_copy(src->badblocks, &dest->badblocks);
2266
2267                 ext2fs_badblocks_list_free(src->badblocks);
2268                 src->badblocks = NULL;
2269         }
2270
2271         io_channel_close(src->io);
2272         return retval;
2273 }
2274
2275 static errcode_t e2fsck_pass1_thread_prepare(e2fsck_t global_ctx, e2fsck_t *thread_ctx,
2276                                              int thread_index, int num_threads)
2277 {
2278         errcode_t               retval;
2279         e2fsck_t                thread_context;
2280         ext2_filsys             thread_fs;
2281         ext2_filsys             global_fs = global_ctx->fs;
2282         struct e2fsck_thread    *tinfo;
2283         dgrp_t                  average_group;
2284
2285         assert(global_ctx->inode_used_map == NULL);
2286         assert(global_ctx->inode_dir_map == NULL);
2287         assert(global_ctx->inode_bb_map == NULL);
2288         assert(global_ctx->inode_imagic_map == NULL);
2289         assert(global_ctx->inode_reg_map == NULL);
2290         assert(global_ctx->inodes_to_rebuild == NULL);
2291
2292         assert(global_ctx->block_found_map == NULL);
2293         assert(global_ctx->block_dup_map == NULL);
2294         assert(global_ctx->block_ea_map == NULL);
2295         assert(global_ctx->block_metadata_map == NULL);
2296         assert(global_ctx->fs->dblist == NULL);
2297
2298         retval = ext2fs_get_mem(sizeof(struct e2fsck_struct), &thread_context);
2299         if (retval) {
2300                 com_err(global_ctx->program_name, retval, "while allocating memory");
2301                 return retval;
2302         }
2303         memcpy(thread_context, global_ctx, sizeof(struct e2fsck_struct));
2304         thread_context->global_ctx = global_ctx;
2305
2306         retval = ext2fs_get_mem(sizeof(struct struct_ext2_filsys), &thread_fs);
2307         if (retval) {
2308                 com_err(global_ctx->program_name, retval, "while allocating memory");
2309                 goto out_context;
2310         }
2311
2312         io_channel_flush_cleanup(global_fs->io);
2313         retval = e2fsck_pass1_copy_fs(thread_fs, global_ctx, global_fs);
2314         if (retval) {
2315                 com_err(global_ctx->program_name, retval, "while copying fs");
2316                 goto out_fs;
2317         }
2318         thread_fs->priv_data = thread_context;
2319
2320         thread_context->thread_info.et_thread_index = thread_index;
2321         set_up_logging(thread_context);
2322
2323         /*
2324          * Distribute work to multiple threads:
2325          * Each thread work on fs->group_desc_count / nthread groups.
2326          */
2327         tinfo = &thread_context->thread_info;
2328         average_group = thread_fs->group_desc_count / num_threads;
2329         if (average_group == 0)
2330                 average_group = 1;
2331         tinfo->et_group_start = average_group * thread_index;
2332         if (thread_index == num_threads - 1)
2333                 tinfo->et_group_end = thread_fs->group_desc_count;
2334         else
2335                 tinfo->et_group_end = average_group * (thread_index + 1);
2336         tinfo->et_group_next = tinfo->et_group_start;
2337         tinfo->et_inode_number = 0;
2338         tinfo->et_log_buf[0] = '\0';
2339         tinfo->et_log_length = 0;
2340         if (thread_context->options & E2F_OPT_MULTITHREAD)
2341                 log_out(thread_context, _("Scan group range [%d, %d)\n"),
2342                         tinfo->et_group_start, tinfo->et_group_end);
2343         thread_context->fs = thread_fs;
2344         *thread_ctx = thread_context;
2345         return 0;
2346 out_fs:
2347         ext2fs_free_mem(&thread_fs);
2348 out_context:
2349         ext2fs_free_mem(&thread_context);
2350         return retval;
2351 }
2352
2353 static int e2fsck_pass1_thread_join_one(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2354 {
2355         errcode_t        retval;
2356         int              flags = global_ctx->flags;
2357         ext2_filsys      thread_fs = thread_ctx->fs;
2358         ext2_filsys      global_fs = global_ctx->fs;
2359         FILE            *global_logf = global_ctx->logf;
2360         FILE            *global_problem_logf = global_ctx->problem_logf;
2361 #ifdef HAVE_SETJMP_H
2362         jmp_buf          old_jmp;
2363
2364         memcpy(old_jmp, global_ctx->abort_loc, sizeof(jmp_buf));
2365 #endif
2366         memcpy(global_ctx, thread_ctx, sizeof(struct e2fsck_struct));
2367 #ifdef HAVE_SETJMP_H
2368         memcpy(global_ctx->abort_loc, old_jmp, sizeof(jmp_buf));
2369 #endif
2370         /* Keep the global singal flags*/
2371         global_ctx->flags |= (flags & E2F_FLAG_SIGNAL_MASK) |
2372                              (global_ctx->flags & E2F_FLAG_SIGNAL_MASK);
2373
2374         retval = e2fsck_pass1_merge_fs(global_fs, thread_fs);
2375         if (retval) {
2376                 com_err(global_ctx->program_name, 0, _("while merging fs\n"));
2377                 return retval;
2378         }
2379         global_fs->priv_data = global_ctx;
2380         global_ctx->fs = global_fs;
2381         global_ctx->logf = global_logf;
2382         global_ctx->problem_logf = global_problem_logf;
2383         global_ctx->global_ctx = NULL;
2384
2385         if (thread_ctx->inode_used_map) {
2386                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2387                                         &thread_ctx->inode_used_map,
2388                                         &global_ctx->inode_used_map);
2389                 if (retval)
2390                         return retval;
2391         }
2392         if (thread_ctx->inode_bad_map) {
2393                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2394                                         &thread_ctx->inode_bad_map,
2395                                         &global_ctx->inode_bad_map);
2396                 if (retval)
2397                         return retval;
2398         }
2399         if (thread_ctx->inode_dir_map) {
2400                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2401                                         &thread_ctx->inode_dir_map,
2402                                         &global_ctx->inode_dir_map);
2403                 if (retval)
2404                         return retval;
2405         }
2406         if (thread_ctx->inode_bb_map) {
2407                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2408                                         &thread_ctx->inode_bb_map,
2409                                         &global_ctx->inode_bb_map);
2410                 if (retval)
2411                         return retval;
2412         }
2413         if (thread_ctx->inode_imagic_map) {
2414                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2415                                         &thread_ctx->inode_imagic_map,
2416                                         &global_ctx->inode_imagic_map);
2417                 if (retval)
2418                         return retval;
2419         }
2420         if (thread_ctx->inode_reg_map) {
2421                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2422                                         &thread_ctx->inode_reg_map,
2423                                         &global_ctx->inode_reg_map);
2424                 if (retval)
2425                         return retval;
2426         }
2427         if (thread_ctx->inodes_to_rebuild) {
2428                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2429                                         &thread_ctx->inodes_to_rebuild,
2430                                         &global_ctx->inodes_to_rebuild);
2431                 if (retval)
2432                         return retval;
2433         }
2434         if (thread_ctx->block_found_map) {
2435                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2436                                         &thread_ctx->block_found_map,
2437                                         &global_ctx->block_found_map);
2438                 if (retval)
2439                         return retval;
2440         }
2441         if (thread_ctx->block_dup_map) {
2442                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2443                                         &thread_ctx->block_dup_map,
2444                                         &global_ctx->block_dup_map);
2445                 if (retval)
2446                         return retval;
2447         }
2448         if (thread_ctx->block_ea_map) {
2449                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2450                                         &thread_ctx->block_ea_map,
2451                                         &global_ctx->block_ea_map);
2452                 if (retval)
2453                         return retval;
2454         }
2455         if (thread_ctx->block_metadata_map) {
2456                 retval = e2fsck_pass1_copy_bitmap(global_fs,
2457                                         &thread_ctx->block_metadata_map,
2458                                         &global_ctx->block_metadata_map);
2459                 if (retval)
2460                         return retval;
2461         }
2462
2463         return 0;
2464 }
2465
2466 static int e2fsck_pass1_thread_join(e2fsck_t global_ctx, e2fsck_t thread_ctx)
2467 {
2468         errcode_t       retval;
2469
2470         retval = e2fsck_pass1_thread_join_one(global_ctx, thread_ctx);
2471         ext2fs_free_mem(&thread_ctx->fs);
2472         if (thread_ctx->logf)
2473                 fclose(thread_ctx->logf);
2474         if (thread_ctx->problem_logf) {
2475                 fputs("</problem_log>\n", thread_ctx->problem_logf);
2476                 fclose(thread_ctx->problem_logf);
2477         }
2478         ext2fs_free_mem(&thread_ctx);
2479
2480         return retval;
2481 }
2482
2483 static int e2fsck_pass1_threads_join(struct e2fsck_thread_info *infos,
2484                                       int num_threads, e2fsck_t global_ctx)
2485 {
2486         errcode_t                        rc;
2487         errcode_t                        ret = 0;
2488         int                              i;
2489         struct e2fsck_thread_info       *pinfo;
2490
2491         for (i = 0; i < num_threads; i++) {
2492                 pinfo = &infos[i];
2493
2494                 if (!pinfo->eti_started)
2495                         continue;
2496
2497                 rc = pthread_join(pinfo->eti_thread_id, NULL);
2498                 if (rc) {
2499                         com_err(global_ctx->program_name, rc,
2500                                 _("while joining thread\n"));
2501                         if (ret == 0)
2502                                 ret = rc;
2503                 }
2504                 e2fsck_pass1_thread_join(global_ctx, infos[i].eti_thread_ctx);
2505         }
2506         free(infos);
2507
2508         return ret;
2509 }
2510
2511 static void *e2fsck_pass1_thread(void *arg)
2512 {
2513         struct e2fsck_thread_info       *info = arg;
2514         e2fsck_t                         thread_ctx = info->eti_thread_ctx;
2515
2516 #ifdef HAVE_SETJMP_H
2517         /*
2518          * When fatal_error() happens, jump to here. The thread
2519          * context's flags will be saved, but its abort_loc will
2520          * be overwritten by original jump buffer for the later
2521          * tests.
2522          */
2523         if (setjmp(thread_ctx->abort_loc)) {
2524                 thread_ctx->flags &= ~E2F_FLAG_SETJMP_OK;
2525                 goto out;
2526         }
2527         thread_ctx->flags |= E2F_FLAG_SETJMP_OK;
2528 #endif
2529
2530         e2fsck_pass1_run(thread_ctx);
2531
2532 out:
2533         if (thread_ctx->options & E2F_OPT_MULTITHREAD)
2534                 log_out(thread_ctx,
2535                         _("Scanned group range [%lu, %lu), inodes %lu\n"),
2536                         thread_ctx->thread_info.et_group_start,
2537                         thread_ctx->thread_info.et_group_end,
2538                         thread_ctx->thread_info.et_inode_number);
2539         return NULL;
2540 }
2541
2542 static int e2fsck_pass1_threads_start(struct e2fsck_thread_info **pinfo,
2543                                       int num_threads, e2fsck_t global_ctx)
2544 {
2545         struct e2fsck_thread_info       *infos;
2546         pthread_attr_t                   attr;
2547         errcode_t                        retval;
2548         errcode_t                        ret;
2549         struct e2fsck_thread_info       *tmp_pinfo;
2550         int                              i;
2551         e2fsck_t                         thread_ctx;
2552
2553         retval = pthread_attr_init(&attr);
2554         if (retval) {
2555                 com_err(global_ctx->program_name, retval,
2556                         _("while setting pthread attribute\n"));
2557                 return retval;
2558         }
2559
2560         infos = calloc(num_threads, sizeof(struct e2fsck_thread_info));
2561         if (infos == NULL) {
2562                 retval = -ENOMEM;
2563                 com_err(global_ctx->program_name, retval,
2564                         _("while allocating memory for threads\n"));
2565                 pthread_attr_destroy(&attr);
2566                 return retval;
2567         }
2568
2569         for (i = 0; i < num_threads; i++) {
2570                 tmp_pinfo = &infos[i];
2571                 tmp_pinfo->eti_thread_index = i;
2572                 retval = e2fsck_pass1_thread_prepare(global_ctx, &thread_ctx,
2573                                                      i, num_threads);
2574                 if (retval) {
2575                         com_err(global_ctx->program_name, retval,
2576                                 _("while preparing pass1 thread\n"));
2577                         break;
2578                 }
2579                 tmp_pinfo->eti_thread_ctx = thread_ctx;
2580
2581                 retval = pthread_create(&tmp_pinfo->eti_thread_id, &attr,
2582                                         &e2fsck_pass1_thread, tmp_pinfo);
2583                 if (retval) {
2584                         com_err(global_ctx->program_name, retval,
2585                                 _("while creating thread\n"));
2586                         e2fsck_pass1_thread_join(global_ctx, thread_ctx);
2587                         break;
2588                 }
2589
2590                 tmp_pinfo->eti_started = 1;
2591         }
2592
2593         /* destroy the thread attribute object, since it is no longer needed */
2594         ret = pthread_attr_destroy(&attr);
2595         if (ret) {
2596                 com_err(global_ctx->program_name, ret,
2597                         _("while destroying thread attribute\n"));
2598                 if (retval == 0)
2599                         retval = ret;
2600         }
2601
2602         if (retval) {
2603                 e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2604                 return retval;
2605         }
2606         *pinfo = infos;
2607         return 0;
2608 }
2609
2610 static void e2fsck_pass1_multithread(e2fsck_t global_ctx)
2611 {
2612         struct e2fsck_thread_info       *infos = NULL;
2613         int                              num_threads = 1;
2614         errcode_t                        retval;
2615
2616         retval = e2fsck_pass1_threads_start(&infos, num_threads, global_ctx);
2617         if (retval) {
2618                 com_err(global_ctx->program_name, retval,
2619                         _("while starting pass1 threads\n"));
2620                 goto out_abort;
2621         }
2622
2623         retval = e2fsck_pass1_threads_join(infos, num_threads, global_ctx);
2624         if (retval) {
2625                 com_err(global_ctx->program_name, retval,
2626                         _("while joining pass1 threads\n"));
2627                 goto out_abort;
2628         }
2629         return;
2630 out_abort:
2631         global_ctx->flags |= E2F_FLAG_ABORT;
2632         return;
2633 }
2634 #endif
2635
2636 void e2fsck_pass1(e2fsck_t ctx)
2637 {
2638
2639 #ifdef HAVE_PTHREAD
2640         e2fsck_pass1_multithread(ctx);
2641 #else
2642         e2fsck_pass1_run(ctx);
2643 #endif
2644 }
2645
2646 #undef FINISH_INODE_LOOP
2647
2648 /*
2649  * When the inode_scan routines call this callback at the end of the
2650  * glock group, call process_inodes.
2651  */
2652 static errcode_t scan_callback(ext2_filsys fs,
2653                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2654                                dgrp_t group, void * priv_data)
2655 {
2656         struct scan_callback_struct *scan_struct;
2657         e2fsck_t ctx;
2658         struct e2fsck_thread *tinfo;
2659
2660         scan_struct = (struct scan_callback_struct *) priv_data;
2661         ctx = scan_struct->ctx;
2662
2663         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2664
2665         if (ctx->progress)
2666                 if ((ctx->progress)(ctx, 1, group+1,
2667                                     ctx->fs->group_desc_count))
2668                         return EXT2_ET_CANCEL_REQUESTED;
2669
2670 #ifdef HAVE_PTHREAD
2671         if (ctx->global_ctx) {
2672                 tinfo = &ctx->thread_info;
2673                 tinfo->et_group_next++;
2674                 if (ctx->options & E2F_OPT_DEBUG &&
2675                     ctx->options & E2F_OPT_MULTITHREAD)
2676                         log_out(ctx, _("group %d finished\n"),
2677                                 tinfo->et_group_next);
2678                 if (tinfo->et_group_next >= tinfo->et_group_end)
2679                         return EXT2_ET_SCAN_FINISHED;
2680         }
2681 #endif
2682
2683         return 0;
2684 }
2685
2686 /*
2687  * Process the inodes in the "inodes to process" list.
2688  */
2689 static void process_inodes(e2fsck_t ctx, char *block_buf)
2690 {
2691         int                     i;
2692         struct ext2_inode       *old_stashed_inode;
2693         ext2_ino_t              old_stashed_ino;
2694         const char              *old_operation;
2695         char                    buf[80];
2696         struct problem_context  pctx;
2697
2698 #if 0
2699         printf("begin process_inodes: ");
2700 #endif
2701         if (process_inode_count == 0)
2702                 return;
2703         old_operation = ehandler_operation(0);
2704         old_stashed_inode = ctx->stashed_inode;
2705         old_stashed_ino = ctx->stashed_ino;
2706         qsort(inodes_to_process, process_inode_count,
2707                       sizeof(struct process_inode_block), process_inode_cmp);
2708         clear_problem_context(&pctx);
2709         for (i=0; i < process_inode_count; i++) {
2710                 pctx.inode = ctx->stashed_inode =
2711                         (struct ext2_inode *) &inodes_to_process[i].inode;
2712                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2713
2714 #if 0
2715                 printf("%u ", pctx.ino);
2716 #endif
2717                 sprintf(buf, _("reading indirect blocks of inode %u"),
2718                         pctx.ino);
2719                 ehandler_operation(buf);
2720                 check_blocks(ctx, &pctx, block_buf,
2721                              &inodes_to_process[i].ea_ibody_quota);
2722                 if (e2fsck_should_abort(ctx))
2723                         break;
2724         }
2725         ctx->stashed_inode = old_stashed_inode;
2726         ctx->stashed_ino = old_stashed_ino;
2727         process_inode_count = 0;
2728 #if 0
2729         printf("end process inodes\n");
2730 #endif
2731         ehandler_operation(old_operation);
2732 }
2733
2734 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2735 {
2736         const struct process_inode_block *ib_a =
2737                 (const struct process_inode_block *) a;
2738         const struct process_inode_block *ib_b =
2739                 (const struct process_inode_block *) b;
2740         int     ret;
2741
2742         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2743                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2744         if (ret == 0)
2745                 /*
2746                  * We only call process_inodes() for non-extent
2747                  * inodes, so it's OK to pass NULL to
2748                  * ext2fs_file_acl_block() here.
2749                  */
2750                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2751                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2752         if (ret == 0)
2753                 ret = ib_a->ino - ib_b->ino;
2754         return ret;
2755 }
2756
2757 /*
2758  * Mark an inode as being bad in some what
2759  */
2760 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2761 {
2762         struct          problem_context pctx;
2763
2764         if (!ctx->inode_bad_map) {
2765                 clear_problem_context(&pctx);
2766
2767                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2768                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2769                                 "inode_bad_map", &ctx->inode_bad_map);
2770                 if (pctx.errcode) {
2771                         pctx.num = 3;
2772                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2773                         /* Should never get here */
2774                         ctx->flags |= E2F_FLAG_ABORT;
2775                         return;
2776                 }
2777         }
2778         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2779 }
2780
2781 static void add_casefolded_dir(e2fsck_t ctx, ino_t ino)
2782 {
2783         struct          problem_context pctx;
2784
2785         if (!ctx->casefolded_dirs) {
2786                 pctx.errcode = ext2fs_u32_list_create(&ctx->casefolded_dirs, 0);
2787                 if (pctx.errcode)
2788                         goto error;
2789         }
2790         pctx.errcode = ext2fs_u32_list_add(ctx->casefolded_dirs, ino);
2791         if (pctx.errcode == 0)
2792                 return;
2793 error:
2794         fix_problem(ctx, PR_1_ALLOCATE_CASEFOLDED_DIRLIST, &pctx);
2795         /* Should never get here */
2796         ctx->flags |= E2F_FLAG_ABORT;
2797 }
2798
2799 /*
2800  * This procedure will allocate the inode "bb" (badblock) map table
2801  */
2802 static void alloc_bb_map(e2fsck_t ctx)
2803 {
2804         struct          problem_context pctx;
2805
2806         clear_problem_context(&pctx);
2807         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2808                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2809                         "inode_bb_map", &ctx->inode_bb_map);
2810         if (pctx.errcode) {
2811                 pctx.num = 4;
2812                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2813                 /* Should never get here */
2814                 ctx->flags |= E2F_FLAG_ABORT;
2815                 return;
2816         }
2817 }
2818
2819 /*
2820  * This procedure will allocate the inode imagic table
2821  */
2822 static void alloc_imagic_map(e2fsck_t ctx)
2823 {
2824         struct          problem_context pctx;
2825
2826         clear_problem_context(&pctx);
2827         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2828                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2829                         "inode_imagic_map", &ctx->inode_imagic_map);
2830         if (pctx.errcode) {
2831                 pctx.num = 5;
2832                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2833                 /* Should never get here */
2834                 ctx->flags |= E2F_FLAG_ABORT;
2835                 return;
2836         }
2837 }
2838
2839 /*
2840  * Marks a block as in use, setting the dup_map if it's been set
2841  * already.  Called by process_block and process_bad_block.
2842  *
2843  * WARNING: Assumes checks have already been done to make sure block
2844  * is valid.  This is true in both process_block and process_bad_block.
2845  */
2846 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2847 {
2848         struct          problem_context pctx;
2849
2850         clear_problem_context(&pctx);
2851
2852         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2853                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2854                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2855                         return;
2856                 }
2857                 if (!ctx->block_dup_map) {
2858                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2859                                         _("multiply claimed block map"),
2860                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2861                                         &ctx->block_dup_map);
2862                         if (pctx.errcode) {
2863                                 pctx.num = 3;
2864                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2865                                             &pctx);
2866                                 /* Should never get here */
2867                                 ctx->flags |= E2F_FLAG_ABORT;
2868                                 return;
2869                         }
2870                 }
2871                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2872         } else {
2873                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2874         }
2875 }
2876
2877 /*
2878  * When cluster size is greater than one block, it is caller's responsibility
2879  * to make sure block parameter starts at a cluster boundary.
2880  */
2881 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2882                                       unsigned int num)
2883 {
2884         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2885                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2886         else {
2887                 unsigned int i;
2888
2889                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2890                         mark_block_used(ctx, block + i);
2891         }
2892 }
2893
2894 /*
2895  * Adjust the extended attribute block's reference counts at the end
2896  * of pass 1, either by subtracting out references for EA blocks that
2897  * are still referenced in ctx->refcount, or by adding references for
2898  * EA blocks that had extra references as accounted for in
2899  * ctx->refcount_extra.
2900  */
2901 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2902                                     char *block_buf, int adjust_sign)
2903 {
2904         struct ext2_ext_attr_header     *header;
2905         struct problem_context          pctx;
2906         ext2_filsys                     fs = ctx->fs;
2907         blk64_t                         blk;
2908         __u32                           should_be;
2909         ea_value_t                      count;
2910
2911         clear_problem_context(&pctx);
2912
2913         ea_refcount_intr_begin(refcount);
2914         while (1) {
2915                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2916                         break;
2917                 pctx.blk = blk;
2918                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2919                                                      pctx.ino);
2920                 if (pctx.errcode) {
2921                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2922                         return;
2923                 }
2924                 header = (struct ext2_ext_attr_header *) block_buf;
2925                 pctx.blkcount = header->h_refcount;
2926                 should_be = header->h_refcount + adjust_sign * (int)count;
2927                 pctx.num = should_be;
2928                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2929                         header->h_refcount = should_be;
2930                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2931                                                              block_buf,
2932                                                              pctx.ino);
2933                         if (pctx.errcode) {
2934                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2935                                             &pctx);
2936                                 continue;
2937                         }
2938                 }
2939         }
2940 }
2941
2942 /*
2943  * Handle processing the extended attribute blocks
2944  */
2945 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2946                            char *block_buf, struct ea_quota *ea_block_quota)
2947 {
2948         ext2_filsys fs = ctx->fs;
2949         ext2_ino_t      ino = pctx->ino;
2950         struct ext2_inode *inode = pctx->inode;
2951         blk64_t         blk;
2952         char *          end;
2953         struct ext2_ext_attr_header *header;
2954         struct ext2_ext_attr_entry *first, *entry;
2955         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2956         __u64           quota_inodes = 0;
2957         region_t        region = 0;
2958         int             failed_csum = 0;
2959
2960         ea_block_quota->blocks = 0;
2961         ea_block_quota->inodes = 0;
2962
2963         blk = ext2fs_file_acl_block(fs, inode);
2964         if (blk == 0)
2965                 return 0;
2966
2967         /*
2968          * If the Extended attribute flag isn't set, then a non-zero
2969          * file acl means that the inode is corrupted.
2970          *
2971          * Or if the extended attribute block is an invalid block,
2972          * then the inode is also corrupted.
2973          */
2974         if (!ext2fs_has_feature_xattr(fs->super) ||
2975             (blk < fs->super->s_first_data_block) ||
2976             (blk >= ext2fs_blocks_count(fs->super))) {
2977                 mark_inode_bad(ctx, ino);
2978                 return 0;
2979         }
2980
2981         /* If ea bitmap hasn't been allocated, create it */
2982         if (!ctx->block_ea_map) {
2983                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2984                                         _("ext attr block map"),
2985                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2986                                         &ctx->block_ea_map);
2987                 if (pctx->errcode) {
2988                         pctx->num = 2;
2989                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2990                         ctx->flags |= E2F_FLAG_ABORT;
2991                         return 0;
2992                 }
2993         }
2994
2995         /* Create the EA refcount structure if necessary */
2996         if (!ctx->refcount) {
2997                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2998                 if (pctx->errcode) {
2999                         pctx->num = 1;
3000                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3001                         ctx->flags |= E2F_FLAG_ABORT;
3002                         return 0;
3003                 }
3004         }
3005
3006 #if 0
3007         /* Debugging text */
3008         printf("Inode %u has EA block %u\n", ino, blk);
3009 #endif
3010
3011         /* Have we seen this EA block before? */
3012         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
3013                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
3014                 ea_block_quota->inodes = 0;
3015
3016                 if (ctx->ea_block_quota_blocks) {
3017                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
3018                                           &quota_blocks);
3019                         if (quota_blocks)
3020                                 ea_block_quota->blocks = quota_blocks;
3021                 }
3022
3023                 if (ctx->ea_block_quota_inodes)
3024                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
3025                                           &ea_block_quota->inodes);
3026
3027                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
3028                         return 1;
3029                 /* Ooops, this EA was referenced more than it stated */
3030                 if (!ctx->refcount_extra) {
3031                         pctx->errcode = ea_refcount_create(0,
3032                                            &ctx->refcount_extra);
3033                         if (pctx->errcode) {
3034                                 pctx->num = 2;
3035                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3036                                 ctx->flags |= E2F_FLAG_ABORT;
3037                                 return 0;
3038                         }
3039                 }
3040                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
3041                 return 1;
3042         }
3043
3044         /*
3045          * OK, we haven't seen this EA block yet.  So we need to
3046          * validate it
3047          */
3048         pctx->blk = blk;
3049         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
3050         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
3051                 pctx->errcode = 0;
3052                 failed_csum = 1;
3053         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
3054                 pctx->errcode = 0;
3055
3056         if (pctx->errcode &&
3057             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
3058                 pctx->errcode = 0;
3059                 goto clear_extattr;
3060         }
3061         header = (struct ext2_ext_attr_header *) block_buf;
3062         pctx->blk = ext2fs_file_acl_block(fs, inode);
3063         if (((ctx->ext_attr_ver == 1) &&
3064              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
3065             ((ctx->ext_attr_ver == 2) &&
3066              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
3067                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
3068                         goto clear_extattr;
3069         }
3070
3071         if (header->h_blocks != 1) {
3072                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
3073                         goto clear_extattr;
3074         }
3075
3076         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
3077                 goto clear_extattr;
3078
3079         region = region_create(0, fs->blocksize);
3080         if (!region) {
3081                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
3082                 ctx->flags |= E2F_FLAG_ABORT;
3083                 return 0;
3084         }
3085         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
3086                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3087                         goto clear_extattr;
3088         }
3089
3090         first = (struct ext2_ext_attr_entry *)(header+1);
3091         end = block_buf + fs->blocksize;
3092         entry = first;
3093         while ((char *)entry < end && *(__u32 *)entry) {
3094                 __u32 hash;
3095
3096                 if (region_allocate(region, (char *)entry - (char *)header,
3097                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
3098                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3099                                 goto clear_extattr;
3100                         break;
3101                 }
3102                 if ((ctx->ext_attr_ver == 1 &&
3103                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
3104                     (ctx->ext_attr_ver == 2 &&
3105                      entry->e_name_index == 0)) {
3106                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
3107                                 goto clear_extattr;
3108                         break;
3109                 }
3110                 if (entry->e_value_inum == 0) {
3111                         if (entry->e_value_size > EXT2_XATTR_SIZE_MAX ||
3112                             (entry->e_value_offs + entry->e_value_size >
3113                              fs->blocksize)) {
3114                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
3115                                         goto clear_extattr;
3116                                 break;
3117                         }
3118                         if (entry->e_value_size &&
3119                             region_allocate(region, entry->e_value_offs,
3120                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
3121                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
3122                                                 pctx))
3123                                         goto clear_extattr;
3124                         }
3125
3126                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
3127                                                           entry->e_value_offs);
3128
3129                         if (entry->e_hash != hash) {
3130                                 pctx->num = entry->e_hash;
3131                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
3132                                         goto clear_extattr;
3133                                 entry->e_hash = hash;
3134                         }
3135                 } else {
3136                         problem_t problem;
3137                         blk64_t entry_quota_blocks;
3138
3139                         problem = check_large_ea_inode(ctx, entry, pctx,
3140                                                        &entry_quota_blocks);
3141                         if (problem && fix_problem(ctx, problem, pctx))
3142                                 goto clear_extattr;
3143
3144                         quota_blocks += entry_quota_blocks;
3145                         quota_inodes++;
3146                 }
3147
3148                 entry = EXT2_EXT_ATTR_NEXT(entry);
3149         }
3150         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
3151                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
3152                         goto clear_extattr;
3153         }
3154         region_free(region);
3155
3156         /*
3157          * We only get here if there was no other errors that were fixed.
3158          * If there was a checksum fail, ask to correct it.
3159          */
3160         if (failed_csum &&
3161             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
3162                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
3163                                                        pctx->ino);
3164                 if (pctx->errcode)
3165                         return 0;
3166         }
3167
3168         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
3169                 if (!ctx->ea_block_quota_blocks) {
3170                         pctx->errcode = ea_refcount_create(0,
3171                                                 &ctx->ea_block_quota_blocks);
3172                         if (pctx->errcode) {
3173                                 pctx->num = 3;
3174                                 goto refcount_fail;
3175                         }
3176                 }
3177                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
3178                                   quota_blocks);
3179         }
3180
3181         if (quota_inodes) {
3182                 if (!ctx->ea_block_quota_inodes) {
3183                         pctx->errcode = ea_refcount_create(0,
3184                                                 &ctx->ea_block_quota_inodes);
3185                         if (pctx->errcode) {
3186                                 pctx->num = 4;
3187 refcount_fail:
3188                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
3189                                 ctx->flags |= E2F_FLAG_ABORT;
3190                                 return 0;
3191                         }
3192                 }
3193
3194                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
3195                                   quota_inodes);
3196         }
3197         ea_block_quota->blocks = quota_blocks;
3198         ea_block_quota->inodes = quota_inodes;
3199
3200         inc_ea_inode_refs(ctx, pctx, first, end);
3201         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
3202         mark_block_used(ctx, blk);
3203         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
3204         return 1;
3205
3206 clear_extattr:
3207         if (region)
3208                 region_free(region);
3209         ext2fs_file_acl_block_set(fs, inode, 0);
3210         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
3211         return 0;
3212 }
3213
3214 /* Returns 1 if bad htree, 0 if OK */
3215 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
3216                         ext2_ino_t ino, struct ext2_inode *inode,
3217                         char *block_buf)
3218 {
3219         struct ext2_dx_root_info        *root;
3220         ext2_filsys                     fs = ctx->fs;
3221         errcode_t                       retval;
3222         blk64_t                         blk;
3223
3224         if ((!LINUX_S_ISDIR(inode->i_mode) &&
3225              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
3226             (!ext2fs_has_feature_dir_index(fs->super) &&
3227              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
3228                 return 1;
3229
3230         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
3231
3232         if ((pctx->errcode) ||
3233             (blk == 0) ||
3234             (blk < fs->super->s_first_data_block) ||
3235             (blk >= ext2fs_blocks_count(fs->super))) {
3236                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3237                         return 1;
3238                 else
3239                         return 0;
3240         }
3241
3242         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
3243         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3244                 return 1;
3245
3246         /* XXX should check that beginning matches a directory */
3247         root = (struct ext2_dx_root_info *) (block_buf + 24);
3248
3249         if ((root->reserved_zero || root->info_length < 8) &&
3250             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
3251                 return 1;
3252
3253         pctx->num = root->hash_version;
3254         if ((root->hash_version != EXT2_HASH_LEGACY) &&
3255             (root->hash_version != EXT2_HASH_HALF_MD4) &&
3256             (root->hash_version != EXT2_HASH_TEA) &&
3257             (root->hash_version != EXT2_HASH_SIPHASH) &&
3258             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
3259                 return 1;
3260
3261         if (ext4_hash_in_dirent(inode)) {
3262                 if (root->hash_version != EXT2_HASH_SIPHASH &&
3263                     fix_problem(ctx, PR_1_HTREE_NEEDS_SIPHASH, pctx))
3264                         return 1;
3265         } else {
3266                 if (root->hash_version == EXT2_HASH_SIPHASH &&
3267                    fix_problem(ctx, PR_1_HTREE_CANNOT_SIPHASH, pctx))
3268                         return 1;
3269         }
3270
3271         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
3272             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
3273                 return 1;
3274
3275         pctx->num = root->indirect_levels;
3276         /* if htree level is clearly too high, consider it to be broken */
3277         if (root->indirect_levels > EXT4_HTREE_LEVEL &&
3278             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3279                 return 1;
3280
3281         /* if level is only maybe too high, LARGE_DIR feature could be unset */
3282         if (root->indirect_levels > ext2_dir_htree_level(fs) &&
3283             !ext2fs_has_feature_largedir(fs->super)) {
3284                 int blockbits = EXT2_BLOCK_SIZE_BITS(fs->super) + 10;
3285                 unsigned idx_pb = 1 << (blockbits - 3);
3286
3287                 /* compare inode size/blocks vs. max-sized 2-level htree */
3288                 if (EXT2_I_SIZE(pctx->inode) <
3289                     (idx_pb - 1) * (idx_pb - 2) << blockbits &&
3290                     pctx->inode->i_blocks <
3291                     (idx_pb - 1) * (idx_pb - 2) << (blockbits - 9) &&
3292                     fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
3293                         return 1;
3294         }
3295
3296         if (root->indirect_levels > EXT4_HTREE_LEVEL_COMPAT ||
3297             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3298                 ctx->large_dirs++;
3299
3300         return 0;
3301 }
3302
3303 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
3304                         struct ext2_inode *inode, int restart_flag,
3305                         const char *source)
3306 {
3307         inode->i_flags = 0;
3308         inode->i_links_count = 0;
3309         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
3310         inode->i_dtime = ctx->now;
3311
3312         /*
3313          * If a special inode has such rotten block mappings that we
3314          * want to clear the whole inode, be sure to actually zap
3315          * the block maps because i_links_count isn't checked for
3316          * special inodes, and we'll end up right back here the next
3317          * time we run fsck.
3318          */
3319         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
3320                 memset(inode->i_block, 0, sizeof(inode->i_block));
3321
3322         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
3323         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
3324         if (ctx->inode_reg_map)
3325                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
3326         if (ctx->inode_bad_map)
3327                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
3328
3329         /*
3330          * If the inode was partially accounted for before processing
3331          * was aborted, we need to restart the pass 1 scan.
3332          */
3333         ctx->flags |= restart_flag;
3334
3335         if (ino == EXT2_BAD_INO)
3336                 memset(inode, 0, sizeof(struct ext2_inode));
3337
3338         e2fsck_write_inode(ctx, ino, inode, source);
3339 }
3340
3341 /*
3342  * Use the multiple-blocks reclamation code to fix alignment problems in
3343  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
3344  * physical cluster, and we want the block offsets within that cluster to
3345  * line up.
3346  */
3347 static int has_unaligned_cluster_map(e2fsck_t ctx,
3348                                      blk64_t last_pblk, blk64_t last_lblk,
3349                                      blk64_t pblk, blk64_t lblk)
3350 {
3351         blk64_t cluster_mask;
3352
3353         if (!ctx->fs->cluster_ratio_bits)
3354                 return 0;
3355         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
3356
3357         /*
3358          * If the block in the logical cluster doesn't align with the block in
3359          * the physical cluster...
3360          */
3361         if ((lblk & cluster_mask) != (pblk & cluster_mask))
3362                 return 1;
3363
3364         /*
3365          * If we cross a physical cluster boundary within a logical cluster...
3366          */
3367         if (last_pblk && (lblk & cluster_mask) != 0 &&
3368             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
3369             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
3370                 return 1;
3371
3372         return 0;
3373 }
3374
3375 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
3376                              struct process_block_struct *pb,
3377                              blk64_t start_block, blk64_t end_block,
3378                              blk64_t eof_block,
3379                              ext2_extent_handle_t ehandle,
3380                              int try_repairs)
3381 {
3382         struct ext2fs_extent    extent;
3383         blk64_t                 blk, last_lblk;
3384         unsigned int            i, n;
3385         int                     is_dir, is_leaf;
3386         problem_t               problem;
3387         struct ext2_extent_info info;
3388         int                     failed_csum = 0;
3389
3390         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
3391                 failed_csum = 1;
3392
3393         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
3394         if (pctx->errcode)
3395                 return;
3396         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
3397             !pb->eti.force_rebuild &&
3398             info.curr_level < MAX_EXTENT_DEPTH_COUNT) {
3399                 struct extent_tree_level *etl;
3400
3401                 etl = pb->eti.ext_info + info.curr_level;
3402                 etl->num_extents += info.num_entries;
3403                 etl->max_extents += info.max_entries;
3404                 /*
3405                  * Implementation wart: Splitting extent blocks when appending
3406                  * will leave the old block with one free entry.  Therefore
3407                  * unless the node is totally full, pretend that a non-root
3408                  * extent block can hold one fewer entry than it actually does,
3409                  * so that we don't repeatedly rebuild the extent tree.
3410                  */
3411                 if (info.curr_level && info.num_entries < info.max_entries)
3412                         etl->max_extents--;
3413         }
3414
3415         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
3416                                           &extent);
3417         while ((pctx->errcode == 0 ||
3418                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
3419                info.num_entries-- > 0) {
3420                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
3421                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
3422                 last_lblk = extent.e_lblk + extent.e_len - 1;
3423
3424                 problem = 0;
3425                 pctx->blk = extent.e_pblk;
3426                 pctx->blk2 = extent.e_lblk;
3427                 pctx->num = extent.e_len;
3428                 pctx->blkcount = extent.e_lblk + extent.e_len;
3429
3430                 if (extent.e_pblk == 0 ||
3431                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
3432                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
3433                         problem = PR_1_EXTENT_BAD_START_BLK;
3434                 else if (extent.e_lblk < start_block)
3435                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
3436                 else if ((end_block && last_lblk > end_block) &&
3437                          !(last_lblk > eof_block &&
3438                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
3439                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
3440                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
3441                 else if (is_leaf && extent.e_len == 0)
3442                         problem = PR_1_EXTENT_LENGTH_ZERO;
3443                 else if (is_leaf &&
3444                          (extent.e_pblk + extent.e_len) >
3445                          ext2fs_blocks_count(ctx->fs->super))
3446                         problem = PR_1_EXTENT_ENDS_BEYOND;
3447                 else if (is_leaf && is_dir && !pctx->inode->i_size_high &&
3448                          !ext2fs_has_feature_largedir(ctx->fs->super) &&
3449                          ((extent.e_lblk + extent.e_len) >
3450                           (1U << (21 - ctx->fs->super->s_log_block_size))))
3451                         problem = PR_1_TOOBIG_DIR;
3452
3453                 if (is_leaf && problem == 0 && extent.e_len > 0) {
3454 #if 0
3455                         printf("extent_region(ino=%u, expect=%llu, "
3456                                "lblk=%llu, len=%u)\n", pb->ino,
3457                                (unsigned long long) pb->next_lblock,
3458                                (unsigned long long) extent.e_lblk,
3459                                extent.e_len);
3460 #endif
3461                         if (extent.e_lblk < pb->next_lblock)
3462                                 problem = PR_1_EXTENT_COLLISION;
3463                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
3464                                 pb->next_lblock = extent.e_lblk + extent.e_len;
3465                 }
3466
3467                 /*
3468                  * Uninitialized blocks in a directory?  Clear the flag and
3469                  * we'll interpret the blocks later.
3470                  */
3471                 if (try_repairs && is_dir && problem == 0 &&
3472                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3473                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
3474                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3475                         pb->inode_modified = 1;
3476                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3477                                                               &extent);
3478                         if (pctx->errcode)
3479                                 return;
3480                         failed_csum = 0;
3481                 }
3482 #ifdef CONFIG_DEVELOPER_FEATURES
3483                 if (try_repairs && !is_dir && problem == 0 &&
3484                     (ctx->options & E2F_OPT_CLEAR_UNINIT) &&
3485                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
3486                     fix_problem(ctx, PR_1_CLEAR_UNINIT_EXTENT, pctx)) {
3487                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
3488                         pb->inode_modified = 1;
3489                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
3490                                                               &extent);
3491                         if (pctx->errcode)
3492                                 return;
3493                         failed_csum = 0;
3494                 }
3495 #endif
3496                 if (try_repairs && problem) {
3497 report_problem:
3498                         if (fix_problem(ctx, problem, pctx)) {
3499                                 if (ctx->invalid_bitmaps) {
3500                                         /*
3501                                          * If fsck knows the bitmaps are bad,
3502                                          * skip to the next extent and
3503                                          * try to clear this extent again
3504                                          * after fixing the bitmaps, by
3505                                          * restarting fsck.
3506                                          */
3507                                         pctx->errcode = ext2fs_extent_get(
3508                                                           ehandle,
3509                                                           EXT2_EXTENT_NEXT_SIB,
3510                                                           &extent);
3511                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3512                                         if (pctx->errcode ==
3513                                                     EXT2_ET_NO_CURRENT_NODE) {
3514                                                 pctx->errcode = 0;
3515                                                 break;
3516                                         }
3517                                         continue;
3518                                 }
3519                                 e2fsck_read_bitmaps(ctx);
3520                                 pb->inode_modified = 1;
3521                                 pctx->errcode =
3522                                         ext2fs_extent_delete(ehandle, 0);
3523                                 if (pctx->errcode) {
3524                                         pctx->str = "ext2fs_extent_delete";
3525                                         return;
3526                                 }
3527                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3528                                 if (pctx->errcode &&
3529                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
3530                                         pctx->str = "ext2fs_extent_fix_parents";
3531                                         return;
3532                                 }
3533                                 pctx->errcode = ext2fs_extent_get(ehandle,
3534                                                                   EXT2_EXTENT_CURRENT,
3535                                                                   &extent);
3536                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
3537                                         pctx->errcode = 0;
3538                                         break;
3539                                 }
3540                                 failed_csum = 0;
3541                                 continue;
3542                         }
3543                         goto next;
3544                 }
3545
3546                 if (!is_leaf) {
3547                         blk64_t lblk = extent.e_lblk;
3548                         int next_try_repairs = 1;
3549
3550                         blk = extent.e_pblk;
3551
3552                         /*
3553                          * If this lower extent block collides with critical
3554                          * metadata, don't try to repair the damage.  Pass 1b
3555                          * will reallocate the block; then we can try again.
3556                          */
3557                         if (pb->ino != EXT2_RESIZE_INO &&
3558                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
3559                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3560                                                       extent.e_pblk)) {
3561                                 next_try_repairs = 0;
3562                                 pctx->blk = blk;
3563                                 fix_problem(ctx,
3564                                             PR_1_CRITICAL_METADATA_COLLISION,
3565                                             pctx);
3566                                 if ((ctx->options & E2F_OPT_NO) == 0)
3567                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3568                         }
3569                         pctx->errcode = ext2fs_extent_get(ehandle,
3570                                                   EXT2_EXTENT_DOWN, &extent);
3571                         if (pctx->errcode &&
3572                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
3573                                 pctx->str = "EXT2_EXTENT_DOWN";
3574                                 problem = PR_1_EXTENT_HEADER_INVALID;
3575                                 if (!next_try_repairs)
3576                                         return;
3577                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
3578                                         goto report_problem;
3579                                 return;
3580                         }
3581                         /* The next extent should match this index's logical start */
3582                         if (extent.e_lblk != lblk) {
3583                                 struct ext2_extent_info e_info;
3584
3585                                 pctx->errcode = ext2fs_extent_get_info(ehandle,
3586                                                                        &e_info);
3587                                 if (pctx->errcode) {
3588                                         pctx->str = "ext2fs_extent_get_info";
3589                                         return;
3590                                 }
3591                                 pctx->blk = lblk;
3592                                 pctx->blk2 = extent.e_lblk;
3593                                 pctx->num = e_info.curr_level - 1;
3594                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
3595                                 if (fix_problem(ctx, problem, pctx)) {
3596                                         pb->inode_modified = 1;
3597                                         pctx->errcode =
3598                                                 ext2fs_extent_fix_parents(ehandle);
3599                                         if (pctx->errcode) {
3600                                                 pctx->str = "ext2fs_extent_fix_parents";
3601                                                 return;
3602                                         }
3603                                 }
3604                         }
3605                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
3606                                          last_lblk, eof_block, ehandle,
3607                                          next_try_repairs);
3608                         if (pctx->errcode)
3609                                 return;
3610                         pctx->errcode = ext2fs_extent_get(ehandle,
3611                                                   EXT2_EXTENT_UP, &extent);
3612                         if (pctx->errcode) {
3613                                 pctx->str = "EXT2_EXTENT_UP";
3614                                 return;
3615                         }
3616                         mark_block_used(ctx, blk);
3617                         pb->num_blocks++;
3618                         goto next;
3619                 }
3620
3621                 if ((pb->previous_block != 0) &&
3622                     (pb->previous_block+1 != extent.e_pblk)) {
3623                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3624                                 char type = '?';
3625
3626                                 if (pb->is_dir)
3627                                         type = 'd';
3628                                 else if (pb->is_reg)
3629                                         type = 'f';
3630
3631                                 printf(("%6lu(%c): expecting %6lu "
3632                                         "actual extent "
3633                                         "phys %6lu log %lu len %lu\n"),
3634                                        (unsigned long) pctx->ino, type,
3635                                        (unsigned long) pb->previous_block+1,
3636                                        (unsigned long) extent.e_pblk,
3637                                        (unsigned long) extent.e_lblk,
3638                                        (unsigned long) extent.e_len);
3639                         }
3640                         pb->fragmented = 1;
3641                 }
3642                 /*
3643                  * If we notice a gap in the logical block mappings of an
3644                  * extent-mapped directory, offer to close the hole by
3645                  * moving the logical block down, otherwise we'll go mad in
3646                  * pass 3 allocating empty directory blocks to fill the hole.
3647                  */
3648                 if (try_repairs && is_dir &&
3649                     pb->last_block + 1 < extent.e_lblk) {
3650                         blk64_t new_lblk;
3651
3652                         new_lblk = pb->last_block + 1;
3653                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3654                                 new_lblk = ((new_lblk +
3655                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3656                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3657                                            (extent.e_pblk &
3658                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3659                         pctx->blk = extent.e_lblk;
3660                         pctx->blk2 = new_lblk;
3661                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3662                                 extent.e_lblk = new_lblk;
3663                                 pb->inode_modified = 1;
3664                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3665                                                                 0, &extent);
3666                                 if (pctx->errcode) {
3667                                         pctx->errcode = 0;
3668                                         goto alloc_later;
3669                                 }
3670                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3671                                 if (pctx->errcode)
3672                                         goto failed_add_dir_block;
3673                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3674                                                                 extent.e_lblk);
3675                                 if (pctx->errcode)
3676                                         goto failed_add_dir_block;
3677                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3678                                 failed_csum = 0;
3679                         }
3680                 }
3681 alloc_later:
3682                 if (is_dir) {
3683                         while (++pb->last_db_block <
3684                                (e2_blkcnt_t) extent.e_lblk) {
3685                                 pctx->errcode = ext2fs_add_dir_block2(
3686                                                         ctx->fs->dblist,
3687                                                         pb->ino, 0,
3688                                                         pb->last_db_block);
3689                                 if (pctx->errcode) {
3690                                         pctx->blk = 0;
3691                                         pctx->num = pb->last_db_block;
3692                                         goto failed_add_dir_block;
3693                                 }
3694                         }
3695
3696                         for (i = 0; i < extent.e_len; i++) {
3697                                 pctx->errcode = ext2fs_add_dir_block2(
3698                                                         ctx->fs->dblist,
3699                                                         pctx->ino,
3700                                                         extent.e_pblk + i,
3701                                                         extent.e_lblk + i);
3702                                 if (pctx->errcode) {
3703                                         pctx->blk = extent.e_pblk + i;
3704                                         pctx->num = extent.e_lblk + i;
3705                                 failed_add_dir_block:
3706                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3707                                         /* Should never get here */
3708                                         ctx->flags |= E2F_FLAG_ABORT;
3709                                         return;
3710                                 }
3711                         }
3712                         if (extent.e_len > 0)
3713                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3714                 }
3715                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3716                                               pb->last_block,
3717                                               extent.e_pblk,
3718                                               extent.e_lblk)) {
3719                         for (i = 0; i < extent.e_len; i++) {
3720                                 pctx->blk = extent.e_lblk + i;
3721                                 pctx->blk2 = extent.e_pblk + i;
3722                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3723                                 mark_block_used(ctx, extent.e_pblk + i);
3724                                 mark_block_used(ctx, extent.e_pblk + i);
3725                         }
3726                 }
3727
3728                 /*
3729                  * Check whether first cluster got marked in previous iteration.
3730                  */
3731                 if (ctx->fs->cluster_ratio_bits &&
3732                     pb->previous_block &&
3733                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3734                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3735                         /* Set blk to the beginning of next cluster. */
3736                         blk = EXT2FS_C2B(
3737                                 ctx->fs,
3738                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3739                 else
3740                         /* Set blk to the beginning of current cluster. */
3741                         blk = EXT2FS_C2B(ctx->fs,
3742                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3743
3744                 if (blk < extent.e_pblk + extent.e_len) {
3745                         mark_blocks_used(ctx, blk,
3746                                          extent.e_pblk + extent.e_len - blk);
3747                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3748                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3749                         pb->num_blocks += n;
3750                 }
3751                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3752                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3753                 start_block = pb->last_block = last_lblk;
3754                 if (is_leaf && !is_dir &&
3755                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3756                         pb->last_init_lblock = last_lblk;
3757         next:
3758                 pctx->errcode = ext2fs_extent_get(ehandle,
3759                                                   EXT2_EXTENT_NEXT_SIB,
3760                                                   &extent);
3761         }
3762
3763         /* Failed csum but passes checks?  Ask to fix checksum. */
3764         if (failed_csum &&
3765             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3766                 pb->inode_modified = 1;
3767                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3768                 if (pctx->errcode)
3769                         return;
3770         }
3771
3772         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3773                 pctx->errcode = 0;
3774 }
3775
3776 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3777                                  struct process_block_struct *pb)
3778 {
3779         struct ext2_extent_info info;
3780         struct ext2_inode       *inode = pctx->inode;
3781         ext2_extent_handle_t    ehandle;
3782         ext2_filsys             fs = ctx->fs;
3783         ext2_ino_t              ino = pctx->ino;
3784         errcode_t               retval;
3785         blk64_t                 eof_lblk;
3786         struct ext3_extent_header       *eh;
3787
3788         /* Check for a proper extent header... */
3789         eh = (struct ext3_extent_header *) &inode->i_block[0];
3790         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3791         if (retval) {
3792                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3793                         e2fsck_clear_inode(ctx, ino, inode, 0,
3794                                            "check_blocks_extents");
3795                 pctx->errcode = 0;
3796                 return;
3797         }
3798
3799         /* ...since this function doesn't fail if i_block is zeroed. */
3800         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3801         if (pctx->errcode) {
3802                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3803                         e2fsck_clear_inode(ctx, ino, inode, 0,
3804                                            "check_blocks_extents");
3805                 pctx->errcode = 0;
3806                 return;
3807         }
3808
3809         retval = ext2fs_extent_get_info(ehandle, &info);
3810         if (retval == 0) {
3811                 int max_depth = info.max_depth;
3812
3813                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3814                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3815                 ctx->extent_depth_count[max_depth]++;
3816         }
3817
3818         /* Check maximum extent depth */
3819         pctx->blk = info.max_depth;
3820         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3821         if (pctx->blk2 < pctx->blk &&
3822             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3823                 pb->eti.force_rebuild = 1;
3824
3825         /* Can we collect extent tree level stats? */
3826         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3827         if (pctx->blk2 > pctx->blk)
3828                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3829         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3830         pb->eti.ino = pb->ino;
3831
3832         pb->next_lblock = 0;
3833
3834         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3835                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3836         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3837         if (pctx->errcode &&
3838             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3839                 pb->num_blocks = 0;
3840                 inode->i_blocks = 0;
3841                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3842                                    "check_blocks_extents");
3843                 pctx->errcode = 0;
3844         }
3845         ext2fs_extent_free(ehandle);
3846
3847         /* Rebuild unless it's a dir and we're rehashing it */
3848         if (LINUX_S_ISDIR(inode->i_mode) &&
3849             e2fsck_dir_will_be_rehashed(ctx, ino))
3850                 return;
3851
3852         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3853                 e2fsck_rebuild_extents_later(ctx, ino);
3854         else
3855                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3856 }
3857
3858 /*
3859  * In fact we don't need to check blocks for an inode with inline data
3860  * because this inode doesn't have any blocks.  In this function all
3861  * we need to do is add this inode into dblist when it is a directory.
3862  */
3863 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3864                                      struct process_block_struct *pb)
3865 {
3866         int     flags;
3867         size_t  inline_data_size = 0;
3868
3869         if (!pb->is_dir) {
3870                 pctx->errcode = 0;
3871                 return;
3872         }
3873
3874         /* Process the dirents in i_block[] as the "first" block. */
3875         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3876         if (pctx->errcode)
3877                 goto err;
3878
3879         /* Process the dirents in the EA as a "second" block. */
3880         flags = ctx->fs->flags;
3881         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3882         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3883                                                 &inline_data_size);
3884         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3885                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3886         if (pctx->errcode) {
3887                 pctx->errcode = 0;
3888                 return;
3889         }
3890
3891         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3892                 return;
3893
3894         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3895         if (pctx->errcode)
3896                 goto err;
3897
3898         return;
3899 err:
3900         pctx->blk = 0;
3901         pctx->num = 0;
3902         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3903         ctx->flags |= E2F_FLAG_ABORT;
3904 }
3905
3906 /*
3907  * This subroutine is called on each inode to account for all of the
3908  * blocks used by that inode.
3909  */
3910 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3911                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3912 {
3913         ext2_filsys fs = ctx->fs;
3914         struct process_block_struct pb;
3915         ext2_ino_t      ino = pctx->ino;
3916         struct ext2_inode *inode = pctx->inode;
3917         unsigned        bad_size = 0;
3918         int             dirty_inode = 0;
3919         int             extent_fs;
3920         int             inlinedata_fs;
3921         __u64           size;
3922         struct ea_quota ea_block_quota;
3923
3924         pb.ino = ino;
3925         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3926                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3927         pb.last_block = ~0;
3928         pb.last_init_lblock = -1;
3929         pb.last_db_block = -1;
3930         pb.num_illegal_blocks = 0;
3931         pb.suppress = 0; pb.clear = 0;
3932         pb.fragmented = 0;
3933         pb.compressed = 0;
3934         pb.previous_block = 0;
3935         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3936         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3937         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3938         pb.inode = inode;
3939         pb.pctx = pctx;
3940         pb.ctx = ctx;
3941         pb.inode_modified = 0;
3942         pb.eti.force_rebuild = 0;
3943         pctx->ino = ino;
3944         pctx->errcode = 0;
3945
3946         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3947         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3948
3949         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3950                 if (e2fsck_should_abort(ctx))
3951                         goto out;
3952                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3953         }
3954
3955         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3956                 check_blocks_inline_data(ctx, pctx, &pb);
3957         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3958                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3959                         check_blocks_extents(ctx, pctx, &pb);
3960                 else {
3961                         int flags;
3962                         /*
3963                          * If we've modified the inode, write it out before
3964                          * iterate() tries to use it.
3965                          */
3966                         if (dirty_inode) {
3967                                 e2fsck_write_inode(ctx, ino, inode,
3968                                                    "check_blocks");
3969                                 dirty_inode = 0;
3970                         }
3971                         flags = fs->flags;
3972                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3973                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3974                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3975                                                 block_buf, process_block, &pb);
3976                         /*
3977                          * We do not have uninitialized extents in non extent
3978                          * files.
3979                          */
3980                         pb.last_init_lblock = pb.last_block;
3981                         /*
3982                          * If iterate() changed a block mapping, we have to
3983                          * re-read the inode.  If we decide to clear the
3984                          * inode after clearing some stuff, we'll re-write the
3985                          * bad mappings into the inode!
3986                          */
3987                         if (pb.inode_modified)
3988                                 e2fsck_read_inode(ctx, ino, inode,
3989                                                   "check_blocks");
3990                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3991                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3992
3993                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3994 #ifdef DEBUG
3995                                 printf("bmap rebuild ino=%d\n", ino);
3996 #endif
3997                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3998                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3999                                         e2fsck_rebuild_extents_later(ctx, ino);
4000                         }
4001                 }
4002         }
4003         end_problem_latch(ctx, PR_LATCH_BLOCK);
4004         end_problem_latch(ctx, PR_LATCH_TOOBIG);
4005         if (e2fsck_should_abort(ctx))
4006                 goto out;
4007         if (pctx->errcode)
4008                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
4009
4010         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
4011                 if (LINUX_S_ISDIR(inode->i_mode))
4012                         ctx->fs_fragmented_dir++;
4013                 else
4014                         ctx->fs_fragmented++;
4015         }
4016
4017         if (pb.clear) {
4018                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
4019                                    "check_blocks");
4020                 return;
4021         }
4022
4023         if (inode->i_flags & EXT2_INDEX_FL) {
4024                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
4025                         inode->i_flags &= ~EXT2_INDEX_FL;
4026                         dirty_inode++;
4027                 } else {
4028                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
4029                 }
4030         }
4031
4032         if (!pb.num_blocks && pb.is_dir &&
4033             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
4034                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
4035                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
4036                         ctx->fs_directory_count--;
4037                         return;
4038                 }
4039         }
4040
4041         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
4042             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
4043             !(inode->i_flags & EXT4_EA_INODE_FL)) {
4044                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
4045                                ino,
4046                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
4047                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
4048                                   ino, (ea_ibody_quota ?
4049                                         ea_ibody_quota->inodes : 0) +
4050                                                 ea_block_quota.inodes + 1);
4051         }
4052
4053         if (!ext2fs_has_feature_huge_file(fs->super) ||
4054             !(inode->i_flags & EXT4_HUGE_FILE_FL))
4055                 pb.num_blocks *= (fs->blocksize / 512);
4056         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
4057 #if 0
4058         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
4059                ino, inode->i_size, (unsigned long long) pb.last_block,
4060                (unsigned long long) ext2fs_inode_i_blocks(fs, inode),
4061                (unsigned long long) pb.num_blocks);
4062 #endif
4063         size = EXT2_I_SIZE(inode);
4064         if (pb.is_dir) {
4065                 unsigned nblock = size >> EXT2_BLOCK_SIZE_BITS(fs->super);
4066                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
4067                         int flags;
4068                         size_t sz = 0;
4069                         errcode_t err;
4070
4071                         flags = ctx->fs->flags;
4072                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
4073                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
4074                                                       &sz);
4075                         ctx->fs->flags = (flags &
4076                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
4077                                          (ctx->fs->flags &
4078                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
4079                         if (err || sz != size) {
4080                                 bad_size = 7;
4081                                 pctx->num = sz;
4082                         }
4083                 } else if (size & (fs->blocksize - 1))
4084                         bad_size = 5;
4085                 else if (nblock > (pb.last_block + 1))
4086                         bad_size = 1;
4087                 else if (nblock < (pb.last_block + 1)) {
4088                         if (((pb.last_block + 1) - nblock) >
4089                             fs->super->s_prealloc_dir_blocks)
4090                                 bad_size = 2;
4091                 }
4092         } else {
4093                 if ((pb.last_init_lblock >= 0) &&
4094                     /* Do not allow initialized allocated blocks past i_size*/
4095                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
4096                     !(inode->i_flags & EXT4_VERITY_FL))
4097                         bad_size = 3;
4098                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4099                          size > ext2_max_sizes[fs->super->s_log_block_size])
4100                         /* too big for a direct/indirect-mapped file */
4101                         bad_size = 4;
4102                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
4103                          size >
4104                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
4105                         /* too big for an extent-based file - 32bit ee_block */
4106                         bad_size = 6;
4107         }
4108         /* i_size for symlinks is checked elsewhere */
4109         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
4110                 /* Did inline_data set pctx->num earlier? */
4111                 if (bad_size != 7)
4112                         pctx->num = (pb.last_block + 1) * fs->blocksize;
4113                 pctx->group = bad_size;
4114                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
4115                         ext2fs_inode_size_set(fs, inode, pctx->num);
4116                         if (EXT2_I_SIZE(inode) == 0 &&
4117                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
4118                                 memset(inode->i_block, 0,
4119                                        sizeof(inode->i_block));
4120                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
4121                         }
4122                         dirty_inode++;
4123                 }
4124                 pctx->num = 0;
4125         }
4126         if (LINUX_S_ISREG(inode->i_mode) &&
4127             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
4128                 ctx->large_files++;
4129         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
4130             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
4131              (ext2fs_has_feature_huge_file(fs->super) &&
4132               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
4133               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
4134                 pctx->num = pb.num_blocks;
4135                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
4136                         inode->i_blocks = pb.num_blocks;
4137                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
4138                         dirty_inode++;
4139                 }
4140                 pctx->num = 0;
4141         }
4142
4143         /*
4144          * The kernel gets mad if we ask it to allocate bigalloc clusters to
4145          * a block mapped file, so rebuild it as an extent file.  We can skip
4146          * symlinks because they're never rewritten.
4147          */
4148         if (ext2fs_has_feature_bigalloc(fs->super) &&
4149             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
4150             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
4151             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
4152             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
4153             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
4154                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
4155                 if (pctx->errcode)
4156                         goto out;
4157         }
4158
4159         if (ctx->dirs_to_hash && pb.is_dir &&
4160             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
4161             !(inode->i_flags & EXT2_INDEX_FL) &&
4162             ((inode->i_size / fs->blocksize) >= 3))
4163                 e2fsck_rehash_dir_later(ctx, ino);
4164
4165 out:
4166         if (dirty_inode)
4167                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
4168 }
4169
4170 #if 0
4171 /*
4172  * Helper function called by process block when an illegal block is
4173  * found.  It returns a description about why the block is illegal
4174  */
4175 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
4176 {
4177         blk64_t super;
4178         int     i;
4179         static char     problem[80];
4180
4181         super = fs->super->s_first_data_block;
4182         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
4183         if (block < super) {
4184                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
4185                 return(problem);
4186         } else if (block >= ext2fs_blocks_count(fs->super)) {
4187                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
4188                 return(problem);
4189         }
4190         for (i = 0; i < fs->group_desc_count; i++) {
4191                 if (block == super) {
4192                         sprintf(problem, "is the superblock in group %d", i);
4193                         break;
4194                 }
4195                 if (block > super &&
4196                     block <= (super + fs->desc_blocks)) {
4197                         sprintf(problem, "is in the group descriptors "
4198                                 "of group %d", i);
4199                         break;
4200                 }
4201                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
4202                         sprintf(problem, "is the block bitmap of group %d", i);
4203                         break;
4204                 }
4205                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
4206                         sprintf(problem, "is the inode bitmap of group %d", i);
4207                         break;
4208                 }
4209                 if (block >= ext2fs_inode_table_loc(fs, i) &&
4210                     (block < ext2fs_inode_table_loc(fs, i)
4211                      + fs->inode_blocks_per_group)) {
4212                         sprintf(problem, "is in the inode table of group %d",
4213                                 i);
4214                         break;
4215                 }
4216                 super += fs->super->s_blocks_per_group;
4217         }
4218         return(problem);
4219 }
4220 #endif
4221
4222 /*
4223  * This is a helper function for check_blocks().
4224  */
4225 static int process_block(ext2_filsys fs,
4226                   blk64_t       *block_nr,
4227                   e2_blkcnt_t blockcnt,
4228                   blk64_t ref_block EXT2FS_ATTR((unused)),
4229                   int ref_offset EXT2FS_ATTR((unused)),
4230                   void *priv_data)
4231 {
4232         struct process_block_struct *p;
4233         struct problem_context *pctx;
4234         blk64_t blk = *block_nr;
4235         int     ret_code = 0;
4236         problem_t       problem = 0;
4237         e2fsck_t        ctx;
4238
4239         p = (struct process_block_struct *) priv_data;
4240         pctx = p->pctx;
4241         ctx = p->ctx;
4242
4243         /*
4244          * For a directory, add logical block zero for processing even if it's
4245          * not mapped or we'll be perennially stuck with broken "." and ".."
4246          * entries.
4247          */
4248         if (p->is_dir && blockcnt == 0 && blk == 0) {
4249                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
4250                 if (pctx->errcode) {
4251                         pctx->blk = blk;
4252                         pctx->num = blockcnt;
4253                         goto failed_add_dir_block;
4254                 }
4255                 p->last_db_block++;
4256         }
4257
4258         if (blk == 0)
4259                 return 0;
4260
4261 #if 0
4262         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
4263                blockcnt);
4264 #endif
4265
4266         /*
4267          * Simplistic fragmentation check.  We merely require that the
4268          * file be contiguous.  (Which can never be true for really
4269          * big files that are greater than a block group.)
4270          */
4271         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
4272                 if (p->previous_block+1 != blk) {
4273                         if (ctx->options & E2F_OPT_FRAGCHECK) {
4274                                 char type = '?';
4275
4276                                 if (p->is_dir)
4277                                         type = 'd';
4278                                 else if (p->is_reg)
4279                                         type = 'f';
4280
4281                                 printf(_("%6lu(%c): expecting %6lu "
4282                                          "got phys %6lu (blkcnt %lld)\n"),
4283                                        (unsigned long) pctx->ino, type,
4284                                        (unsigned long) p->previous_block+1,
4285                                        (unsigned long) blk,
4286                                        (long long) blockcnt);
4287                         }
4288                         p->fragmented = 1;
4289                 }
4290         }
4291
4292         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
4293             !pctx->inode->i_size_high &&
4294             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
4295                 problem = PR_1_TOOBIG_DIR;
4296         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
4297                 problem = PR_1_TOOBIG_DIR;
4298         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
4299                 problem = PR_1_TOOBIG_REG;
4300         if (!p->is_dir && !p->is_reg && blockcnt > 0)
4301                 problem = PR_1_TOOBIG_SYMLINK;
4302
4303         if (blk < fs->super->s_first_data_block ||
4304             blk >= ext2fs_blocks_count(fs->super))
4305                 problem = PR_1_ILLEGAL_BLOCK_NUM;
4306
4307         /*
4308          * If this IND/DIND/TIND block is squatting atop some critical metadata
4309          * (group descriptors, superblock, bitmap, inode table), any write to
4310          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
4311          * fix that and restart fsck.
4312          */
4313         if (blockcnt < 0 &&
4314             p->ino != EXT2_RESIZE_INO &&
4315             blk < ctx->fs->super->s_blocks_count &&
4316             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
4317                 pctx->blk = blk;
4318                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
4319                 if ((ctx->options & E2F_OPT_NO) == 0)
4320                         ctx->flags |= E2F_FLAG_RESTART_LATER;
4321         }
4322
4323         if (problem) {
4324                 p->num_illegal_blocks++;
4325                 /*
4326                  * A bit of subterfuge here -- we're trying to fix a block
4327                  * mapping, but the IND/DIND/TIND block could have collided
4328                  * with some critical metadata.  So, fix the in-core mapping so
4329                  * iterate won't go insane, but return 0 instead of
4330                  * BLOCK_CHANGED so that it won't write the remapping out to
4331                  * our multiply linked block.
4332                  *
4333                  * Even if we previously determined that an *IND block
4334                  * conflicts with critical metadata, we must still try to
4335                  * iterate the *IND block as if it is an *IND block to find and
4336                  * mark the blocks it points to.  Better to be overly cautious
4337                  * with the used_blocks map so that we don't move the *IND
4338                  * block to a block that's really in use!
4339                  */
4340                 if (p->ino != EXT2_RESIZE_INO &&
4341                     ref_block != 0 &&
4342                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
4343                                               ref_block)) {
4344                         *block_nr = 0;
4345                         return 0;
4346                 }
4347                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
4348                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
4349                                 p->clear = 1;
4350                                 return BLOCK_ABORT;
4351                         }
4352                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
4353                                 p->suppress = 1;
4354                                 set_latch_flags(PR_LATCH_BLOCK,
4355                                                 PRL_SUPPRESS, 0);
4356                         }
4357                 }
4358                 pctx->blk = blk;
4359                 pctx->blkcount = blockcnt;
4360                 if (fix_problem(ctx, problem, pctx)) {
4361                         blk = *block_nr = 0;
4362                         ret_code = BLOCK_CHANGED;
4363                         p->inode_modified = 1;
4364                         /*
4365                          * If the directory block is too big and is beyond the
4366                          * end of the FS, don't bother trying to add it for
4367                          * processing -- the kernel would never have created a
4368                          * directory this large, and we risk an ENOMEM abort.
4369                          * In any case, the toobig handler for extent-based
4370                          * directories also doesn't feed toobig blocks to
4371                          * pass 2.
4372                          */
4373                         if (problem == PR_1_TOOBIG_DIR)
4374                                 return ret_code;
4375                         goto mark_dir;
4376                 } else
4377                         return 0;
4378         }
4379
4380         if (p->ino == EXT2_RESIZE_INO) {
4381                 /*
4382                  * The resize inode has already be sanity checked
4383                  * during pass #0 (the superblock checks).  All we
4384                  * have to do is mark the double indirect block as
4385                  * being in use; all of the other blocks are handled
4386                  * by mark_table_blocks()).
4387                  */
4388                 if (blockcnt == BLOCK_COUNT_DIND)
4389                         mark_block_used(ctx, blk);
4390                 p->num_blocks++;
4391         } else if (!(ctx->fs->cluster_ratio_bits &&
4392                      p->previous_block &&
4393                      (EXT2FS_B2C(ctx->fs, blk) ==
4394                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
4395                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
4396                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
4397                 mark_block_used(ctx, blk);
4398                 p->num_blocks++;
4399         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
4400                                              p->last_block, blk, blockcnt)) {
4401                 pctx->blk = blockcnt;
4402                 pctx->blk2 = blk;
4403                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
4404                 mark_block_used(ctx, blk);
4405                 mark_block_used(ctx, blk);
4406         }
4407         if (blockcnt >= 0)
4408                 p->last_block = blockcnt;
4409         p->previous_block = blk;
4410 mark_dir:
4411         if (p->is_dir && (blockcnt >= 0)) {
4412                 while (++p->last_db_block < blockcnt) {
4413                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
4414                                                               p->ino, 0,
4415                                                               p->last_db_block);
4416                         if (pctx->errcode) {
4417                                 pctx->blk = 0;
4418                                 pctx->num = p->last_db_block;
4419                                 goto failed_add_dir_block;
4420                         }
4421                 }
4422                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
4423                                                       blk, blockcnt);
4424                 if (pctx->errcode) {
4425                         pctx->blk = blk;
4426                         pctx->num = blockcnt;
4427                 failed_add_dir_block:
4428                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
4429                         /* Should never get here */
4430                         ctx->flags |= E2F_FLAG_ABORT;
4431                         return BLOCK_ABORT;
4432                 }
4433         }
4434         return ret_code;
4435 }
4436
4437 static int process_bad_block(ext2_filsys fs,
4438                       blk64_t *block_nr,
4439                       e2_blkcnt_t blockcnt,
4440                       blk64_t ref_block EXT2FS_ATTR((unused)),
4441                       int ref_offset EXT2FS_ATTR((unused)),
4442                       void *priv_data)
4443 {
4444         struct process_block_struct *p;
4445         blk64_t         blk = *block_nr;
4446         blk64_t         first_block;
4447         dgrp_t          i;
4448         struct problem_context *pctx;
4449         e2fsck_t        ctx;
4450
4451         if (!blk)
4452                 return 0;
4453
4454         p = (struct process_block_struct *) priv_data;
4455         ctx = p->ctx;
4456         pctx = p->pctx;
4457
4458         pctx->ino = EXT2_BAD_INO;
4459         pctx->blk = blk;
4460         pctx->blkcount = blockcnt;
4461
4462         if ((blk < fs->super->s_first_data_block) ||
4463             (blk >= ext2fs_blocks_count(fs->super))) {
4464                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
4465                         *block_nr = 0;
4466                         return BLOCK_CHANGED;
4467                 } else
4468                         return 0;
4469         }
4470
4471         if (blockcnt < 0) {
4472                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
4473                         p->bbcheck = 1;
4474                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
4475                                 *block_nr = 0;
4476                                 return BLOCK_CHANGED;
4477                         }
4478                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4479                                                     blk)) {
4480                         p->bbcheck = 1;
4481                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
4482                                         pctx)) {
4483                                 *block_nr = 0;
4484                                 return BLOCK_CHANGED;
4485                         }
4486                         if (e2fsck_should_abort(ctx))
4487                                 return BLOCK_ABORT;
4488                 } else
4489                         mark_block_used(ctx, blk);
4490                 return 0;
4491         }
4492 #if 0
4493         printf ("DEBUG: Marking %u as bad.\n", blk);
4494 #endif
4495         ctx->fs_badblocks_count++;
4496         /*
4497          * If the block is not used, then mark it as used and return.
4498          * If it is already marked as found, this must mean that
4499          * there's an overlap between the filesystem table blocks
4500          * (bitmaps and inode table) and the bad block list.
4501          */
4502         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
4503                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4504                 return 0;
4505         }
4506         /*
4507          * Try to find the where the filesystem block was used...
4508          */
4509         first_block = fs->super->s_first_data_block;
4510
4511         for (i = 0; i < fs->group_desc_count; i++ ) {
4512                 pctx->group = i;
4513                 pctx->blk = blk;
4514                 if (!ext2fs_bg_has_super(fs, i))
4515                         goto skip_super;
4516                 if (blk == first_block) {
4517                         if (i == 0) {
4518                                 if (fix_problem(ctx,
4519                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
4520                                                 pctx)) {
4521                                         *block_nr = 0;
4522                                         return BLOCK_CHANGED;
4523                                 }
4524                                 return 0;
4525                         }
4526                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
4527                         return 0;
4528                 }
4529                 if ((blk > first_block) &&
4530                     (blk <= first_block + fs->desc_blocks)) {
4531                         if (i == 0) {
4532                                 pctx->blk = *block_nr;
4533                                 if (fix_problem(ctx,
4534                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
4535                                         *block_nr = 0;
4536                                         return BLOCK_CHANGED;
4537                                 }
4538                                 return 0;
4539                         }
4540                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
4541                         return 0;
4542                 }
4543         skip_super:
4544                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
4545                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
4546                                 ctx->invalid_block_bitmap_flag[i]++;
4547                                 ctx->invalid_bitmaps++;
4548                         }
4549                         return 0;
4550                 }
4551                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
4552                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
4553                                 ctx->invalid_inode_bitmap_flag[i]++;
4554                                 ctx->invalid_bitmaps++;
4555                         }
4556                         return 0;
4557                 }
4558                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
4559                     (blk < (ext2fs_inode_table_loc(fs, i) +
4560                             fs->inode_blocks_per_group))) {
4561                         /*
4562                          * If there are bad blocks in the inode table,
4563                          * the inode scan code will try to do
4564                          * something reasonable automatically.
4565                          */
4566                         return 0;
4567                 }
4568                 first_block += fs->super->s_blocks_per_group;
4569         }
4570         /*
4571          * If we've gotten to this point, then the only
4572          * possibility is that the bad block inode meta data
4573          * is using a bad block.
4574          */
4575         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
4576             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
4577             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
4578                 p->bbcheck = 1;
4579                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
4580                         *block_nr = 0;
4581                         return BLOCK_CHANGED;
4582                 }
4583                 if (e2fsck_should_abort(ctx))
4584                         return BLOCK_ABORT;
4585                 return 0;
4586         }
4587
4588         pctx->group = -1;
4589
4590         /* Warn user that the block wasn't claimed */
4591         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
4592
4593         return 0;
4594 }
4595
4596 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
4597                             const char *name, int num, blk64_t *new_block)
4598 {
4599         ext2_filsys fs = ctx->fs;
4600         dgrp_t          last_grp;
4601         blk64_t         old_block = *new_block;
4602         blk64_t         last_block;
4603         dgrp_t          flexbg;
4604         unsigned        flexbg_size;
4605         int             i, is_flexbg;
4606         char            *buf;
4607         struct problem_context  pctx;
4608
4609         clear_problem_context(&pctx);
4610
4611         pctx.group = group;
4612         pctx.blk = old_block;
4613         pctx.str = name;
4614
4615         /*
4616          * For flex_bg filesystems, first try to allocate the metadata
4617          * within the flex_bg, and if that fails then try finding the
4618          * space anywhere in the filesystem.
4619          */
4620         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
4621         if (is_flexbg) {
4622                 flexbg_size = 1U << fs->super->s_log_groups_per_flex;
4623                 flexbg = group / flexbg_size;
4624                 first_block = ext2fs_group_first_block2(fs,
4625                                                         flexbg_size * flexbg);
4626                 last_grp = group | (flexbg_size - 1);
4627                 if (last_grp >= fs->group_desc_count)
4628                         last_grp = fs->group_desc_count - 1;
4629                 last_block = ext2fs_group_last_block2(fs, last_grp);
4630         } else
4631                 last_block = ext2fs_group_last_block2(fs, group);
4632         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4633                                                num, ctx->block_found_map,
4634                                                new_block);
4635         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4636                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4637                                 fs->super->s_first_data_block,
4638                                 ext2fs_blocks_count(fs->super),
4639                                 num, ctx->block_found_map, new_block);
4640         if (pctx.errcode) {
4641                 pctx.num = num;
4642                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4643                 ext2fs_unmark_valid(fs);
4644                 ctx->flags |= E2F_FLAG_ABORT;
4645                 return;
4646         }
4647         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4648         if (pctx.errcode) {
4649                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4650                 ext2fs_unmark_valid(fs);
4651                 ctx->flags |= E2F_FLAG_ABORT;
4652                 return;
4653         }
4654         ext2fs_mark_super_dirty(fs);
4655         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4656         pctx.blk2 = *new_block;
4657         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4658                           PR_1_RELOC_TO), &pctx);
4659         pctx.blk2 = 0;
4660         for (i = 0; i < num; i++) {
4661                 pctx.blk = i;
4662                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4663                 if (old_block) {
4664                         pctx.errcode = io_channel_read_blk64(fs->io,
4665                                    old_block + i, 1, buf);
4666                         if (pctx.errcode)
4667                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4668                         pctx.blk = (*new_block) + i;
4669                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4670                                                               1, buf);
4671                 } else {
4672                         pctx.blk = (*new_block) + i;
4673                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4674                                                            NULL, NULL);
4675                 }
4676
4677                 if (pctx.errcode)
4678                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4679         }
4680         ext2fs_free_mem(&buf);
4681 }
4682
4683 /*
4684  * This routine gets called at the end of pass 1 if bad blocks are
4685  * detected in the superblock, group descriptors, inode_bitmaps, or
4686  * block bitmaps.  At this point, all of the blocks have been mapped
4687  * out, so we can try to allocate new block(s) to replace the bad
4688  * blocks.
4689  */
4690 static void handle_fs_bad_blocks(e2fsck_t ctx)
4691 {
4692         ext2_filsys fs = ctx->fs;
4693         dgrp_t          i;
4694         blk64_t         first_block;
4695         blk64_t         new_blk;
4696
4697         for (i = 0; i < fs->group_desc_count; i++) {
4698                 first_block = ext2fs_group_first_block2(fs, i);
4699
4700                 if (ctx->invalid_block_bitmap_flag[i]) {
4701                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4702                         new_table_block(ctx, first_block, i, _("block bitmap"),
4703                                         1, &new_blk);
4704                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4705                 }
4706                 if (ctx->invalid_inode_bitmap_flag[i]) {
4707                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4708                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4709                                         1, &new_blk);
4710                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4711                 }
4712                 if (ctx->invalid_inode_table_flag[i]) {
4713                         new_blk = ext2fs_inode_table_loc(fs, i);
4714                         new_table_block(ctx, first_block, i, _("inode table"),
4715                                         fs->inode_blocks_per_group,
4716                                         &new_blk);
4717                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4718                         ctx->flags |= E2F_FLAG_RESTART;
4719                 }
4720         }
4721         ctx->invalid_bitmaps = 0;
4722 }
4723
4724 /*
4725  * This routine marks all blocks which are used by the superblock,
4726  * group descriptors, inode bitmaps, and block bitmaps.
4727  */
4728 static void mark_table_blocks(e2fsck_t ctx)
4729 {
4730         ext2_filsys fs = ctx->fs;
4731         blk64_t b;
4732         dgrp_t  i;
4733         unsigned int    j;
4734         struct problem_context pctx;
4735
4736         clear_problem_context(&pctx);
4737
4738         for (i = 0; i < fs->group_desc_count; i++) {
4739                 pctx.group = i;
4740
4741                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4742                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4743
4744                 /*
4745                  * Mark the blocks used for the inode table
4746                  */
4747                 if (ext2fs_inode_table_loc(fs, i)) {
4748                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4749                              j < fs->inode_blocks_per_group;
4750                              j++, b++) {
4751                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4752                                                              b)) {
4753                                         pctx.blk = b;
4754                                         if (!ctx->invalid_inode_table_flag[i] &&
4755                                             fix_problem(ctx,
4756                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4757                                                 ctx->invalid_inode_table_flag[i]++;
4758                                                 ctx->invalid_bitmaps++;
4759                                         }
4760                                 } else {
4761                                     ext2fs_mark_block_bitmap2(
4762                                                 ctx->block_found_map, b);
4763                                     ext2fs_mark_block_bitmap2(
4764                                                 ctx->block_metadata_map, b);
4765                                 }
4766                         }
4767                 }
4768
4769                 /*
4770                  * Mark block used for the block bitmap
4771                  */
4772                 if (ext2fs_block_bitmap_loc(fs, i)) {
4773                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4774                                      ext2fs_block_bitmap_loc(fs, i))) {
4775                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4776                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4777                                         ctx->invalid_block_bitmap_flag[i]++;
4778                                         ctx->invalid_bitmaps++;
4779                                 }
4780                         } else {
4781                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4782                                      ext2fs_block_bitmap_loc(fs, i));
4783                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4784                                      ext2fs_block_bitmap_loc(fs, i));
4785                         }
4786                 }
4787                 /*
4788                  * Mark block used for the inode bitmap
4789                  */
4790                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4791                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4792                                      ext2fs_inode_bitmap_loc(fs, i))) {
4793                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4794                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4795                                         ctx->invalid_inode_bitmap_flag[i]++;
4796                                         ctx->invalid_bitmaps++;
4797                                 }
4798                         } else {
4799                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4800                                      ext2fs_inode_bitmap_loc(fs, i));
4801                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4802                                      ext2fs_inode_bitmap_loc(fs, i));
4803                         }
4804                 }
4805         }
4806 }
4807
4808 /*
4809  * These subroutines short circuits ext2fs_get_blocks and
4810  * ext2fs_check_directory; we use them since we already have the inode
4811  * structure, so there's no point in letting the ext2fs library read
4812  * the inode again.
4813  */
4814 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4815                                   blk_t *blocks)
4816 {
4817         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4818         int     i;
4819
4820         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4821                 return EXT2_ET_CALLBACK_NOTHANDLED;
4822
4823         for (i=0; i < EXT2_N_BLOCKS; i++)
4824                 blocks[i] = ctx->stashed_inode->i_block[i];
4825         return 0;
4826 }
4827
4828 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4829                                   struct ext2_inode *inode)
4830 {
4831         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4832
4833         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4834                 return EXT2_ET_CALLBACK_NOTHANDLED;
4835         *inode = *ctx->stashed_inode;
4836         return 0;
4837 }
4838
4839 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4840                             struct ext2_inode *inode)
4841 {
4842         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4843
4844         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4845                 (inode != ctx->stashed_inode))
4846                 *ctx->stashed_inode = *inode;
4847         return EXT2_ET_CALLBACK_NOTHANDLED;
4848 }
4849
4850 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4851 {
4852         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4853
4854         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4855                 return EXT2_ET_CALLBACK_NOTHANDLED;
4856
4857         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4858                 return EXT2_ET_NO_DIRECTORY;
4859         return 0;
4860 }
4861
4862 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4863                                         blk64_t *ret)
4864 {
4865         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4866         errcode_t       retval;
4867         blk64_t         new_block;
4868
4869         if (ctx->block_found_map) {
4870                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4871                                            &new_block);
4872                 if (retval)
4873                         return retval;
4874                 if (fs->block_map) {
4875                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4876                         ext2fs_mark_bb_dirty(fs);
4877                 }
4878         } else {
4879                 if (!fs->block_map) {
4880                         retval = ext2fs_read_block_bitmap(fs);
4881                         if (retval)
4882                                 return retval;
4883                 }
4884
4885                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4886                 if (retval)
4887                         return retval;
4888         }
4889
4890         *ret = new_block;
4891         return (0);
4892 }
4893
4894 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4895                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4896 {
4897         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4898         errcode_t       retval;
4899
4900         if (ctx->block_found_map)
4901                 return ext2fs_new_range(fs, flags, goal, len,
4902                                         ctx->block_found_map, pblk, plen);
4903
4904         if (!fs->block_map) {
4905                 retval = ext2fs_read_block_bitmap(fs);
4906                 if (retval)
4907                         return retval;
4908         }
4909
4910         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4911                                 pblk, plen);
4912 }
4913
4914 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4915 {
4916         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4917
4918         /* Never free a critical metadata block */
4919         if (ctx->block_found_map &&
4920             ctx->block_metadata_map &&
4921             inuse < 0 &&
4922             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4923                 return;
4924
4925         if (ctx->block_found_map) {
4926                 if (inuse > 0)
4927                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4928                 else
4929                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4930         }
4931 }
4932
4933 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4934                                            blk_t num, int inuse)
4935 {
4936         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4937
4938         /* Never free a critical metadata block */
4939         if (ctx->block_found_map &&
4940             ctx->block_metadata_map &&
4941             inuse < 0 &&
4942             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4943                 return;
4944
4945         if (ctx->block_found_map) {
4946                 if (inuse > 0)
4947                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4948                                                         blk, num);
4949                 else
4950                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4951                                                         blk, num);
4952         }
4953 }
4954
4955 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4956 {
4957         ext2_filsys fs = ctx->fs;
4958
4959         if (use_shortcuts) {
4960                 fs->get_blocks = pass1_get_blocks;
4961                 fs->check_directory = pass1_check_directory;
4962                 fs->read_inode = pass1_read_inode;
4963                 fs->write_inode = pass1_write_inode;
4964                 ctx->stashed_ino = 0;
4965         } else {
4966                 fs->get_blocks = 0;
4967                 fs->check_directory = 0;
4968                 fs->read_inode = 0;
4969                 fs->write_inode = 0;
4970         }
4971 }
4972
4973 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4974 {
4975         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4976         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4977                                                 e2fsck_block_alloc_stats, 0);
4978         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4979         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4980                                         e2fsck_block_alloc_stats_range, NULL);
4981 }