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 blocks are in use.          (block_found_map)
  27  *      - A bitmap of which blocks are in use by two inodes     (block_dup_map)
  28  *      - The data blocks of the directory inodes.      (dir_map)
  29  *      - Ref counts for ea_inodes.                     (ea_inode_refs)
  30  *
  31  * Pass 1 is designed to stash away enough information so that the
  32  * other passes should not need to read in the inode information
  33  * during the normal course of a filesystem check.  (Although if an
  34  * inconsistency is detected, other passes may need to read in an
  35  * inode to fix it.)
  36  *
  37  * Note that pass 1B will be invoked if there are any duplicate blocks
  38  * found.
  39  */
  40
  41 #define _GNU_SOURCE 1 /* get strnlen() */
  42 #include "config.h"
  43 #include <string.h>
  44 #include <time.h>
  45 #ifdef HAVE_ERRNO_H
  46 #include <errno.h>
  47 #endif
  48
  49 #include "e2fsck.h"
  50 #include <ext2fs/ext2_ext_attr.h>
  51 #include <e2p/e2p.h>
  52
  53 #include "problem.h"
  54
  55 #ifdef NO_INLINE_FUNCS
  56 #define _INLINE_
  57 #else
  58 #define _INLINE_ inline
  59 #endif
  60
  61 #undef DEBUG
  62
  63 struct ea_quota {
  64         blk64_t blocks;
  65         __u64 inodes;
  66 };
  67
  68 static int process_block(ext2_filsys fs, blk64_t        *blocknr,
  69                          e2_blkcnt_t blockcnt, blk64_t ref_blk,
  70                          int ref_offset, void *priv_data);
  71 static int process_bad_block(ext2_filsys fs, blk64_t *block_nr,
  72                              e2_blkcnt_t blockcnt, blk64_t ref_blk,
  73                              int ref_offset, void *priv_data);
  74 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
  75                          char *block_buf,
  76                          const struct ea_quota *ea_ibody_quota);
  77 static void mark_table_blocks(e2fsck_t ctx);
  78 static void alloc_bb_map(e2fsck_t ctx);
  79 static void alloc_imagic_map(e2fsck_t ctx);
  80 static void mark_inode_bad(e2fsck_t ctx, ino_t ino);
  81 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino);
  82 static void handle_fs_bad_blocks(e2fsck_t ctx);
  83 static void process_inodes(e2fsck_t ctx, char *block_buf);
  84 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b);
  85 static errcode_t scan_callback(ext2_filsys fs, ext2_inode_scan scan,
  86                                   dgrp_t group, void * priv_data);
  87 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
  88                                     char *block_buf, int adjust_sign);
  89 /* static char *describe_illegal_block(ext2_filsys fs, blk64_t block); */
  90
  91 struct process_block_struct {
  92         ext2_ino_t      ino;
  93         unsigned        is_dir:1, is_reg:1, clear:1, suppress:1,
  94                                 fragmented:1, compressed:1, bbcheck:1,
  95                                 inode_modified:1;
  96         blk64_t         num_blocks;
  97         blk64_t         max_blocks;
  98         blk64_t         last_block;
  99         e2_blkcnt_t     last_init_lblock;
 100         e2_blkcnt_t     last_db_block;
 101         int             num_illegal_blocks;
 102         blk64_t         previous_block;
 103         struct ext2_inode *inode;
 104         struct problem_context *pctx;
 105         ext2fs_block_bitmap fs_meta_blocks;
 106         e2fsck_t        ctx;
 107         blk64_t         next_lblock;
 108         struct extent_tree_info eti;
 109 };
 110
 111 struct process_inode_block {
 112         ext2_ino_t ino;
 113         struct ea_quota ea_ibody_quota;
 114         struct ext2_inode_large inode;
 115 };
 116
 117 struct scan_callback_struct {
 118         e2fsck_t        ctx;
 119         char            *block_buf;
 120 };
 121
 122 /*
 123  * For the inodes to process list.
 124  */
 125 static struct process_inode_block *inodes_to_process;
 126 static int process_inode_count;
 127
 128 static __u64 ext2_max_sizes[EXT2_MAX_BLOCK_LOG_SIZE -
 129                             EXT2_MIN_BLOCK_LOG_SIZE + 1];
 130
 131 /*
 132  * Free all memory allocated by pass1 in preparation for restarting
 133  * things.
 134  */
 135 static void unwind_pass1(ext2_filsys fs EXT2FS_ATTR((unused)))
 136 {
 137         ext2fs_free_mem(&inodes_to_process);
 138         inodes_to_process = 0;
 139 }
 140
 141 /*
 142  * Check to make sure a device inode is real.  Returns 1 if the device
 143  * checks out, 0 if not.
 144  *
 145  * Note: this routine is now also used to check FIFO's and Sockets,
 146  * since they have the same requirement; the i_block fields should be
 147  * zero.
 148  */
 149 int e2fsck_pass1_check_device_inode(ext2_filsys fs EXT2FS_ATTR((unused)),
 150                                     struct ext2_inode *inode)
 151 {
 152         int     i;
 153
 154         /*
 155          * If the index or extents flag is set, then this is a bogus
 156          * device/fifo/socket
 157          */
 158         if (inode->i_flags & (EXT2_INDEX_FL | EXT4_EXTENTS_FL))
 159                 return 0;
 160
 161         /*
 162          * We should be able to do the test below all the time, but
 163          * because the kernel doesn't forcibly clear the device
 164          * inode's additional i_block fields, there are some rare
 165          * occasions when a legitimate device inode will have non-zero
 166          * additional i_block fields.  So for now, we only complain
 167          * when the immutable flag is set, which should never happen
 168          * for devices.  (And that's when the problem is caused, since
 169          * you can't set or clear immutable flags for devices.)  Once
 170          * the kernel has been fixed we can change this...
 171          */
 172         if (inode->i_flags & (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)) {
 173                 for (i=4; i < EXT2_N_BLOCKS; i++)
 174                         if (inode->i_block[i])
 175                                 return 0;
 176         }
 177         return 1;
 178 }
 179
 180 /*
 181  * Check to make sure a symlink inode is real.  Returns 1 if the symlink
 182  * checks out, 0 if not.
 183  */
 184 int e2fsck_pass1_check_symlink(ext2_filsys fs, ext2_ino_t ino,
 185                                struct ext2_inode *inode, char *buf)
 186 {
 187         unsigned int buflen;
 188         unsigned int len;
 189
 190         if ((inode->i_size_high || inode->i_size == 0) ||
 191             (inode->i_flags & EXT2_INDEX_FL))
 192                 return 0;
 193
 194         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
 195                 size_t inline_size;
 196
 197                 if (inode->i_flags & EXT4_EXTENTS_FL)
 198                         return 0;
 199                 if (ext2fs_inline_data_size(fs, ino, &inline_size))
 200                         return 0;
 201                 if (inode->i_size != inline_size)
 202                         return 0;
 203
 204                 return 1;
 205         }
 206
 207         if (ext2fs_is_fast_symlink(inode)) {
 208                 if (inode->i_flags & EXT4_EXTENTS_FL)
 209                         return 0;
 210                 buf = (char *)inode->i_block;
 211                 buflen = sizeof(inode->i_block);
 212         } else {
 213                 ext2_extent_handle_t    handle;
 214                 struct ext2_extent_info info;
 215                 struct ext2fs_extent    extent;
 216                 blk64_t blk;
 217                 int i;
 218
 219                 if (inode->i_flags & EXT4_EXTENTS_FL) {
 220                         if (ext2fs_extent_open2(fs, ino, inode, &handle))
 221                                 return 0;
 222                         if (ext2fs_extent_get_info(handle, &info) ||
 223                             (info.num_entries != 1) ||
 224                             (info.max_depth != 0)) {
 225                                 ext2fs_extent_free(handle);
 226                                 return 0;
 227                         }
 228                         if (ext2fs_extent_get(handle, EXT2_EXTENT_ROOT,
 229                                               &extent) ||
 230                             (extent.e_lblk != 0) ||
 231                             (extent.e_len != 1)) {
 232                                 ext2fs_extent_free(handle);
 233                                 return 0;
 234                         }
 235                         blk = extent.e_pblk;
 236                         ext2fs_extent_free(handle);
 237                 } else {
 238                         blk = inode->i_block[0];
 239
 240                         for (i = 1; i < EXT2_N_BLOCKS; i++)
 241                                 if (inode->i_block[i])
 242                                         return 0;
 243                 }
 244
 245                 if (blk < fs->super->s_first_data_block ||
 246                     blk >= ext2fs_blocks_count(fs->super))
 247                         return 0;
 248
 249                 if (io_channel_read_blk64(fs->io, blk, 1, buf))
 250                         return 0;
 251
 252                 buflen = fs->blocksize;
 253         }
 254
 255         if (inode->i_flags & EXT4_ENCRYPT_FL)
 256                 len = ext2fs_le16_to_cpu(*(__u16 *)buf) + 2;
 257         else
 258                 len = strnlen(buf, buflen);
 259
 260         if (len >= buflen)
 261                 return 0;
 262
 263         if (len != inode->i_size)
 264                 return 0;
 265         return 1;
 266 }
 267
 268 /*
 269  * If the extents or inlinedata flags are set on the inode, offer to clear 'em.
 270  */
 271 #define BAD_SPECIAL_FLAGS (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)
 272 static void check_extents_inlinedata(e2fsck_t ctx,
 273                                      struct problem_context *pctx)
 274 {
 275         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 276                 return;
 277
 278         if (!fix_problem(ctx, PR_1_SPECIAL_EXTENTS_IDATA, pctx))
 279                 return;
 280
 281         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 282         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 283 }
 284 #undef BAD_SPECIAL_FLAGS
 285
 286 /*
 287  * If the immutable (or append-only) flag is set on the inode, offer
 288  * to clear it.
 289  */
 290 #define BAD_SPECIAL_FLAGS (EXT2_IMMUTABLE_FL | EXT2_APPEND_FL)
 291 static void check_immutable(e2fsck_t ctx, struct problem_context *pctx)
 292 {
 293         if (!(pctx->inode->i_flags & BAD_SPECIAL_FLAGS))
 294                 return;
 295
 296         if (!fix_problem(ctx, PR_1_SET_IMMUTABLE, pctx))
 297                 return;
 298
 299         pctx->inode->i_flags &= ~BAD_SPECIAL_FLAGS;
 300         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 301 }
 302
 303 /*
 304  * If device, fifo or socket, check size is zero -- if not offer to
 305  * clear it
 306  */
 307 static void check_size(e2fsck_t ctx, struct problem_context *pctx)
 308 {
 309         struct ext2_inode *inode = pctx->inode;
 310
 311         if (EXT2_I_SIZE(inode) == 0)
 312                 return;
 313
 314         if (!fix_problem(ctx, PR_1_SET_NONZSIZE, pctx))
 315                 return;
 316
 317         ext2fs_inode_size_set(ctx->fs, inode, 0);
 318         e2fsck_write_inode(ctx, pctx->ino, pctx->inode, "pass1");
 319 }
 320
 321 /*
 322  * For a given size, calculate how many blocks would be charged towards quota.
 323  */
 324 static blk64_t size_to_quota_blocks(ext2_filsys fs, size_t size)
 325 {
 326         blk64_t clusters;
 327
 328         clusters = DIV_ROUND_UP(size, fs->blocksize << fs->cluster_ratio_bits);
 329         return EXT2FS_C2B(fs, clusters);
 330 }
 331
 332 /*
 333  * Check validity of EA inode. Return 0 if EA inode is valid, otherwise return
 334  * the problem code.
 335  */
 336 static problem_t check_large_ea_inode(e2fsck_t ctx,
 337                                       struct ext2_ext_attr_entry *entry,
 338                                       struct problem_context *pctx,
 339                                       blk64_t *quota_blocks)
 340 {
 341         struct ext2_inode inode;
 342         __u32 hash;
 343         errcode_t retval;
 344
 345         /* Check if inode is within valid range */
 346         if ((entry->e_value_inum < EXT2_FIRST_INODE(ctx->fs->super)) ||
 347             (entry->e_value_inum > ctx->fs->super->s_inodes_count)) {
 348                 pctx->num = entry->e_value_inum;
 349                 return PR_1_ATTR_VALUE_EA_INODE;
 350         }
 351
 352         e2fsck_read_inode(ctx, entry->e_value_inum, &inode, "pass1");
 353
 354         retval = ext2fs_ext_attr_hash_entry2(ctx->fs, entry, NULL, &hash);
 355         if (retval) {
 356                 com_err("check_large_ea_inode", retval,
 357                         _("while hashing entry with e_value_inum = %u"),
 358                         entry->e_value_inum);
 359                 fatal_error(ctx, 0);
 360         }
 361
 362         if (hash == entry->e_hash) {
 363                 *quota_blocks = size_to_quota_blocks(ctx->fs,
 364                                                      entry->e_value_size);
 365         } else {
 366                 /* This might be an old Lustre-style ea_inode reference. */
 367                 if (inode.i_mtime == pctx->ino &&
 368                     inode.i_generation == pctx->inode->i_generation) {
 369                         *quota_blocks = 0;
 370                 } else {
 371                         /* If target inode is also missing EA_INODE flag,
 372                          * this is likely to be a bad reference.
 373                          */
 374                         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 375                                 pctx->num = entry->e_value_inum;
 376                                 return PR_1_ATTR_VALUE_EA_INODE;
 377                         } else {
 378                                 pctx->num = entry->e_hash;
 379                                 return PR_1_ATTR_HASH;
 380                         }
 381                 }
 382         }
 383
 384         if (!(inode.i_flags & EXT4_EA_INODE_FL)) {
 385                 pctx->num = entry->e_value_inum;
 386                 if (fix_problem(ctx, PR_1_ATTR_SET_EA_INODE_FL, pctx)) {
 387                         inode.i_flags |= EXT4_EA_INODE_FL;
 388                         ext2fs_write_inode(ctx->fs, entry->e_value_inum,
 389                                            &inode);
 390                 } else {
 391                         return PR_1_ATTR_NO_EA_INODE_FL;
 392                 }
 393         }
 394         return 0;
 395 }
 396
 397 static void inc_ea_inode_refs(e2fsck_t ctx, struct problem_context *pctx,
 398                               struct ext2_ext_attr_entry *first, void *end)
 399 {
 400         struct ext2_ext_attr_entry *entry;
 401
 402         for (entry = first;
 403              (void *)entry < end && !EXT2_EXT_IS_LAST_ENTRY(entry);
 404              entry = EXT2_EXT_ATTR_NEXT(entry)) {
 405                 if (!entry->e_value_inum)
 406                         continue;
 407                 if (!ctx->ea_inode_refs) {
 408                         pctx->errcode = ea_refcount_create(0,
 409                                                            &ctx->ea_inode_refs);
 410                         if (pctx->errcode) {
 411                                 pctx->num = 4;
 412                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
 413                                 ctx->flags |= E2F_FLAG_ABORT;
 414                                 return;
 415                         }
 416                 }
 417                 ea_refcount_increment(ctx->ea_inode_refs, entry->e_value_inum,
 418                                       0);
 419         }
 420 }
 421
 422 static void check_ea_in_inode(e2fsck_t ctx, struct problem_context *pctx,
 423                               struct ea_quota *ea_ibody_quota)
 424 {
 425         struct ext2_super_block *sb = ctx->fs->super;
 426         struct ext2_inode_large *inode;
 427         struct ext2_ext_attr_entry *entry;
 428         char *start, *header, *end;
 429         unsigned int storage_size, remain;
 430         problem_t problem = 0;
 431         region_t region = 0;
 432
 433         ea_ibody_quota->blocks = 0;
 434         ea_ibody_quota->inodes = 0;
 435
 436         inode = (struct ext2_inode_large *) pctx->inode;
 437         storage_size = EXT2_INODE_SIZE(ctx->fs->super) - EXT2_GOOD_OLD_INODE_SIZE -
 438                 inode->i_extra_isize;
 439         header = ((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 440                  inode->i_extra_isize;
 441         end = header + storage_size;
 442         start = header + sizeof(__u32);
 443         entry = (struct ext2_ext_attr_entry *) start;
 444
 445         /* scan all entry's headers first */
 446
 447         /* take finish entry 0UL into account */
 448         remain = storage_size - sizeof(__u32);
 449
 450         region = region_create(0, storage_size);
 451         if (!region) {
 452                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
 453                 problem = 0;
 454                 ctx->flags |= E2F_FLAG_ABORT;
 455                 return;
 456         }
 457         if (region_allocate(region, 0, sizeof(__u32))) {
 458                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 459                 goto fix;
 460         }
 461
 462         while (remain >= sizeof(struct ext2_ext_attr_entry) &&
 463                !EXT2_EXT_IS_LAST_ENTRY(entry)) {
 464                 __u32 hash;
 465
 466                 if (region_allocate(region, (char *)entry - (char *)header,
 467                                     EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
 468                         problem = PR_1_INODE_EA_ALLOC_COLLISION;
 469                         goto fix;
 470                 }
 471
 472                 /* header eats this space */
 473                 remain -= sizeof(struct ext2_ext_attr_entry);
 474
 475                 /* is attribute name valid? */
 476                 if (EXT2_EXT_ATTR_SIZE(entry->e_name_len) > remain) {
 477                         pctx->num = entry->e_name_len;
 478                         problem = PR_1_ATTR_NAME_LEN;
 479                         goto fix;
 480                 }
 481
 482                 /* attribute len eats this space */
 483                 remain -= EXT2_EXT_ATTR_SIZE(entry->e_name_len);
 484
 485                 if (entry->e_value_inum == 0) {
 486                         /* check value size */
 487                         if (entry->e_value_size > remain) {
 488                                 pctx->num = entry->e_value_size;
 489                                 problem = PR_1_ATTR_VALUE_SIZE;
 490                                 goto fix;
 491                         }
 492
 493                         if (entry->e_value_size &&
 494                             region_allocate(region,
 495                                             sizeof(__u32) + entry->e_value_offs,
 496                                             EXT2_EXT_ATTR_SIZE(
 497                                                 entry->e_value_size))) {
 498                                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 499                                 goto fix;
 500                         }
 501
 502                         hash = ext2fs_ext_attr_hash_entry(entry,
 503                                                           start + entry->e_value_offs);
 504
 505                         /* e_hash may be 0 in older inode's ea */
 506                         if (entry->e_hash != 0 && entry->e_hash != hash) {
 507                                 pctx->num = entry->e_hash;
 508                                 problem = PR_1_ATTR_HASH;
 509                                 goto fix;
 510                         }
 511                 } else {
 512                         blk64_t quota_blocks;
 513
 514                         problem = check_large_ea_inode(ctx, entry, pctx,
 515                                                        &quota_blocks);
 516                         if (problem != 0)
 517                                 goto fix;
 518
 519                         ea_ibody_quota->blocks += quota_blocks;
 520                         ea_ibody_quota->inodes++;
 521                 }
 522
 523                 /* If EA value is stored in external inode then it does not
 524                  * consume space here */
 525                 if (entry->e_value_inum == 0)
 526                         remain -= entry->e_value_size;
 527
 528                 entry = EXT2_EXT_ATTR_NEXT(entry);
 529         }
 530
 531         if (region_allocate(region, (char *)entry - (char *)header,
 532                             sizeof(__u32))) {
 533                 problem = PR_1_INODE_EA_ALLOC_COLLISION;
 534                 goto fix;
 535         }
 536 fix:
 537         if (region)
 538                 region_free(region);
 539         /*
 540          * it seems like a corruption. it's very unlikely we could repair
 541          * EA(s) in automatic fashion -bzzz
 542          */
 543         if (problem == 0 || !fix_problem(ctx, problem, pctx)) {
 544                 inc_ea_inode_refs(ctx, pctx,
 545                                   (struct ext2_ext_attr_entry *)start, end);
 546                 return;
 547         }
 548
 549         /* simply remove all possible EA(s) */
 550         *((__u32 *)header) = 0UL;
 551         e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 552                                 EXT2_INODE_SIZE(sb), "pass1");
 553         ea_ibody_quota->blocks = 0;
 554         ea_ibody_quota->inodes = 0;
 555 }
 556
 557 static int check_inode_extra_negative_epoch(__u32 xtime, __u32 extra) {
 558         return (xtime & (1U << 31)) != 0 &&
 559                 (extra & EXT4_EPOCH_MASK) == EXT4_EPOCH_MASK;
 560 }
 561
 562 #define CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, xtime) \
 563         check_inode_extra_negative_epoch(inode->i_##xtime, \
 564                                          inode->i_##xtime##_extra)
 565
 566 /* When today's date is earlier than 2242, we assume that atimes,
 567  * ctimes, crtimes, and mtimes with years in the range 2310..2378 are
 568  * actually pre-1970 dates mis-encoded.
 569  */
 570 #define EXT4_EXTRA_NEGATIVE_DATE_CUTOFF 2 * (1LL << 32)
 571
 572 static void check_inode_extra_space(e2fsck_t ctx, struct problem_context *pctx,
 573                                     struct ea_quota *ea_ibody_quota)
 574 {
 575         struct ext2_super_block *sb = ctx->fs->super;
 576         struct ext2_inode_large *inode;
 577         __u32 *eamagic;
 578         int min, max;
 579
 580         ea_ibody_quota->blocks = 0;
 581         ea_ibody_quota->inodes = 0;
 582
 583         inode = (struct ext2_inode_large *) pctx->inode;
 584         if (EXT2_INODE_SIZE(sb) == EXT2_GOOD_OLD_INODE_SIZE) {
 585                 /* this isn't large inode. so, nothing to check */
 586                 return;
 587         }
 588
 589 #if 0
 590         printf("inode #%u, i_extra_size %d\n", pctx->ino,
 591                         inode->i_extra_isize);
 592 #endif
 593         /* i_extra_isize must cover i_extra_isize + i_checksum_hi at least */
 594         min = sizeof(inode->i_extra_isize) + sizeof(inode->i_checksum_hi);
 595         max = EXT2_INODE_SIZE(sb) - EXT2_GOOD_OLD_INODE_SIZE;
 596         /*
 597          * For now we will allow i_extra_isize to be 0, but really
 598          * implementations should never allow i_extra_isize to be 0
 599          */
 600         if (inode->i_extra_isize &&
 601             (inode->i_extra_isize < min || inode->i_extra_isize > max ||
 602              inode->i_extra_isize & 3)) {
 603                 if (!fix_problem(ctx, PR_1_EXTRA_ISIZE, pctx))
 604                         return;
 605                 if (inode->i_extra_isize < min || inode->i_extra_isize > max)
 606                         inode->i_extra_isize = sb->s_want_extra_isize;
 607                 else
 608                         inode->i_extra_isize = (inode->i_extra_isize + 3) & ~3;
 609                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 610                                         EXT2_INODE_SIZE(sb), "pass1");
 611         }
 612
 613         /* check if there is no place for an EA header */
 614         if (inode->i_extra_isize >= max - sizeof(__u32))
 615                 return;
 616
 617         eamagic = (__u32 *) (((char *) inode) + EXT2_GOOD_OLD_INODE_SIZE +
 618                         inode->i_extra_isize);
 619         if (*eamagic == EXT2_EXT_ATTR_MAGIC) {
 620                 /* it seems inode has an extended attribute(s) in body */
 621                 check_ea_in_inode(ctx, pctx, ea_ibody_quota);
 622         }
 623
 624         /*
 625          * If the inode's extended atime (ctime, crtime, mtime) is stored in
 626          * the old, invalid format, repair it.
 627          */
 628         if (((sizeof(time_t) <= 4) ||
 629              (((sizeof(time_t) > 4) &&
 630                ctx->now < EXT4_EXTRA_NEGATIVE_DATE_CUTOFF))) &&
 631             (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime) ||
 632              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime) ||
 633              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime) ||
 634              CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))) {
 635
 636                 if (!fix_problem(ctx, PR_1_EA_TIME_OUT_OF_RANGE, pctx))
 637                         return;
 638
 639                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, atime))
 640                         inode->i_atime_extra &= ~EXT4_EPOCH_MASK;
 641                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, ctime))
 642                         inode->i_ctime_extra &= ~EXT4_EPOCH_MASK;
 643                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, crtime))
 644                         inode->i_crtime_extra &= ~EXT4_EPOCH_MASK;
 645                 if (CHECK_INODE_EXTRA_NEGATIVE_EPOCH(inode, mtime))
 646                         inode->i_mtime_extra &= ~EXT4_EPOCH_MASK;
 647                 e2fsck_write_inode_full(ctx, pctx->ino, pctx->inode,
 648                                         EXT2_INODE_SIZE(sb), "pass1");
 649         }
 650
 651 }
 652
 653 /*
 654  * Check to see if the inode might really be a directory, despite i_mode
 655  *
 656  * This is a lot of complexity for something for which I'm not really
 657  * convinced happens frequently in the wild.  If for any reason this
 658  * causes any problems, take this code out.
 659  * [tytso:20070331.0827EDT]
 660  */
 661 static void check_is_really_dir(e2fsck_t ctx, struct problem_context *pctx,
 662                                 char *buf)
 663 {
 664         struct ext2_inode *inode = pctx->inode;
 665         struct ext2_dir_entry   *dirent;
 666         errcode_t               retval;
 667         blk64_t                 blk;
 668         unsigned int            i, rec_len, not_device = 0;
 669         int                     extent_fs;
 670         int                     inlinedata_fs;
 671
 672         /*
 673          * If the mode looks OK, we believe it.  If the first block in
 674          * the i_block array is 0, this cannot be a directory. If the
 675          * inode is extent-mapped, it is still the case that the latter
 676          * cannot be 0 - the magic number in the extent header would make
 677          * it nonzero.
 678          */
 679         if (LINUX_S_ISDIR(inode->i_mode) || LINUX_S_ISREG(inode->i_mode) ||
 680             LINUX_S_ISLNK(inode->i_mode) || inode->i_block[0] == 0)
 681                 return;
 682
 683         /*
 684          * Check the block numbers in the i_block array for validity:
 685          * zero blocks are skipped (but the first one cannot be zero -
 686          * see above), other blocks are checked against the first and
 687          * max data blocks (from the the superblock) and against the
 688          * block bitmap. Any invalid block found means this cannot be
 689          * a directory.
 690          *
 691          * If there are non-zero blocks past the fourth entry, then
 692          * this cannot be a device file: we remember that for the next
 693          * check.
 694          *
 695          * For extent mapped files, we don't do any sanity checking:
 696          * just try to get the phys block of logical block 0 and run
 697          * with it.
 698          *
 699          * For inline data files, we just try to get the size of inline
 700          * data.  If it's true, we will treat it as a directory.
 701          */
 702
 703         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
 704         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
 705         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL)) {
 706                 size_t size;
 707                 __u32 dotdot;
 708                 unsigned int rec_len2;
 709                 struct ext2_dir_entry de;
 710
 711                 if (ext2fs_inline_data_size(ctx->fs, pctx->ino, &size))
 712                         return;
 713                 /*
 714                  * If the size isn't a multiple of 4, it's probably not a
 715                  * directory??
 716                  */
 717                 if (size & 3)
 718                         return;
 719                 /*
 720                  * If the first 10 bytes don't look like a directory entry,
 721                  * it's probably not a directory.
 722                  */
 723                 memcpy(&dotdot, inode->i_block, sizeof(dotdot));
 724                 memcpy(&de, ((char *)inode->i_block) + EXT4_INLINE_DATA_DOTDOT_SIZE,
 725                        EXT2_DIR_REC_LEN(0));
 726                 dotdot = ext2fs_le32_to_cpu(dotdot);
 727                 de.inode = ext2fs_le32_to_cpu(de.inode);
 728                 de.rec_len = ext2fs_le16_to_cpu(de.rec_len);
 729                 ext2fs_get_rec_len(ctx->fs, &de, &rec_len2);
 730                 if (dotdot >= ctx->fs->super->s_inodes_count ||
 731                     (dotdot < EXT2_FIRST_INO(ctx->fs->super) &&
 732                      dotdot != EXT2_ROOT_INO) ||
 733                     de.inode >= ctx->fs->super->s_inodes_count ||
 734                     (de.inode < EXT2_FIRST_INO(ctx->fs->super) &&
 735                      de.inode != 0) ||
 736                     rec_len2 > EXT4_MIN_INLINE_DATA_SIZE -
 737                               EXT4_INLINE_DATA_DOTDOT_SIZE)
 738                         return;
 739                 /* device files never have a "system.data" entry */
 740                 goto isdir;
 741         } else if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) {
 742                 /* extent mapped */
 743                 if  (ext2fs_bmap2(ctx->fs, pctx->ino, inode, 0, 0, 0, 0,
 744                                  &blk))
 745                         return;
 746                 /* device files are never extent mapped */
 747                 not_device++;
 748         } else {
 749                 for (i=0; i < EXT2_N_BLOCKS; i++) {
 750                         blk = inode->i_block[i];
 751                         if (!blk)
 752                                 continue;
 753                         if (i >= 4)
 754                                 not_device++;
 755
 756                         if (blk < ctx->fs->super->s_first_data_block ||
 757                             blk >= ext2fs_blocks_count(ctx->fs->super) ||
 758                             ext2fs_fast_test_block_bitmap2(ctx->block_found_map,
 759                                                            blk))
 760                                 return; /* Invalid block, can't be dir */
 761                 }
 762                 blk = inode->i_block[0];
 763         }
 764
 765         /*
 766          * If the mode says this is a device file and the i_links_count field
 767          * is sane and we have not ruled it out as a device file previously,
 768          * we declare it a device file, not a directory.
 769          */
 770         if ((LINUX_S_ISCHR(inode->i_mode) || LINUX_S_ISBLK(inode->i_mode)) &&
 771             (inode->i_links_count == 1) && !not_device)
 772                 return;
 773
 774         /* read the first block */
 775         ehandler_operation(_("reading directory block"));
 776         retval = ext2fs_read_dir_block4(ctx->fs, blk, buf, 0, pctx->ino);
 777         ehandler_operation(0);
 778         if (retval)
 779                 return;
 780
 781         dirent = (struct ext2_dir_entry *) buf;
 782         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 783         if (retval)
 784                 return;
 785         if ((ext2fs_dirent_name_len(dirent) != 1) ||
 786             (dirent->name[0] != '.') ||
 787             (dirent->inode != pctx->ino) ||
 788             (rec_len < 12) ||
 789             (rec_len % 4) ||
 790             (rec_len >= ctx->fs->blocksize - 12))
 791                 return;
 792
 793         dirent = (struct ext2_dir_entry *) (buf + rec_len);
 794         retval = ext2fs_get_rec_len(ctx->fs, dirent, &rec_len);
 795         if (retval)
 796                 return;
 797         if ((ext2fs_dirent_name_len(dirent) != 2) ||
 798             (dirent->name[0] != '.') ||
 799             (dirent->name[1] != '.') ||
 800             (rec_len < 12) ||
 801             (rec_len % 4))
 802                 return;
 803
 804 isdir:
 805         if (fix_problem(ctx, PR_1_TREAT_AS_DIRECTORY, pctx)) {
 806                 inode->i_mode = (inode->i_mode & 07777) | LINUX_S_IFDIR;
 807                 e2fsck_write_inode_full(ctx, pctx->ino, inode,
 808                                         EXT2_INODE_SIZE(ctx->fs->super),
 809                                         "check_is_really_dir");
 810         }
 811 }
 812
 813 extern errcode_t e2fsck_setup_icount(e2fsck_t ctx, const char *icount_name,
 814                                      int flags, ext2_icount_t hint,
 815                                      ext2_icount_t *ret)
 816 {
 817         unsigned int            threshold;
 818         unsigned int            save_type;
 819         ext2_ino_t              num_dirs;
 820         errcode_t               retval;
 821         char                    *tdb_dir;
 822         int                     enable;
 823
 824         *ret = 0;
 825
 826         profile_get_string(ctx->profile, "scratch_files", "directory", 0, 0,
 827                            &tdb_dir);
 828         profile_get_uint(ctx->profile, "scratch_files",
 829                          "numdirs_threshold", 0, 0, &threshold);
 830         profile_get_boolean(ctx->profile, "scratch_files",
 831                             "icount", 0, 1, &enable);
 832
 833         retval = ext2fs_get_num_dirs(ctx->fs, &num_dirs);
 834         if (retval)
 835                 num_dirs = 1024;        /* Guess */
 836
 837         if (enable && tdb_dir && !access(tdb_dir, W_OK) &&
 838             (!threshold || num_dirs > threshold)) {
 839                 retval = ext2fs_create_icount_tdb(ctx->fs, tdb_dir,
 840                                                   flags, ret);
 841                 if (retval == 0)
 842                         return 0;
 843         }
 844         e2fsck_set_bitmap_type(ctx->fs, EXT2FS_BMAP64_RBTREE, icount_name,
 845                                &save_type);
 846         if (ctx->options & E2F_OPT_ICOUNT_FULLMAP)
 847                 flags |= EXT2_ICOUNT_OPT_FULLMAP;
 848         retval = ext2fs_create_icount2(ctx->fs, flags, 0, hint, ret);
 849         ctx->fs->default_bitmap_type = save_type;
 850         return retval;
 851 }
 852
 853 static errcode_t recheck_bad_inode_checksum(ext2_filsys fs, ext2_ino_t ino,
 854                                             e2fsck_t ctx,
 855                                             struct problem_context *pctx)
 856 {
 857         errcode_t retval;
 858         struct ext2_inode_large inode;
 859
 860         /*
 861          * Reread inode.  If we don't see checksum error, then this inode
 862          * has been fixed elsewhere.
 863          */
 864         ctx->stashed_ino = 0;
 865         retval = ext2fs_read_inode_full(fs, ino, (struct ext2_inode *)&inode,
 866                                         sizeof(inode));
 867         if (retval && retval != EXT2_ET_INODE_CSUM_INVALID)
 868                 return retval;
 869         if (!retval)
 870                 return 0;
 871
 872         /*
 873          * Checksum still doesn't match.  That implies that the inode passes
 874          * all the sanity checks, so maybe the checksum is simply corrupt.
 875          * See if the user will go for fixing that.
 876          */
 877         if (!fix_problem(ctx, PR_1_INODE_ONLY_CSUM_INVALID, pctx))
 878                 return 0;
 879
 880         retval = ext2fs_write_inode_full(fs, ino, (struct ext2_inode *)&inode,
 881                                          sizeof(inode));
 882         return retval;
 883 }
 884
 885 static void reserve_block_for_root_repair(e2fsck_t ctx)
 886 {
 887         blk64_t         blk = 0;
 888         errcode_t       err;
 889         ext2_filsys     fs = ctx->fs;
 890
 891         ctx->root_repair_block = 0;
 892         if (ext2fs_test_inode_bitmap2(ctx->inode_used_map, EXT2_ROOT_INO))
 893                 return;
 894
 895         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 896         if (err)
 897                 return;
 898         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 899         ctx->root_repair_block = blk;
 900 }
 901
 902 static void reserve_block_for_lnf_repair(e2fsck_t ctx)
 903 {
 904         blk64_t         blk = 0;
 905         errcode_t       err;
 906         ext2_filsys     fs = ctx->fs;
 907         static const char name[] = "lost+found";
 908         ext2_ino_t      ino;
 909
 910         ctx->lnf_repair_block = 0;
 911         if (!ext2fs_lookup(fs, EXT2_ROOT_INO, name, sizeof(name)-1, 0, &ino))
 912                 return;
 913
 914         err = ext2fs_new_block2(fs, 0, ctx->block_found_map, &blk);
 915         if (err)
 916                 return;
 917         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
 918         ctx->lnf_repair_block = blk;
 919 }
 920
 921 static errcode_t get_inline_data_ea_size(ext2_filsys fs, ext2_ino_t ino,
 922                                          size_t *sz)
 923 {
 924         void *p;
 925         struct ext2_xattr_handle *handle;
 926         errcode_t retval;
 927
 928         retval = ext2fs_xattrs_open(fs, ino, &handle);
 929         if (retval)
 930                 return retval;
 931
 932         retval = ext2fs_xattrs_read(handle);
 933         if (retval)
 934                 goto err;
 935
 936         retval = ext2fs_xattr_get(handle, "system.data", &p, sz);
 937         if (retval)
 938                 goto err;
 939         ext2fs_free_mem(&p);
 940 err:
 941         (void) ext2fs_xattrs_close(&handle);
 942         return retval;
 943 }
 944
 945 static void finish_processing_inode(e2fsck_t ctx, ext2_ino_t ino,
 946                                     struct problem_context *pctx,
 947                                     int failed_csum)
 948 {
 949         if (!failed_csum)
 950                 return;
 951
 952         /*
 953          * If the inode failed the checksum and the user didn't
 954          * clear the inode, test the checksum again -- if it still
 955          * fails, ask the user if the checksum should be corrected.
 956          */
 957         pctx->errcode = recheck_bad_inode_checksum(ctx->fs, ino, ctx, pctx);
 958         if (pctx->errcode)
 959                 ctx->flags |= E2F_FLAG_ABORT;
 960 }
 961 #define FINISH_INODE_LOOP(ctx, ino, pctx, failed_csum) \
 962         do { \
 963                 finish_processing_inode((ctx), (ino), (pctx), (failed_csum)); \
 964                 if ((ctx)->flags & E2F_FLAG_ABORT) \
 965                         return; \
 966         } while (0)
 967
 968 static int could_be_block_map(ext2_filsys fs, struct ext2_inode *inode)
 969 {
 970         __u32 x;
 971         int i;
 972
 973         for (i = 0; i < EXT2_N_BLOCKS; i++) {
 974                 x = inode->i_block[i];
 975 #ifdef WORDS_BIGENDIAN
 976                 x = ext2fs_swab32(x);
 977 #endif
 978                 if (x >= ext2fs_blocks_count(fs->super))
 979                         return 0;
 980         }
 981
 982         return 1;
 983 }
 984
 985 /*
 986  * Figure out what to do with an inode that has both extents and inline data
 987  * inode flags set.  Returns -1 if we decide to erase the inode, 0 otherwise.
 988  */
 989 static int fix_inline_data_extents_file(e2fsck_t ctx,
 990                                         ext2_ino_t ino,
 991                                         struct ext2_inode *inode,
 992                                         int inode_size,
 993                                         struct problem_context *pctx)
 994 {
 995         size_t max_inline_ea_size;
 996         ext2_filsys fs = ctx->fs;
 997         int dirty = 0;
 998
 999         /* Both feature flags not set?  Just run the regular checks */
1000         if (!ext2fs_has_feature_extents(fs->super) &&
1001             !ext2fs_has_feature_inline_data(fs->super))
1002                 return 0;
1003
1004         /* Clear both flags if it's a special file */
1005         if (LINUX_S_ISCHR(inode->i_mode) ||
1006             LINUX_S_ISBLK(inode->i_mode) ||
1007             LINUX_S_ISFIFO(inode->i_mode) ||
1008             LINUX_S_ISSOCK(inode->i_mode)) {
1009                 check_extents_inlinedata(ctx, pctx);
1010                 return 0;
1011         }
1012
1013         /* If it looks like an extent tree, try to clear inlinedata */
1014         if (ext2fs_extent_header_verify(inode->i_block,
1015                                  sizeof(inode->i_block)) == 0 &&
1016             fix_problem(ctx, PR_1_CLEAR_INLINE_DATA_FOR_EXTENT, pctx)) {
1017                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1018                 dirty = 1;
1019                 goto out;
1020         }
1021
1022         /* If it looks short enough to be inline data, try to clear extents */
1023         if (inode_size > EXT2_GOOD_OLD_INODE_SIZE)
1024                 max_inline_ea_size = inode_size -
1025                                      (EXT2_GOOD_OLD_INODE_SIZE +
1026                                       ((struct ext2_inode_large *)inode)->i_extra_isize);
1027         else
1028                 max_inline_ea_size = 0;
1029         if (EXT2_I_SIZE(inode) <
1030             EXT4_MIN_INLINE_DATA_SIZE + max_inline_ea_size &&
1031             fix_problem(ctx, PR_1_CLEAR_EXTENT_FOR_INLINE_DATA, pctx)) {
1032                 inode->i_flags &= ~EXT4_EXTENTS_FL;
1033                 dirty = 1;
1034                 goto out;
1035         }
1036
1037         /*
1038          * Too big for inline data, but no evidence of extent tree -
1039          * maybe it's a block map file?  If the mappings all look valid?
1040          */
1041         if (could_be_block_map(fs, inode) &&
1042             fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_FLAGS, pctx)) {
1043 #ifdef WORDS_BIGENDIAN
1044                 int i;
1045
1046                 for (i = 0; i < EXT2_N_BLOCKS; i++)
1047                         inode->i_block[i] = ext2fs_swab32(inode->i_block[i]);
1048 #endif
1049
1050                 inode->i_flags &= ~(EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL);
1051                 dirty = 1;
1052                 goto out;
1053         }
1054
1055         /* Oh well, just clear the busted inode. */
1056         if (fix_problem(ctx, PR_1_CLEAR_EXTENT_INLINE_DATA_INODE, pctx)) {
1057                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1058                 return -1;
1059         }
1060
1061 out:
1062         if (dirty)
1063                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1064
1065         return 0;
1066 }
1067
1068 static void pass1_readahead(e2fsck_t ctx, dgrp_t *group, ext2_ino_t *next_ino)
1069 {
1070         ext2_ino_t inodes_in_group = 0, inodes_per_block, inodes_per_buffer;
1071         dgrp_t start = *group, grp;
1072         blk64_t blocks_to_read = 0;
1073         errcode_t err = EXT2_ET_INVALID_ARGUMENT;
1074
1075         if (ctx->readahead_kb == 0)
1076                 goto out;
1077
1078         /* Keep iterating groups until we have enough to readahead */
1079         inodes_per_block = EXT2_INODES_PER_BLOCK(ctx->fs->super);
1080         for (grp = start; grp < ctx->fs->group_desc_count; grp++) {
1081                 if (ext2fs_bg_flags_test(ctx->fs, grp, EXT2_BG_INODE_UNINIT))
1082                         continue;
1083                 inodes_in_group = ctx->fs->super->s_inodes_per_group -
1084                                         ext2fs_bg_itable_unused(ctx->fs, grp);
1085                 blocks_to_read += (inodes_in_group + inodes_per_block - 1) /
1086                                         inodes_per_block;
1087                 if (blocks_to_read * ctx->fs->blocksize >
1088                     ctx->readahead_kb * 1024)
1089                         break;
1090         }
1091
1092         err = e2fsck_readahead(ctx->fs, E2FSCK_READA_ITABLE, start,
1093                                grp - start + 1);
1094         if (err == EAGAIN) {
1095                 ctx->readahead_kb /= 2;
1096                 err = 0;
1097         }
1098
1099 out:
1100         if (err) {
1101                 /* Error; disable itable readahead */
1102                 *group = ctx->fs->group_desc_count;
1103                 *next_ino = ctx->fs->super->s_inodes_count;
1104         } else {
1105                 /*
1106                  * Don't do more readahead until we've reached the first inode
1107                  * of the last inode scan buffer block for the last group.
1108                  */
1109                 *group = grp + 1;
1110                 inodes_per_buffer = (ctx->inode_buffer_blocks ?
1111                                      ctx->inode_buffer_blocks :
1112                                      EXT2_INODE_SCAN_DEFAULT_BUFFER_BLOCKS) *
1113                                     ctx->fs->blocksize /
1114                                     EXT2_INODE_SIZE(ctx->fs->super);
1115                 inodes_in_group--;
1116                 *next_ino = inodes_in_group -
1117                             (inodes_in_group % inodes_per_buffer) + 1 +
1118                             (grp * ctx->fs->super->s_inodes_per_group);
1119         }
1120 }
1121
1122 /*
1123  * Check if the passed ino is one of the used superblock quota inodes.
1124  *
1125  * Before the quota inodes were journaled, older superblock quota inodes
1126  * were just regular files in the filesystem and not reserved inodes.  This
1127  * checks if the passed ino is one of the s_*_quota_inum superblock fields,
1128  * which may not always be the same as the EXT4_*_QUOTA_INO fields.
1129  */
1130 static int quota_inum_is_super(struct ext2_super_block *sb, ext2_ino_t ino)
1131 {
1132         enum quota_type qtype;
1133
1134         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1135                 if (*quota_sb_inump(sb, qtype) == ino)
1136                         return 1;
1137
1138         return 0;
1139 }
1140
1141 /*
1142  * Check if the passed ino is one of the reserved quota inodes.
1143  * This checks if the inode number is one of the reserved EXT4_*_QUOTA_INO
1144  * inodes.  These inodes may or may not be in use by the quota feature.
1145  */
1146 static int quota_inum_is_reserved(ext2_filsys fs, ext2_ino_t ino)
1147 {
1148         enum quota_type qtype;
1149
1150         for (qtype = 0; qtype < MAXQUOTAS; qtype++)
1151                 if (quota_type2inum(qtype, fs->super) == ino)
1152                         return 1;
1153
1154         return 0;
1155 }
1156
1157 void e2fsck_pass1(e2fsck_t ctx)
1158 {
1159         int     i;
1160         __u64   max_sizes;
1161         ext2_filsys fs = ctx->fs;
1162         ext2_ino_t      ino = 0;
1163         struct ext2_inode *inode = NULL;
1164         ext2_inode_scan scan = NULL;
1165         char            *block_buf = NULL;
1166 #ifdef RESOURCE_TRACK
1167         struct resource_track   rtrack;
1168 #endif
1169         unsigned char   frag, fsize;
1170         struct          problem_context pctx;
1171         struct          scan_callback_struct scan_struct;
1172         struct ext2_super_block *sb = ctx->fs->super;
1173         const char      *old_op;
1174         const char      *eop_next_inode = _("getting next inode from scan");
1175         int             imagic_fs, extent_fs, inlinedata_fs, casefold_fs;
1176         int             low_dtime_check = 1;
1177         unsigned int    inode_size = EXT2_INODE_SIZE(fs->super);
1178         unsigned int    bufsize;
1179         int             failed_csum = 0;
1180         ext2_ino_t      ino_threshold = 0;
1181         dgrp_t          ra_group = 0;
1182         struct ea_quota ea_ibody_quota;
1183
1184         init_resource_track(&rtrack, ctx->fs->io);
1185         clear_problem_context(&pctx);
1186
1187         /* If we can do readahead, figure out how many groups to pull in. */
1188         if (!e2fsck_can_readahead(ctx->fs))
1189                 ctx->readahead_kb = 0;
1190         else if (ctx->readahead_kb == ~0ULL)
1191                 ctx->readahead_kb = e2fsck_guess_readahead(ctx->fs);
1192         pass1_readahead(ctx, &ra_group, &ino_threshold);
1193
1194         if (!(ctx->options & E2F_OPT_PREEN))
1195                 fix_problem(ctx, PR_1_PASS_HEADER, &pctx);
1196
1197         if (ext2fs_has_feature_dir_index(fs->super) &&
1198             !(ctx->options & E2F_OPT_NO)) {
1199                 if (ext2fs_u32_list_create(&ctx->dirs_to_hash, 50))
1200                         ctx->dirs_to_hash = 0;
1201         }
1202
1203 #ifdef MTRACE
1204         mtrace_print("Pass 1");
1205 #endif
1206
1207 #define EXT2_BPP(bits) (1ULL << ((bits) - 2))
1208
1209         for (i = EXT2_MIN_BLOCK_LOG_SIZE; i <= EXT2_MAX_BLOCK_LOG_SIZE; i++) {
1210                 max_sizes = EXT2_NDIR_BLOCKS + EXT2_BPP(i);
1211                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i);
1212                 max_sizes = max_sizes + EXT2_BPP(i) * EXT2_BPP(i) * EXT2_BPP(i);
1213                 max_sizes = (max_sizes * (1UL << i));
1214                 ext2_max_sizes[i - EXT2_MIN_BLOCK_LOG_SIZE] = max_sizes;
1215         }
1216 #undef EXT2_BPP
1217
1218         imagic_fs = ext2fs_has_feature_imagic_inodes(sb);
1219         extent_fs = ext2fs_has_feature_extents(sb);
1220         inlinedata_fs = ext2fs_has_feature_inline_data(sb);
1221         casefold_fs = ext2fs_has_feature_casefold(sb);
1222
1223         /*
1224          * Allocate bitmaps structures
1225          */
1226         pctx.errcode = e2fsck_allocate_inode_bitmap(fs, _("in-use inode map"),
1227                                                     EXT2FS_BMAP64_RBTREE,
1228                                                     "inode_used_map",
1229                                                     &ctx->inode_used_map);
1230         if (pctx.errcode) {
1231                 pctx.num = 1;
1232                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1233                 ctx->flags |= E2F_FLAG_ABORT;
1234                 return;
1235         }
1236         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1237                         _("directory inode map"),
1238                         EXT2FS_BMAP64_AUTODIR,
1239                         "inode_dir_map", &ctx->inode_dir_map);
1240         if (pctx.errcode) {
1241                 pctx.num = 2;
1242                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1243                 ctx->flags |= E2F_FLAG_ABORT;
1244                 return;
1245         }
1246         pctx.errcode = e2fsck_allocate_inode_bitmap(fs,
1247                         _("regular file inode map"), EXT2FS_BMAP64_RBTREE,
1248                         "inode_reg_map", &ctx->inode_reg_map);
1249         if (pctx.errcode) {
1250                 pctx.num = 6;
1251                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
1252                 ctx->flags |= E2F_FLAG_ABORT;
1253                 return;
1254         }
1255         pctx.errcode = e2fsck_allocate_subcluster_bitmap(fs,
1256                         _("in-use block map"), EXT2FS_BMAP64_RBTREE,
1257                         "block_found_map", &ctx->block_found_map);
1258         if (pctx.errcode) {
1259                 pctx.num = 1;
1260                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1261                 ctx->flags |= E2F_FLAG_ABORT;
1262                 return;
1263         }
1264         pctx.errcode = e2fsck_allocate_block_bitmap(fs,
1265                         _("metadata block map"), EXT2FS_BMAP64_RBTREE,
1266                         "block_metadata_map", &ctx->block_metadata_map);
1267         if (pctx.errcode) {
1268                 pctx.num = 1;
1269                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1270                 ctx->flags |= E2F_FLAG_ABORT;
1271                 return;
1272         }
1273         pctx.errcode = e2fsck_setup_icount(ctx, "inode_link_info", 0, NULL,
1274                                            &ctx->inode_link_info);
1275         if (pctx.errcode) {
1276                 fix_problem(ctx, PR_1_ALLOCATE_ICOUNT, &pctx);
1277                 ctx->flags |= E2F_FLAG_ABORT;
1278                 return;
1279         }
1280         bufsize = inode_size;
1281         if (bufsize < sizeof(struct ext2_inode_large))
1282                 bufsize = sizeof(struct ext2_inode_large);
1283         inode = (struct ext2_inode *)
1284                 e2fsck_allocate_memory(ctx, bufsize, "scratch inode");
1285
1286         inodes_to_process = (struct process_inode_block *)
1287                 e2fsck_allocate_memory(ctx,
1288                                        (ctx->process_inode_size *
1289                                         sizeof(struct process_inode_block)),
1290                                        "array of inodes to process");
1291         process_inode_count = 0;
1292
1293         pctx.errcode = ext2fs_init_dblist(fs, 0);
1294         if (pctx.errcode) {
1295                 fix_problem(ctx, PR_1_ALLOCATE_DBCOUNT, &pctx);
1296                 ctx->flags |= E2F_FLAG_ABORT;
1297                 goto endit;
1298         }
1299
1300         /*
1301          * If the last orphan field is set, clear it, since the pass1
1302          * processing will automatically find and clear the orphans.
1303          * In the future, we may want to try using the last_orphan
1304          * linked list ourselves, but for now, we clear it so that the
1305          * ext3 mount code won't get confused.
1306          */
1307         if (!(ctx->options & E2F_OPT_READONLY)) {
1308                 if (fs->super->s_last_orphan) {
1309                         fs->super->s_last_orphan = 0;
1310                         ext2fs_mark_super_dirty(fs);
1311                 }
1312         }
1313
1314         mark_table_blocks(ctx);
1315         pctx.errcode = ext2fs_convert_subcluster_bitmap(fs,
1316                                                 &ctx->block_found_map);
1317         if (pctx.errcode) {
1318                 fix_problem(ctx, PR_1_CONVERT_SUBCLUSTER, &pctx);
1319                 ctx->flags |= E2F_FLAG_ABORT;
1320                 goto endit;
1321         }
1322         block_buf = (char *) e2fsck_allocate_memory(ctx, fs->blocksize * 3,
1323                                                     "block interate buffer");
1324         if (EXT2_INODE_SIZE(fs->super) == EXT2_GOOD_OLD_INODE_SIZE)
1325                 e2fsck_use_inode_shortcuts(ctx, 1);
1326         e2fsck_intercept_block_allocations(ctx);
1327         old_op = ehandler_operation(_("opening inode scan"));
1328         pctx.errcode = ext2fs_open_inode_scan(fs, ctx->inode_buffer_blocks,
1329                                               &scan);
1330         ehandler_operation(old_op);
1331         if (pctx.errcode) {
1332                 fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1333                 ctx->flags |= E2F_FLAG_ABORT;
1334                 goto endit;
1335         }
1336         ext2fs_inode_scan_flags(scan, EXT2_SF_SKIP_MISSING_ITABLE |
1337                                       EXT2_SF_WARN_GARBAGE_INODES, 0);
1338         ctx->stashed_inode = inode;
1339         scan_struct.ctx = ctx;
1340         scan_struct.block_buf = block_buf;
1341         ext2fs_set_inode_callback(scan, scan_callback, &scan_struct);
1342         if (ctx->progress && ((ctx->progress)(ctx, 1, 0,
1343                                               ctx->fs->group_desc_count)))
1344                 goto endit;
1345         if ((fs->super->s_wtime &&
1346              fs->super->s_wtime < fs->super->s_inodes_count) ||
1347             (fs->super->s_mtime &&
1348              fs->super->s_mtime < fs->super->s_inodes_count) ||
1349             (fs->super->s_mkfs_time &&
1350              fs->super->s_mkfs_time < fs->super->s_inodes_count))
1351                 low_dtime_check = 0;
1352
1353         if (ext2fs_has_feature_mmp(fs->super) &&
1354             fs->super->s_mmp_block > fs->super->s_first_data_block &&
1355             fs->super->s_mmp_block < ext2fs_blocks_count(fs->super))
1356                 ext2fs_mark_block_bitmap2(ctx->block_found_map,
1357                                           fs->super->s_mmp_block);
1358
1359         /* Set up ctx->lost_and_found if possible */
1360         (void) e2fsck_get_lost_and_found(ctx, 0);
1361
1362         while (1) {
1363                 if (ino % (fs->super->s_inodes_per_group * 4) == 1) {
1364                         if (e2fsck_mmp_update(fs))
1365                                 fatal_error(ctx, 0);
1366                 }
1367                 old_op = ehandler_operation(eop_next_inode);
1368                 pctx.errcode = ext2fs_get_next_inode_full(scan, &ino,
1369                                                           inode, inode_size);
1370                 if (ino > ino_threshold)
1371                         pass1_readahead(ctx, &ra_group, &ino_threshold);
1372                 ehandler_operation(old_op);
1373                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1374                         goto endit;
1375                 if (pctx.errcode == EXT2_ET_BAD_BLOCK_IN_INODE_TABLE) {
1376                         /*
1377                          * If badblocks says badblocks is bad, offer to clear
1378                          * the list, update the in-core bb list, and restart
1379                          * the inode scan.
1380                          */
1381                         if (ino == EXT2_BAD_INO &&
1382                             fix_problem(ctx, PR_1_BADBLOCKS_IN_BADBLOCKS,
1383                                         &pctx)) {
1384                                 errcode_t err;
1385
1386                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1387                                 ext2fs_badblocks_list_free(ctx->fs->badblocks);
1388                                 ctx->fs->badblocks = NULL;
1389                                 err = ext2fs_read_bb_inode(ctx->fs,
1390                                                         &ctx->fs->badblocks);
1391                                 if (err) {
1392                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1393                                                     &pctx);
1394                                         ctx->flags |= E2F_FLAG_ABORT;
1395                                         goto endit;
1396                                 }
1397                                 err = ext2fs_inode_scan_goto_blockgroup(scan,
1398                                                                         0);
1399                                 if (err) {
1400                                         fix_problem(ctx, PR_1_ISCAN_ERROR,
1401                                                     &pctx);
1402                                         ctx->flags |= E2F_FLAG_ABORT;
1403                                         goto endit;
1404                                 }
1405                                 continue;
1406                         }
1407                         if (!ctx->inode_bb_map)
1408                                 alloc_bb_map(ctx);
1409                         ext2fs_mark_inode_bitmap2(ctx->inode_bb_map, ino);
1410                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1411                         continue;
1412                 }
1413                 if (pctx.errcode &&
1414                     pctx.errcode != EXT2_ET_INODE_CSUM_INVALID &&
1415                     pctx.errcode != EXT2_ET_INODE_IS_GARBAGE) {
1416                         fix_problem(ctx, PR_1_ISCAN_ERROR, &pctx);
1417                         ctx->flags |= E2F_FLAG_ABORT;
1418                         goto endit;
1419                 }
1420                 if (!ino)
1421                         break;
1422                 pctx.ino = ino;
1423                 pctx.inode = inode;
1424                 ctx->stashed_ino = ino;
1425
1426                 /* Clear trashed inode? */
1427                 if (pctx.errcode == EXT2_ET_INODE_IS_GARBAGE &&
1428                     inode->i_links_count > 0 &&
1429                     fix_problem(ctx, PR_1_INODE_IS_GARBAGE, &pctx)) {
1430                         pctx.errcode = 0;
1431                         e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1432                 }
1433                 failed_csum = pctx.errcode != 0;
1434
1435                 /*
1436                  * Check for inodes who might have been part of the
1437                  * orphaned list linked list.  They should have gotten
1438                  * dealt with by now, unless the list had somehow been
1439                  * corrupted.
1440                  *
1441                  * FIXME: In the future, inodes which are still in use
1442                  * (and which are therefore) pending truncation should
1443                  * be handled specially.  Right now we just clear the
1444                  * dtime field, and the normal e2fsck handling of
1445                  * inodes where i_size and the inode blocks are
1446                  * inconsistent is to fix i_size, instead of releasing
1447                  * the extra blocks.  This won't catch the inodes that
1448                  * was at the end of the orphan list, but it's better
1449                  * than nothing.  The right answer is that there
1450                  * shouldn't be any bugs in the orphan list handling.  :-)
1451                  */
1452                 if (inode->i_dtime && low_dtime_check &&
1453                     inode->i_dtime < ctx->fs->super->s_inodes_count) {
1454                         if (fix_problem(ctx, PR_1_LOW_DTIME, &pctx)) {
1455                                 inode->i_dtime = inode->i_links_count ?
1456                                         0 : ctx->now;
1457                                 e2fsck_write_inode(ctx, ino, inode,
1458                                                    "pass1");
1459                                 failed_csum = 0;
1460                         }
1461                 }
1462
1463                 if (inode->i_links_count) {
1464                         pctx.errcode = ext2fs_icount_store(ctx->inode_link_info,
1465                                            ino, inode->i_links_count);
1466                         if (pctx.errcode) {
1467                                 pctx.num = inode->i_links_count;
1468                                 fix_problem(ctx, PR_1_ICOUNT_STORE, &pctx);
1469                                 ctx->flags |= E2F_FLAG_ABORT;
1470                                 goto endit;
1471                         }
1472                 } else if ((ino >= EXT2_FIRST_INODE(fs->super)) &&
1473                            !quota_inum_is_reserved(fs, ino)) {
1474                         if (!inode->i_dtime && inode->i_mode) {
1475                                 if (fix_problem(ctx,
1476                                             PR_1_ZERO_DTIME, &pctx)) {
1477                                         inode->i_dtime = ctx->now;
1478                                         e2fsck_write_inode(ctx, ino, inode,
1479                                                            "pass1");
1480                                         failed_csum = 0;
1481                                 }
1482                         }
1483                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1484                         continue;
1485                 }
1486
1487                 if ((inode->i_flags & EXT4_CASEFOLD_FL) &&
1488                     ((!LINUX_S_ISDIR(inode->i_mode) &&
1489                       fix_problem(ctx, PR_1_CASEFOLD_NONDIR, &pctx)) ||
1490                      (!casefold_fs &&
1491                       fix_problem(ctx, PR_1_CASEFOLD_FEATURE, &pctx)))) {
1492                         inode->i_flags &= ~EXT4_CASEFOLD_FL;
1493                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1494                 }
1495
1496                 /* Conflicting inlinedata/extents inode flags? */
1497                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) &&
1498                     (inode->i_flags & EXT4_EXTENTS_FL)) {
1499                         int res = fix_inline_data_extents_file(ctx, ino, inode,
1500                                                                inode_size,
1501                                                                &pctx);
1502                         if (res < 0) {
1503                                 /* skip FINISH_INODE_LOOP */
1504                                 continue;
1505                         }
1506                 }
1507
1508                 /* Test for incorrect inline_data flags settings. */
1509                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && !inlinedata_fs &&
1510                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1511                         size_t size = 0;
1512
1513                         pctx.errcode = get_inline_data_ea_size(fs, ino, &size);
1514                         if (!pctx.errcode &&
1515                             fix_problem(ctx, PR_1_INLINE_DATA_FEATURE, &pctx)) {
1516                                 ext2fs_set_feature_inline_data(sb);
1517                                 ext2fs_mark_super_dirty(fs);
1518                                 inlinedata_fs = 1;
1519                         } else if (fix_problem(ctx, PR_1_INLINE_DATA_SET, &pctx)) {
1520                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1521                                 /* skip FINISH_INODE_LOOP */
1522                                 continue;
1523                         }
1524                 }
1525
1526                 /* Test for inline data flag but no attr */
1527                 if ((inode->i_flags & EXT4_INLINE_DATA_FL) && inlinedata_fs &&
1528                     (ino >= EXT2_FIRST_INODE(fs->super))) {
1529                         size_t size = 0;
1530                         errcode_t err;
1531                         int flags;
1532
1533                         flags = fs->flags;
1534                         if (failed_csum)
1535                                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
1536                         err = get_inline_data_ea_size(fs, ino, &size);
1537                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
1538                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
1539
1540                         switch (err) {
1541                         case 0:
1542                                 /* Everything is awesome... */
1543                                 break;
1544                         case EXT2_ET_BAD_EA_BLOCK_NUM:
1545                         case EXT2_ET_BAD_EA_HASH:
1546                         case EXT2_ET_BAD_EA_HEADER:
1547                         case EXT2_ET_EA_BAD_NAME_LEN:
1548                         case EXT2_ET_EA_BAD_VALUE_SIZE:
1549                         case EXT2_ET_EA_KEY_NOT_FOUND:
1550                         case EXT2_ET_EA_NO_SPACE:
1551                         case EXT2_ET_MISSING_EA_FEATURE:
1552                         case EXT2_ET_INLINE_DATA_CANT_ITERATE:
1553                         case EXT2_ET_INLINE_DATA_NO_BLOCK:
1554                         case EXT2_ET_INLINE_DATA_NO_SPACE:
1555                         case EXT2_ET_NO_INLINE_DATA:
1556                         case EXT2_ET_EXT_ATTR_CSUM_INVALID:
1557                         case EXT2_ET_EA_BAD_VALUE_OFFSET:
1558                         case EXT2_ET_EA_INODE_CORRUPTED:
1559                                 /* broken EA or no system.data EA; truncate */
1560                                 if (fix_problem(ctx, PR_1_INLINE_DATA_NO_ATTR,
1561                                                 &pctx)) {
1562                                         err = ext2fs_inode_size_set(fs, inode, 0);
1563                                         if (err) {
1564                                                 pctx.errcode = err;
1565                                                 ctx->flags |= E2F_FLAG_ABORT;
1566                                                 goto endit;
1567                                         }
1568                                         inode->i_flags &= ~EXT4_INLINE_DATA_FL;
1569                                         memset(&inode->i_block, 0,
1570                                                sizeof(inode->i_block));
1571                                         e2fsck_write_inode(ctx, ino, inode,
1572                                                            "pass1");
1573                                         failed_csum = 0;
1574                                 }
1575                                 break;
1576                         default:
1577                                 /* Some other kind of non-xattr error? */
1578                                 pctx.errcode = err;
1579                                 ctx->flags |= E2F_FLAG_ABORT;
1580                                 goto endit;
1581                         }
1582                 }
1583
1584                 /*
1585                  * Test for incorrect extent flag settings.
1586                  *
1587                  * On big-endian machines we must be careful:
1588                  * When the inode is read, the i_block array is not swapped
1589                  * if the extent flag is set.  Therefore if we are testing
1590                  * for or fixing a wrongly-set flag, we must potentially
1591                  * (un)swap before testing, or after fixing.
1592                  */
1593
1594                 /*
1595                  * In this case the extents flag was set when read, so
1596                  * extent_header_verify is ok.  If the inode is cleared,
1597                  * no need to swap... so no extra swapping here.
1598                  */
1599                 if ((inode->i_flags & EXT4_EXTENTS_FL) && !extent_fs &&
1600                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1601                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO))) {
1602                         if ((ext2fs_extent_header_verify(inode->i_block,
1603                                                  sizeof(inode->i_block)) == 0) &&
1604                             fix_problem(ctx, PR_1_EXTENT_FEATURE, &pctx)) {
1605                                 ext2fs_set_feature_extents(sb);
1606                                 ext2fs_mark_super_dirty(fs);
1607                                 extent_fs = 1;
1608                         } else if (fix_problem(ctx, PR_1_EXTENTS_SET, &pctx)) {
1609                         clear_inode:
1610                                 e2fsck_clear_inode(ctx, ino, inode, 0, "pass1");
1611                                 if (ino == EXT2_BAD_INO)
1612                                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map,
1613                                                                  ino);
1614                                 /* skip FINISH_INODE_LOOP */
1615                                 continue;
1616                         }
1617                 }
1618
1619                 /*
1620                  * For big-endian machines:
1621                  * If the inode didn't have the extents flag set when it
1622                  * was read, then the i_blocks array was swapped.  To test
1623                  * as an extents header, we must swap it back first.
1624                  * IF we then set the extents flag, the entire i_block
1625                  * array must be un/re-swapped to make it proper extents data.
1626                  */
1627                 if (extent_fs && !(inode->i_flags & EXT4_EXTENTS_FL) &&
1628                     (inode->i_links_count || (ino == EXT2_BAD_INO) ||
1629                      (ino == EXT2_ROOT_INO) || (ino == EXT2_JOURNAL_INO)) &&
1630                     (LINUX_S_ISREG(inode->i_mode) ||
1631                      LINUX_S_ISDIR(inode->i_mode))) {
1632                         void *ehp;
1633 #ifdef WORDS_BIGENDIAN
1634                         __u32 tmp_block[EXT2_N_BLOCKS];
1635
1636                         for (i = 0; i < EXT2_N_BLOCKS; i++)
1637                                 tmp_block[i] = ext2fs_swab32(inode->i_block[i]);
1638                         ehp = tmp_block;
1639 #else
1640                         ehp = inode->i_block;
1641 #endif
1642                         if ((ext2fs_extent_header_verify(ehp,
1643                                          sizeof(inode->i_block)) == 0) &&
1644                             (fix_problem(ctx, PR_1_UNSET_EXTENT_FL, &pctx))) {
1645                                 inode->i_flags |= EXT4_EXTENTS_FL;
1646 #ifdef WORDS_BIGENDIAN
1647                                 memcpy(inode->i_block, tmp_block,
1648                                        sizeof(inode->i_block));
1649 #endif
1650                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1651                                 failed_csum = 0;
1652                         }
1653                 }
1654
1655                 if (ino == EXT2_BAD_INO) {
1656                         struct process_block_struct pb;
1657
1658                         if ((failed_csum || inode->i_mode || inode->i_uid ||
1659                              inode->i_gid || inode->i_links_count ||
1660                              (inode->i_flags & EXT4_INLINE_DATA_FL) ||
1661                              inode->i_file_acl) &&
1662                             fix_problem(ctx, PR_1_INVALID_BAD_INODE, &pctx)) {
1663                                 memset(inode, 0, sizeof(struct ext2_inode));
1664                                 e2fsck_write_inode(ctx, ino, inode,
1665                                                    "clear bad inode");
1666                                 failed_csum = 0;
1667                         }
1668
1669                         pctx.errcode = ext2fs_copy_bitmap(ctx->block_found_map,
1670                                                           &pb.fs_meta_blocks);
1671                         if (pctx.errcode) {
1672                                 pctx.num = 4;
1673                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, &pctx);
1674                                 ctx->flags |= E2F_FLAG_ABORT;
1675                                 goto endit;
1676                         }
1677                         pb.ino = EXT2_BAD_INO;
1678                         pb.num_blocks = pb.last_block = 0;
1679                         pb.last_db_block = -1;
1680                         pb.num_illegal_blocks = 0;
1681                         pb.suppress = 0; pb.clear = 0; pb.is_dir = 0;
1682                         pb.is_reg = 0; pb.fragmented = 0; pb.bbcheck = 0;
1683                         pb.inode = inode;
1684                         pb.pctx = &pctx;
1685                         pb.ctx = ctx;
1686                         pctx.errcode = ext2fs_block_iterate3(fs, ino, 0,
1687                                      block_buf, process_bad_block, &pb);
1688                         ext2fs_free_block_bitmap(pb.fs_meta_blocks);
1689                         if (pctx.errcode) {
1690                                 fix_problem(ctx, PR_1_BLOCK_ITERATE, &pctx);
1691                                 ctx->flags |= E2F_FLAG_ABORT;
1692                                 goto endit;
1693                         }
1694                         if (pb.bbcheck)
1695                                 if (!fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK_PROMPT, &pctx)) {
1696                                 ctx->flags |= E2F_FLAG_ABORT;
1697                                 goto endit;
1698                         }
1699                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1700                         clear_problem_context(&pctx);
1701                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1702                         continue;
1703                 } else if (ino == EXT2_ROOT_INO) {
1704                         /*
1705                          * Make sure the root inode is a directory; if
1706                          * not, offer to clear it.  It will be
1707                          * regenerated in pass #3.
1708                          */
1709                         if (!LINUX_S_ISDIR(inode->i_mode)) {
1710                                 if (fix_problem(ctx, PR_1_ROOT_NO_DIR, &pctx))
1711                                         goto clear_inode;
1712                         }
1713                         /*
1714                          * If dtime is set, offer to clear it.  mke2fs
1715                          * version 0.2b created filesystems with the
1716                          * dtime field set for the root and lost+found
1717                          * directories.  We won't worry about
1718                          * /lost+found, since that can be regenerated
1719                          * easily.  But we will fix the root directory
1720                          * as a special case.
1721                          */
1722                         if (inode->i_dtime && inode->i_links_count) {
1723                                 if (fix_problem(ctx, PR_1_ROOT_DTIME, &pctx)) {
1724                                         inode->i_dtime = 0;
1725                                         e2fsck_write_inode(ctx, ino, inode,
1726                                                            "pass1");
1727                                         failed_csum = 0;
1728                                 }
1729                         }
1730                 } else if (ino == EXT2_JOURNAL_INO) {
1731                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1732                         if (fs->super->s_journal_inum == EXT2_JOURNAL_INO) {
1733                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1734                                     fix_problem(ctx, PR_1_JOURNAL_BAD_MODE,
1735                                                 &pctx)) {
1736                                         inode->i_mode = LINUX_S_IFREG;
1737                                         e2fsck_write_inode(ctx, ino, inode,
1738                                                            "pass1");
1739                                         failed_csum = 0;
1740                                 }
1741                                 check_blocks(ctx, &pctx, block_buf, NULL);
1742                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1743                                 continue;
1744                         }
1745                         if ((inode->i_links_count ||
1746                              inode->i_blocks || inode->i_block[0]) &&
1747                             fix_problem(ctx, PR_1_JOURNAL_INODE_NOT_CLEAR,
1748                                         &pctx)) {
1749                                 memset(inode, 0, inode_size);
1750                                 ext2fs_icount_store(ctx->inode_link_info,
1751                                                     ino, 0);
1752                                 e2fsck_write_inode_full(ctx, ino, inode,
1753                                                         inode_size, "pass1");
1754                                 failed_csum = 0;
1755                         }
1756                 } else if (quota_inum_is_reserved(fs, ino)) {
1757                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1758                         if (ext2fs_has_feature_quota(fs->super) &&
1759                             quota_inum_is_super(fs->super, ino)) {
1760                                 if (!LINUX_S_ISREG(inode->i_mode) &&
1761                                     fix_problem(ctx, PR_1_QUOTA_BAD_MODE,
1762                                                         &pctx)) {
1763                                         inode->i_mode = LINUX_S_IFREG;
1764                                         e2fsck_write_inode(ctx, ino, inode,
1765                                                         "pass1");
1766                                         failed_csum = 0;
1767                                 }
1768                                 check_blocks(ctx, &pctx, block_buf, NULL);
1769                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1770                                 continue;
1771                         }
1772                         if ((inode->i_links_count ||
1773                              inode->i_blocks || inode->i_block[0]) &&
1774                             fix_problem(ctx, PR_1_QUOTA_INODE_NOT_CLEAR,
1775                                         &pctx)) {
1776                                 memset(inode, 0, inode_size);
1777                                 ext2fs_icount_store(ctx->inode_link_info,
1778                                                     ino, 0);
1779                                 e2fsck_write_inode_full(ctx, ino, inode,
1780                                                         inode_size, "pass1");
1781                                 failed_csum = 0;
1782                         }
1783                 } else if (ino < EXT2_FIRST_INODE(fs->super)) {
1784                         problem_t problem = 0;
1785
1786                         ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1787                         if (ino == EXT2_BOOT_LOADER_INO) {
1788                                 if (LINUX_S_ISDIR(inode->i_mode))
1789                                         problem = PR_1_RESERVED_BAD_MODE;
1790                         } else if (ino == EXT2_RESIZE_INO) {
1791                                 if (inode->i_mode &&
1792                                     !LINUX_S_ISREG(inode->i_mode))
1793                                         problem = PR_1_RESERVED_BAD_MODE;
1794                         } else {
1795                                 if (inode->i_mode != 0)
1796                                         problem = PR_1_RESERVED_BAD_MODE;
1797                         }
1798                         if (problem) {
1799                                 if (fix_problem(ctx, problem, &pctx)) {
1800                                         inode->i_mode = 0;
1801                                         e2fsck_write_inode(ctx, ino, inode,
1802                                                            "pass1");
1803                                         failed_csum = 0;
1804                                 }
1805                         }
1806                         check_blocks(ctx, &pctx, block_buf, NULL);
1807                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1808                         continue;
1809                 }
1810
1811                 if (!inode->i_links_count) {
1812                         FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1813                         continue;
1814                 }
1815                 /*
1816                  * n.b.  0.3c ext2fs code didn't clear i_links_count for
1817                  * deleted files.  Oops.
1818                  *
1819                  * Since all new ext2 implementations get this right,
1820                  * we now assume that the case of non-zero
1821                  * i_links_count and non-zero dtime means that we
1822                  * should keep the file, not delete it.
1823                  *
1824                  */
1825                 if (inode->i_dtime) {
1826                         if (fix_problem(ctx, PR_1_SET_DTIME, &pctx)) {
1827                                 inode->i_dtime = 0;
1828                                 e2fsck_write_inode(ctx, ino, inode, "pass1");
1829                                 failed_csum = 0;
1830                         }
1831                 }
1832
1833                 ext2fs_mark_inode_bitmap2(ctx->inode_used_map, ino);
1834                 switch (fs->super->s_creator_os) {
1835                     case EXT2_OS_HURD:
1836                         frag = inode->osd2.hurd2.h_i_frag;
1837                         fsize = inode->osd2.hurd2.h_i_fsize;
1838                         break;
1839                     default:
1840                         frag = fsize = 0;
1841                 }
1842
1843                 if (inode->i_faddr || frag || fsize ||
1844                     (!ext2fs_has_feature_largedir(fs->super) &&
1845                     (LINUX_S_ISDIR(inode->i_mode) && inode->i_size_high)))
1846                         mark_inode_bad(ctx, ino);
1847                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1848                     !ext2fs_has_feature_64bit(fs->super) &&
1849                     inode->osd2.linux2.l_i_file_acl_high != 0)
1850                         mark_inode_bad(ctx, ino);
1851                 if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
1852                     !ext2fs_has_feature_huge_file(fs->super) &&
1853                     (inode->osd2.linux2.l_i_blocks_hi != 0))
1854                         mark_inode_bad(ctx, ino);
1855                 if (inode->i_flags & EXT2_IMAGIC_FL) {
1856                         if (imagic_fs) {
1857                                 if (!ctx->inode_imagic_map)
1858                                         alloc_imagic_map(ctx);
1859                                 ext2fs_mark_inode_bitmap2(ctx->inode_imagic_map,
1860                                                          ino);
1861                         } else {
1862                                 if (fix_problem(ctx, PR_1_SET_IMAGIC, &pctx)) {
1863                                         inode->i_flags &= ~EXT2_IMAGIC_FL;
1864                                         e2fsck_write_inode(ctx, ino,
1865                                                            inode, "pass1");
1866                                         failed_csum = 0;
1867                                 }
1868                         }
1869                 }
1870
1871                 check_inode_extra_space(ctx, &pctx, &ea_ibody_quota);
1872                 check_is_really_dir(ctx, &pctx, block_buf);
1873
1874                 /*
1875                  * ext2fs_inode_has_valid_blocks2 does not actually look
1876                  * at i_block[] values, so not endian-sensitive here.
1877                  */
1878                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL) &&
1879                     LINUX_S_ISLNK(inode->i_mode) &&
1880                     !ext2fs_inode_has_valid_blocks2(fs, inode) &&
1881                     fix_problem(ctx, PR_1_FAST_SYMLINK_EXTENT_FL, &pctx)) {
1882                         inode->i_flags &= ~EXT4_EXTENTS_FL;
1883                         e2fsck_write_inode(ctx, ino, inode, "pass1");
1884                         failed_csum = 0;
1885                 }
1886
1887                 if (LINUX_S_ISDIR(inode->i_mode)) {
1888                         ext2fs_mark_inode_bitmap2(ctx->inode_dir_map, ino);
1889                         e2fsck_add_dir_info(ctx, ino, 0);
1890                         ctx->fs_directory_count++;
1891                         if (inode->i_flags & EXT4_ENCRYPT_FL)
1892                                 add_encrypted_dir(ctx, ino);
1893                 } else if (LINUX_S_ISREG (inode->i_mode)) {
1894                         ext2fs_mark_inode_bitmap2(ctx->inode_reg_map, ino);
1895                         ctx->fs_regular_count++;
1896                 } else if (LINUX_S_ISCHR (inode->i_mode) &&
1897                            e2fsck_pass1_check_device_inode(fs, inode)) {
1898                         check_extents_inlinedata(ctx, &pctx);
1899                         check_immutable(ctx, &pctx);
1900                         check_size(ctx, &pctx);
1901                         ctx->fs_chardev_count++;
1902                 } else if (LINUX_S_ISBLK (inode->i_mode) &&
1903                            e2fsck_pass1_check_device_inode(fs, inode)) {
1904                         check_extents_inlinedata(ctx, &pctx);
1905                         check_immutable(ctx, &pctx);
1906                         check_size(ctx, &pctx);
1907                         ctx->fs_blockdev_count++;
1908                 } else if (LINUX_S_ISLNK (inode->i_mode) &&
1909                            e2fsck_pass1_check_symlink(fs, ino, inode,
1910                                                       block_buf)) {
1911                         check_immutable(ctx, &pctx);
1912                         ctx->fs_symlinks_count++;
1913                         if (inode->i_flags & EXT4_INLINE_DATA_FL) {
1914                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1915                                 continue;
1916                         } else if (ext2fs_is_fast_symlink(inode)) {
1917                                 ctx->fs_fast_symlinks_count++;
1918                                 check_blocks(ctx, &pctx, block_buf,
1919                                              &ea_ibody_quota);
1920                                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1921                                 continue;
1922                         }
1923                 }
1924                 else if (LINUX_S_ISFIFO (inode->i_mode) &&
1925                          e2fsck_pass1_check_device_inode(fs, inode)) {
1926                         check_extents_inlinedata(ctx, &pctx);
1927                         check_immutable(ctx, &pctx);
1928                         check_size(ctx, &pctx);
1929                         ctx->fs_fifo_count++;
1930                 } else if ((LINUX_S_ISSOCK (inode->i_mode)) &&
1931                            e2fsck_pass1_check_device_inode(fs, inode)) {
1932                         check_extents_inlinedata(ctx, &pctx);
1933                         check_immutable(ctx, &pctx);
1934                         check_size(ctx, &pctx);
1935                         ctx->fs_sockets_count++;
1936                 } else
1937                         mark_inode_bad(ctx, ino);
1938                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1939                     !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
1940                         if (inode->i_block[EXT2_IND_BLOCK])
1941                                 ctx->fs_ind_count++;
1942                         if (inode->i_block[EXT2_DIND_BLOCK])
1943                                 ctx->fs_dind_count++;
1944                         if (inode->i_block[EXT2_TIND_BLOCK])
1945                                 ctx->fs_tind_count++;
1946                 }
1947                 if (!(inode->i_flags & EXT4_EXTENTS_FL) &&
1948                     !(inode->i_flags & EXT4_INLINE_DATA_FL) &&
1949                     (inode->i_block[EXT2_IND_BLOCK] ||
1950                      inode->i_block[EXT2_DIND_BLOCK] ||
1951                      inode->i_block[EXT2_TIND_BLOCK] ||
1952                      ext2fs_file_acl_block(fs, inode))) {
1953                         struct process_inode_block *itp;
1954
1955                         itp = &inodes_to_process[process_inode_count];
1956                         itp->ino = ino;
1957                         itp->ea_ibody_quota = ea_ibody_quota;
1958                         if (inode_size < sizeof(struct ext2_inode_large))
1959                                 memcpy(&itp->inode, inode, inode_size);
1960                         else
1961                                 memcpy(&itp->inode, inode, sizeof(itp->inode));
1962                         process_inode_count++;
1963                 } else
1964                         check_blocks(ctx, &pctx, block_buf, &ea_ibody_quota);
1965
1966                 FINISH_INODE_LOOP(ctx, ino, &pctx, failed_csum);
1967
1968                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1969                         goto endit;
1970
1971                 if (process_inode_count >= ctx->process_inode_size) {
1972                         process_inodes(ctx, block_buf);
1973
1974                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
1975                                 goto endit;
1976                 }
1977         }
1978         process_inodes(ctx, block_buf);
1979         ext2fs_close_inode_scan(scan);
1980         scan = NULL;
1981
1982         reserve_block_for_root_repair(ctx);
1983         reserve_block_for_lnf_repair(ctx);
1984
1985         /*
1986          * If any extended attribute blocks' reference counts need to
1987          * be adjusted, either up (ctx->refcount_extra), or down
1988          * (ctx->refcount), then fix them.
1989          */
1990         if (ctx->refcount) {
1991                 adjust_extattr_refcount(ctx, ctx->refcount, block_buf, -1);
1992                 ea_refcount_free(ctx->refcount);
1993                 ctx->refcount = 0;
1994         }
1995         if (ctx->refcount_extra) {
1996                 adjust_extattr_refcount(ctx, ctx->refcount_extra,
1997                                         block_buf, +1);
1998                 ea_refcount_free(ctx->refcount_extra);
1999                 ctx->refcount_extra = 0;
2000         }
2001
2002         if (ctx->ea_block_quota_blocks) {
2003                 ea_refcount_free(ctx->ea_block_quota_blocks);
2004                 ctx->ea_block_quota_blocks = 0;
2005         }
2006
2007         if (ctx->ea_block_quota_inodes) {
2008                 ea_refcount_free(ctx->ea_block_quota_inodes);
2009                 ctx->ea_block_quota_inodes = 0;
2010         }
2011
2012         if (ctx->invalid_bitmaps)
2013                 handle_fs_bad_blocks(ctx);
2014
2015         /* We don't need the block_ea_map any more */
2016         if (ctx->block_ea_map) {
2017                 ext2fs_free_block_bitmap(ctx->block_ea_map);
2018                 ctx->block_ea_map = 0;
2019         }
2020
2021         if (ctx->flags & E2F_FLAG_RESIZE_INODE) {
2022                 clear_problem_context(&pctx);
2023                 pctx.errcode = ext2fs_create_resize_inode(fs);
2024                 if (pctx.errcode) {
2025                         if (!fix_problem(ctx, PR_1_RESIZE_INODE_CREATE,
2026                                          &pctx)) {
2027                                 ctx->flags |= E2F_FLAG_ABORT;
2028                                 goto endit;
2029                         }
2030                         pctx.errcode = 0;
2031                 }
2032                 if (!pctx.errcode) {
2033                         e2fsck_read_inode(ctx, EXT2_RESIZE_INO, inode,
2034                                           "recreate inode");
2035                         inode->i_mtime = ctx->now;
2036                         e2fsck_write_inode(ctx, EXT2_RESIZE_INO, inode,
2037                                            "recreate inode");
2038                 }
2039                 ctx->flags &= ~E2F_FLAG_RESIZE_INODE;
2040         }
2041
2042         if (ctx->flags & E2F_FLAG_RESTART) {
2043                 /*
2044                  * Only the master copy of the superblock and block
2045                  * group descriptors are going to be written during a
2046                  * restart, so set the superblock to be used to be the
2047                  * master superblock.
2048                  */
2049                 ctx->use_superblock = 0;
2050                 unwind_pass1(fs);
2051                 goto endit;
2052         }
2053
2054         if (ctx->block_dup_map) {
2055                 if (ctx->options & E2F_OPT_PREEN) {
2056                         clear_problem_context(&pctx);
2057                         fix_problem(ctx, PR_1_DUP_BLOCKS_PREENSTOP, &pctx);
2058                 }
2059                 e2fsck_pass1_dupblocks(ctx, block_buf);
2060         }
2061         ctx->flags |= E2F_FLAG_ALLOC_OK;
2062         ext2fs_free_mem(&inodes_to_process);
2063 endit:
2064         e2fsck_use_inode_shortcuts(ctx, 0);
2065
2066         if (scan)
2067                 ext2fs_close_inode_scan(scan);
2068         if (block_buf)
2069                 ext2fs_free_mem(&block_buf);
2070         if (inode)
2071                 ext2fs_free_mem(&inode);
2072
2073         /*
2074          * The l+f inode may have been cleared, so zap it now and
2075          * later passes will recalculate it if necessary
2076          */
2077         ctx->lost_and_found = 0;
2078
2079         if ((ctx->flags & E2F_FLAG_SIGNAL_MASK) == 0)
2080                 print_resource_track(ctx, _("Pass 1"), &rtrack, ctx->fs->io);
2081         else
2082                 ctx->invalid_bitmaps++;
2083 }
2084 #undef FINISH_INODE_LOOP
2085
2086 /*
2087  * When the inode_scan routines call this callback at the end of the
2088  * glock group, call process_inodes.
2089  */
2090 static errcode_t scan_callback(ext2_filsys fs,
2091                                ext2_inode_scan scan EXT2FS_ATTR((unused)),
2092                                dgrp_t group, void * priv_data)
2093 {
2094         struct scan_callback_struct *scan_struct;
2095         e2fsck_t ctx;
2096
2097         scan_struct = (struct scan_callback_struct *) priv_data;
2098         ctx = scan_struct->ctx;
2099
2100         process_inodes((e2fsck_t) fs->priv_data, scan_struct->block_buf);
2101
2102         if (ctx->progress)
2103                 if ((ctx->progress)(ctx, 1, group+1,
2104                                     ctx->fs->group_desc_count))
2105                         return EXT2_ET_CANCEL_REQUESTED;
2106
2107         return 0;
2108 }
2109
2110 /*
2111  * Process the inodes in the "inodes to process" list.
2112  */
2113 static void process_inodes(e2fsck_t ctx, char *block_buf)
2114 {
2115         int                     i;
2116         struct ext2_inode       *old_stashed_inode;
2117         ext2_ino_t              old_stashed_ino;
2118         const char              *old_operation;
2119         char                    buf[80];
2120         struct problem_context  pctx;
2121
2122 #if 0
2123         printf("begin process_inodes: ");
2124 #endif
2125         if (process_inode_count == 0)
2126                 return;
2127         old_operation = ehandler_operation(0);
2128         old_stashed_inode = ctx->stashed_inode;
2129         old_stashed_ino = ctx->stashed_ino;
2130         qsort(inodes_to_process, process_inode_count,
2131                       sizeof(struct process_inode_block), process_inode_cmp);
2132         clear_problem_context(&pctx);
2133         for (i=0; i < process_inode_count; i++) {
2134                 pctx.inode = ctx->stashed_inode =
2135                         (struct ext2_inode *) &inodes_to_process[i].inode;
2136                 pctx.ino = ctx->stashed_ino = inodes_to_process[i].ino;
2137
2138 #if 0
2139                 printf("%u ", pctx.ino);
2140 #endif
2141                 sprintf(buf, _("reading indirect blocks of inode %u"),
2142                         pctx.ino);
2143                 ehandler_operation(buf);
2144                 check_blocks(ctx, &pctx, block_buf,
2145                              &inodes_to_process[i].ea_ibody_quota);
2146                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
2147                         break;
2148         }
2149         ctx->stashed_inode = old_stashed_inode;
2150         ctx->stashed_ino = old_stashed_ino;
2151         process_inode_count = 0;
2152 #if 0
2153         printf("end process inodes\n");
2154 #endif
2155         ehandler_operation(old_operation);
2156 }
2157
2158 static EXT2_QSORT_TYPE process_inode_cmp(const void *a, const void *b)
2159 {
2160         const struct process_inode_block *ib_a =
2161                 (const struct process_inode_block *) a;
2162         const struct process_inode_block *ib_b =
2163                 (const struct process_inode_block *) b;
2164         int     ret;
2165
2166         ret = (ib_a->inode.i_block[EXT2_IND_BLOCK] -
2167                ib_b->inode.i_block[EXT2_IND_BLOCK]);
2168         if (ret == 0)
2169                 /*
2170                  * We only call process_inodes() for non-extent
2171                  * inodes, so it's OK to pass NULL to
2172                  * ext2fs_file_acl_block() here.
2173                  */
2174                 ret = ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_a->inode)) -
2175                         ext2fs_file_acl_block(0, ext2fs_const_inode(&ib_b->inode));
2176         if (ret == 0)
2177                 ret = ib_a->ino - ib_b->ino;
2178         return ret;
2179 }
2180
2181 /*
2182  * Mark an inode as being bad in some what
2183  */
2184 static void mark_inode_bad(e2fsck_t ctx, ino_t ino)
2185 {
2186         struct          problem_context pctx;
2187
2188         if (!ctx->inode_bad_map) {
2189                 clear_problem_context(&pctx);
2190
2191                 pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2192                                 _("bad inode map"), EXT2FS_BMAP64_RBTREE,
2193                                 "inode_bad_map", &ctx->inode_bad_map);
2194                 if (pctx.errcode) {
2195                         pctx.num = 3;
2196                         fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2197                         /* Should never get here */
2198                         ctx->flags |= E2F_FLAG_ABORT;
2199                         return;
2200                 }
2201         }
2202         ext2fs_mark_inode_bitmap2(ctx->inode_bad_map, ino);
2203 }
2204
2205 static void add_encrypted_dir(e2fsck_t ctx, ino_t ino)
2206 {
2207         struct          problem_context pctx;
2208
2209         if (!ctx->encrypted_dirs) {
2210                 pctx.errcode = ext2fs_u32_list_create(&ctx->encrypted_dirs, 0);
2211                 if (pctx.errcode)
2212                         goto error;
2213         }
2214         pctx.errcode = ext2fs_u32_list_add(ctx->encrypted_dirs, ino);
2215         if (pctx.errcode == 0)
2216                 return;
2217 error:
2218         fix_problem(ctx, PR_1_ALLOCATE_ENCRYPTED_DIRLIST, &pctx);
2219         /* Should never get here */
2220         ctx->flags |= E2F_FLAG_ABORT;
2221 }
2222
2223 /*
2224  * This procedure will allocate the inode "bb" (badblock) map table
2225  */
2226 static void alloc_bb_map(e2fsck_t ctx)
2227 {
2228         struct          problem_context pctx;
2229
2230         clear_problem_context(&pctx);
2231         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2232                         _("inode in bad block map"), EXT2FS_BMAP64_RBTREE,
2233                         "inode_bb_map", &ctx->inode_bb_map);
2234         if (pctx.errcode) {
2235                 pctx.num = 4;
2236                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2237                 /* Should never get here */
2238                 ctx->flags |= E2F_FLAG_ABORT;
2239                 return;
2240         }
2241 }
2242
2243 /*
2244  * This procedure will allocate the inode imagic table
2245  */
2246 static void alloc_imagic_map(e2fsck_t ctx)
2247 {
2248         struct          problem_context pctx;
2249
2250         clear_problem_context(&pctx);
2251         pctx.errcode = e2fsck_allocate_inode_bitmap(ctx->fs,
2252                         _("imagic inode map"), EXT2FS_BMAP64_RBTREE,
2253                         "inode_imagic_map", &ctx->inode_imagic_map);
2254         if (pctx.errcode) {
2255                 pctx.num = 5;
2256                 fix_problem(ctx, PR_1_ALLOCATE_IBITMAP_ERROR, &pctx);
2257                 /* Should never get here */
2258                 ctx->flags |= E2F_FLAG_ABORT;
2259                 return;
2260         }
2261 }
2262
2263 /*
2264  * Marks a block as in use, setting the dup_map if it's been set
2265  * already.  Called by process_block and process_bad_block.
2266  *
2267  * WARNING: Assumes checks have already been done to make sure block
2268  * is valid.  This is true in both process_block and process_bad_block.
2269  */
2270 static _INLINE_ void mark_block_used(e2fsck_t ctx, blk64_t block)
2271 {
2272         struct          problem_context pctx;
2273
2274         clear_problem_context(&pctx);
2275
2276         if (ext2fs_fast_test_block_bitmap2(ctx->block_found_map, block)) {
2277                 if (ext2fs_has_feature_shared_blocks(ctx->fs->super) &&
2278                     !(ctx->options & E2F_OPT_UNSHARE_BLOCKS)) {
2279                         return;
2280                 }
2281                 if (!ctx->block_dup_map) {
2282                         pctx.errcode = e2fsck_allocate_block_bitmap(ctx->fs,
2283                                         _("multiply claimed block map"),
2284                                         EXT2FS_BMAP64_RBTREE, "block_dup_map",
2285                                         &ctx->block_dup_map);
2286                         if (pctx.errcode) {
2287                                 pctx.num = 3;
2288                                 fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR,
2289                                             &pctx);
2290                                 /* Should never get here */
2291                                 ctx->flags |= E2F_FLAG_ABORT;
2292                                 return;
2293                         }
2294                 }
2295                 ext2fs_fast_mark_block_bitmap2(ctx->block_dup_map, block);
2296         } else {
2297                 ext2fs_fast_mark_block_bitmap2(ctx->block_found_map, block);
2298         }
2299 }
2300
2301 /*
2302  * When cluster size is greater than one block, it is caller's responsibility
2303  * to make sure block parameter starts at a cluster boundary.
2304  */
2305 static _INLINE_ void mark_blocks_used(e2fsck_t ctx, blk64_t block,
2306                                       unsigned int num)
2307 {
2308         if (ext2fs_test_block_bitmap_range2(ctx->block_found_map, block, num))
2309                 ext2fs_mark_block_bitmap_range2(ctx->block_found_map, block, num);
2310         else {
2311                 unsigned int i;
2312
2313                 for (i = 0; i < num; i += EXT2FS_CLUSTER_RATIO(ctx->fs))
2314                         mark_block_used(ctx, block + i);
2315         }
2316 }
2317
2318 /*
2319  * Adjust the extended attribute block's reference counts at the end
2320  * of pass 1, either by subtracting out references for EA blocks that
2321  * are still referenced in ctx->refcount, or by adding references for
2322  * EA blocks that had extra references as accounted for in
2323  * ctx->refcount_extra.
2324  */
2325 static void adjust_extattr_refcount(e2fsck_t ctx, ext2_refcount_t refcount,
2326                                     char *block_buf, int adjust_sign)
2327 {
2328         struct ext2_ext_attr_header     *header;
2329         struct problem_context          pctx;
2330         ext2_filsys                     fs = ctx->fs;
2331         blk64_t                         blk;
2332         __u32                           should_be;
2333         ea_value_t                      count;
2334
2335         clear_problem_context(&pctx);
2336
2337         ea_refcount_intr_begin(refcount);
2338         while (1) {
2339                 if ((blk = ea_refcount_intr_next(refcount, &count)) == 0)
2340                         break;
2341                 pctx.blk = blk;
2342                 pctx.errcode = ext2fs_read_ext_attr3(fs, blk, block_buf,
2343                                                      pctx.ino);
2344                 if (pctx.errcode) {
2345                         fix_problem(ctx, PR_1_EXTATTR_READ_ABORT, &pctx);
2346                         return;
2347                 }
2348                 header = (struct ext2_ext_attr_header *) block_buf;
2349                 pctx.blkcount = header->h_refcount;
2350                 should_be = header->h_refcount + adjust_sign * (int)count;
2351                 pctx.num = should_be;
2352                 if (fix_problem(ctx, PR_1_EXTATTR_REFCOUNT, &pctx)) {
2353                         header->h_refcount = should_be;
2354                         pctx.errcode = ext2fs_write_ext_attr3(fs, blk,
2355                                                              block_buf,
2356                                                              pctx.ino);
2357                         if (pctx.errcode) {
2358                                 fix_problem(ctx, PR_1_EXTATTR_WRITE_ABORT,
2359                                             &pctx);
2360                                 continue;
2361                         }
2362                 }
2363         }
2364 }
2365
2366 /*
2367  * Handle processing the extended attribute blocks
2368  */
2369 static int check_ext_attr(e2fsck_t ctx, struct problem_context *pctx,
2370                            char *block_buf, struct ea_quota *ea_block_quota)
2371 {
2372         ext2_filsys fs = ctx->fs;
2373         ext2_ino_t      ino = pctx->ino;
2374         struct ext2_inode *inode = pctx->inode;
2375         blk64_t         blk;
2376         char *          end;
2377         struct ext2_ext_attr_header *header;
2378         struct ext2_ext_attr_entry *first, *entry;
2379         blk64_t         quota_blocks = EXT2FS_C2B(fs, 1);
2380         __u64           quota_inodes = 0;
2381         region_t        region = 0;
2382         int             failed_csum = 0;
2383
2384         ea_block_quota->blocks = 0;
2385         ea_block_quota->inodes = 0;
2386
2387         blk = ext2fs_file_acl_block(fs, inode);
2388         if (blk == 0)
2389                 return 0;
2390
2391         /*
2392          * If the Extended attribute flag isn't set, then a non-zero
2393          * file acl means that the inode is corrupted.
2394          *
2395          * Or if the extended attribute block is an invalid block,
2396          * then the inode is also corrupted.
2397          */
2398         if (!ext2fs_has_feature_xattr(fs->super) ||
2399             (blk < fs->super->s_first_data_block) ||
2400             (blk >= ext2fs_blocks_count(fs->super))) {
2401                 mark_inode_bad(ctx, ino);
2402                 return 0;
2403         }
2404
2405         /* If ea bitmap hasn't been allocated, create it */
2406         if (!ctx->block_ea_map) {
2407                 pctx->errcode = e2fsck_allocate_block_bitmap(fs,
2408                                         _("ext attr block map"),
2409                                         EXT2FS_BMAP64_RBTREE, "block_ea_map",
2410                                         &ctx->block_ea_map);
2411                 if (pctx->errcode) {
2412                         pctx->num = 2;
2413                         fix_problem(ctx, PR_1_ALLOCATE_BBITMAP_ERROR, pctx);
2414                         ctx->flags |= E2F_FLAG_ABORT;
2415                         return 0;
2416                 }
2417         }
2418
2419         /* Create the EA refcount structure if necessary */
2420         if (!ctx->refcount) {
2421                 pctx->errcode = ea_refcount_create(0, &ctx->refcount);
2422                 if (pctx->errcode) {
2423                         pctx->num = 1;
2424                         fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2425                         ctx->flags |= E2F_FLAG_ABORT;
2426                         return 0;
2427                 }
2428         }
2429
2430 #if 0
2431         /* Debugging text */
2432         printf("Inode %u has EA block %u\n", ino, blk);
2433 #endif
2434
2435         /* Have we seen this EA block before? */
2436         if (ext2fs_fast_test_block_bitmap2(ctx->block_ea_map, blk)) {
2437                 ea_block_quota->blocks = EXT2FS_C2B(fs, 1);
2438                 ea_block_quota->inodes = 0;
2439
2440                 if (ctx->ea_block_quota_blocks) {
2441                         ea_refcount_fetch(ctx->ea_block_quota_blocks, blk,
2442                                           &quota_blocks);
2443                         if (quota_blocks)
2444                                 ea_block_quota->blocks = quota_blocks;
2445                 }
2446
2447                 if (ctx->ea_block_quota_inodes)
2448                         ea_refcount_fetch(ctx->ea_block_quota_inodes, blk,
2449                                           &ea_block_quota->inodes);
2450
2451                 if (ea_refcount_decrement(ctx->refcount, blk, 0) == 0)
2452                         return 1;
2453                 /* Ooops, this EA was referenced more than it stated */
2454                 if (!ctx->refcount_extra) {
2455                         pctx->errcode = ea_refcount_create(0,
2456                                            &ctx->refcount_extra);
2457                         if (pctx->errcode) {
2458                                 pctx->num = 2;
2459                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2460                                 ctx->flags |= E2F_FLAG_ABORT;
2461                                 return 0;
2462                         }
2463                 }
2464                 ea_refcount_increment(ctx->refcount_extra, blk, 0);
2465                 return 1;
2466         }
2467
2468         /*
2469          * OK, we haven't seen this EA block yet.  So we need to
2470          * validate it
2471          */
2472         pctx->blk = blk;
2473         pctx->errcode = ext2fs_read_ext_attr3(fs, blk, block_buf, pctx->ino);
2474         if (pctx->errcode == EXT2_ET_EXT_ATTR_CSUM_INVALID) {
2475                 pctx->errcode = 0;
2476                 failed_csum = 1;
2477         } else if (pctx->errcode == EXT2_ET_BAD_EA_HEADER)
2478                 pctx->errcode = 0;
2479
2480         if (pctx->errcode &&
2481             fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx)) {
2482                 pctx->errcode = 0;
2483                 goto clear_extattr;
2484         }
2485         header = (struct ext2_ext_attr_header *) block_buf;
2486         pctx->blk = ext2fs_file_acl_block(fs, inode);
2487         if (((ctx->ext_attr_ver == 1) &&
2488              (header->h_magic != EXT2_EXT_ATTR_MAGIC_v1)) ||
2489             ((ctx->ext_attr_ver == 2) &&
2490              (header->h_magic != EXT2_EXT_ATTR_MAGIC))) {
2491                 if (fix_problem(ctx, PR_1_BAD_EA_BLOCK, pctx))
2492                         goto clear_extattr;
2493         }
2494
2495         if (header->h_blocks != 1) {
2496                 if (fix_problem(ctx, PR_1_EA_MULTI_BLOCK, pctx))
2497                         goto clear_extattr;
2498         }
2499
2500         if (pctx->errcode && fix_problem(ctx, PR_1_READ_EA_BLOCK, pctx))
2501                 goto clear_extattr;
2502
2503         region = region_create(0, fs->blocksize);
2504         if (!region) {
2505                 fix_problem(ctx, PR_1_EA_ALLOC_REGION_ABORT, pctx);
2506                 ctx->flags |= E2F_FLAG_ABORT;
2507                 return 0;
2508         }
2509         if (region_allocate(region, 0, sizeof(struct ext2_ext_attr_header))) {
2510                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2511                         goto clear_extattr;
2512         }
2513
2514         first = (struct ext2_ext_attr_entry *)(header+1);
2515         end = block_buf + fs->blocksize;
2516         entry = first;
2517         while ((char *)entry < end && *(__u32 *)entry) {
2518                 __u32 hash;
2519
2520                 if (region_allocate(region, (char *)entry - (char *)header,
2521                                    EXT2_EXT_ATTR_LEN(entry->e_name_len))) {
2522                         if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2523                                 goto clear_extattr;
2524                         break;
2525                 }
2526                 if ((ctx->ext_attr_ver == 1 &&
2527                      (entry->e_name_len == 0 || entry->e_name_index != 0)) ||
2528                     (ctx->ext_attr_ver == 2 &&
2529                      entry->e_name_index == 0)) {
2530                         if (fix_problem(ctx, PR_1_EA_BAD_NAME, pctx))
2531                                 goto clear_extattr;
2532                         break;
2533                 }
2534                 if (entry->e_value_inum == 0) {
2535                         if (entry->e_value_offs + entry->e_value_size >
2536                             fs->blocksize) {
2537                                 if (fix_problem(ctx, PR_1_EA_BAD_VALUE, pctx))
2538                                         goto clear_extattr;
2539                                 break;
2540                         }
2541                         if (entry->e_value_size &&
2542                             region_allocate(region, entry->e_value_offs,
2543                                             EXT2_EXT_ATTR_SIZE(entry->e_value_size))) {
2544                                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION,
2545                                                 pctx))
2546                                         goto clear_extattr;
2547                         }
2548
2549                         hash = ext2fs_ext_attr_hash_entry(entry, block_buf +
2550                                                           entry->e_value_offs);
2551
2552                         if (entry->e_hash != hash) {
2553                                 pctx->num = entry->e_hash;
2554                                 if (fix_problem(ctx, PR_1_ATTR_HASH, pctx))
2555                                         goto clear_extattr;
2556                                 entry->e_hash = hash;
2557                         }
2558                 } else {
2559                         problem_t problem;
2560                         blk64_t entry_quota_blocks;
2561
2562                         problem = check_large_ea_inode(ctx, entry, pctx,
2563                                                        &entry_quota_blocks);
2564                         if (problem && fix_problem(ctx, problem, pctx))
2565                                 goto clear_extattr;
2566
2567                         quota_blocks += entry_quota_blocks;
2568                         quota_inodes++;
2569                 }
2570
2571                 entry = EXT2_EXT_ATTR_NEXT(entry);
2572         }
2573         if (region_allocate(region, (char *)entry - (char *)header, 4)) {
2574                 if (fix_problem(ctx, PR_1_EA_ALLOC_COLLISION, pctx))
2575                         goto clear_extattr;
2576         }
2577         region_free(region);
2578
2579         /*
2580          * We only get here if there was no other errors that were fixed.
2581          * If there was a checksum fail, ask to correct it.
2582          */
2583         if (failed_csum &&
2584             fix_problem(ctx, PR_1_EA_BLOCK_ONLY_CSUM_INVALID, pctx)) {
2585                 pctx->errcode = ext2fs_write_ext_attr3(fs, blk, block_buf,
2586                                                        pctx->ino);
2587                 if (pctx->errcode)
2588                         return 0;
2589         }
2590
2591         if (quota_blocks != EXT2FS_C2B(fs, 1U)) {
2592                 if (!ctx->ea_block_quota_blocks) {
2593                         pctx->errcode = ea_refcount_create(0,
2594                                                 &ctx->ea_block_quota_blocks);
2595                         if (pctx->errcode) {
2596                                 pctx->num = 3;
2597                                 goto refcount_fail;
2598                         }
2599                 }
2600                 ea_refcount_store(ctx->ea_block_quota_blocks, blk,
2601                                   quota_blocks);
2602         }
2603
2604         if (quota_inodes) {
2605                 if (!ctx->ea_block_quota_inodes) {
2606                         pctx->errcode = ea_refcount_create(0,
2607                                                 &ctx->ea_block_quota_inodes);
2608                         if (pctx->errcode) {
2609                                 pctx->num = 4;
2610 refcount_fail:
2611                                 fix_problem(ctx, PR_1_ALLOCATE_REFCOUNT, pctx);
2612                                 ctx->flags |= E2F_FLAG_ABORT;
2613                                 return 0;
2614                         }
2615                 }
2616
2617                 ea_refcount_store(ctx->ea_block_quota_inodes, blk,
2618                                   quota_inodes);
2619         }
2620         ea_block_quota->blocks = quota_blocks;
2621         ea_block_quota->inodes = quota_inodes;
2622
2623         inc_ea_inode_refs(ctx, pctx, first, end);
2624         ea_refcount_store(ctx->refcount, blk, header->h_refcount - 1);
2625         mark_block_used(ctx, blk);
2626         ext2fs_fast_mark_block_bitmap2(ctx->block_ea_map, blk);
2627         return 1;
2628
2629 clear_extattr:
2630         if (region)
2631                 region_free(region);
2632         ext2fs_file_acl_block_set(fs, inode, 0);
2633         e2fsck_write_inode(ctx, ino, inode, "check_ext_attr");
2634         return 0;
2635 }
2636
2637 /* Returns 1 if bad htree, 0 if OK */
2638 static int handle_htree(e2fsck_t ctx, struct problem_context *pctx,
2639                         ext2_ino_t ino, struct ext2_inode *inode,
2640                         char *block_buf)
2641 {
2642         struct ext2_dx_root_info        *root;
2643         ext2_filsys                     fs = ctx->fs;
2644         errcode_t                       retval;
2645         blk64_t                         blk;
2646
2647         if ((!LINUX_S_ISDIR(inode->i_mode) &&
2648              fix_problem(ctx, PR_1_HTREE_NODIR, pctx)) ||
2649             (!ext2fs_has_feature_dir_index(fs->super) &&
2650              fix_problem(ctx, PR_1_HTREE_SET, pctx)))
2651                 return 1;
2652
2653         pctx->errcode = ext2fs_bmap2(fs, ino, inode, 0, 0, 0, 0, &blk);
2654
2655         if ((pctx->errcode) ||
2656             (blk == 0) ||
2657             (blk < fs->super->s_first_data_block) ||
2658             (blk >= ext2fs_blocks_count(fs->super))) {
2659                 if (fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2660                         return 1;
2661                 else
2662                         return 0;
2663         }
2664
2665         retval = io_channel_read_blk64(fs->io, blk, 1, block_buf);
2666         if (retval && fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2667                 return 1;
2668
2669         /* XXX should check that beginning matches a directory */
2670         root = (struct ext2_dx_root_info *) (block_buf + 24);
2671
2672         if ((root->reserved_zero || root->info_length < 8) &&
2673             fix_problem(ctx, PR_1_HTREE_BADROOT, pctx))
2674                 return 1;
2675
2676         pctx->num = root->hash_version;
2677         if ((root->hash_version != EXT2_HASH_LEGACY) &&
2678             (root->hash_version != EXT2_HASH_HALF_MD4) &&
2679             (root->hash_version != EXT2_HASH_TEA) &&
2680             fix_problem(ctx, PR_1_HTREE_HASHV, pctx))
2681                 return 1;
2682
2683         if ((root->unused_flags & EXT2_HASH_FLAG_INCOMPAT) &&
2684             fix_problem(ctx, PR_1_HTREE_INCOMPAT, pctx))
2685                 return 1;
2686
2687         pctx->num = root->indirect_levels;
2688         if ((root->indirect_levels > ext2_dir_htree_level(fs)) &&
2689             fix_problem(ctx, PR_1_HTREE_DEPTH, pctx))
2690                 return 1;
2691
2692         return 0;
2693 }
2694
2695 void e2fsck_clear_inode(e2fsck_t ctx, ext2_ino_t ino,
2696                         struct ext2_inode *inode, int restart_flag,
2697                         const char *source)
2698 {
2699         inode->i_flags = 0;
2700         inode->i_links_count = 0;
2701         ext2fs_icount_store(ctx->inode_link_info, ino, 0);
2702         inode->i_dtime = ctx->now;
2703
2704         /*
2705          * If a special inode has such rotten block mappings that we
2706          * want to clear the whole inode, be sure to actually zap
2707          * the block maps because i_links_count isn't checked for
2708          * special inodes, and we'll end up right back here the next
2709          * time we run fsck.
2710          */
2711         if (ino < EXT2_FIRST_INODE(ctx->fs->super))
2712                 memset(inode->i_block, 0, sizeof(inode->i_block));
2713
2714         ext2fs_unmark_inode_bitmap2(ctx->inode_dir_map, ino);
2715         ext2fs_unmark_inode_bitmap2(ctx->inode_used_map, ino);
2716         if (ctx->inode_reg_map)
2717                 ext2fs_unmark_inode_bitmap2(ctx->inode_reg_map, ino);
2718         if (ctx->inode_bad_map)
2719                 ext2fs_unmark_inode_bitmap2(ctx->inode_bad_map, ino);
2720
2721         /*
2722          * If the inode was partially accounted for before processing
2723          * was aborted, we need to restart the pass 1 scan.
2724          */
2725         ctx->flags |= restart_flag;
2726
2727         if (ino == EXT2_BAD_INO)
2728                 memset(inode, 0, sizeof(struct ext2_inode));
2729
2730         e2fsck_write_inode(ctx, ino, inode, source);
2731 }
2732
2733 /*
2734  * Use the multiple-blocks reclamation code to fix alignment problems in
2735  * a bigalloc filesystem.  We want a logical cluster to map to *only* one
2736  * physical cluster, and we want the block offsets within that cluster to
2737  * line up.
2738  */
2739 static int has_unaligned_cluster_map(e2fsck_t ctx,
2740                                      blk64_t last_pblk, blk64_t last_lblk,
2741                                      blk64_t pblk, blk64_t lblk)
2742 {
2743         blk64_t cluster_mask;
2744
2745         if (!ctx->fs->cluster_ratio_bits)
2746                 return 0;
2747         cluster_mask = EXT2FS_CLUSTER_MASK(ctx->fs);
2748
2749         /*
2750          * If the block in the logical cluster doesn't align with the block in
2751          * the physical cluster...
2752          */
2753         if ((lblk & cluster_mask) != (pblk & cluster_mask))
2754                 return 1;
2755
2756         /*
2757          * If we cross a physical cluster boundary within a logical cluster...
2758          */
2759         if (last_pblk && (lblk & cluster_mask) != 0 &&
2760             EXT2FS_B2C(ctx->fs, lblk) == EXT2FS_B2C(ctx->fs, last_lblk) &&
2761             EXT2FS_B2C(ctx->fs, pblk) != EXT2FS_B2C(ctx->fs, last_pblk))
2762                 return 1;
2763
2764         return 0;
2765 }
2766
2767 static void scan_extent_node(e2fsck_t ctx, struct problem_context *pctx,
2768                              struct process_block_struct *pb,
2769                              blk64_t start_block, blk64_t end_block,
2770                              blk64_t eof_block,
2771                              ext2_extent_handle_t ehandle,
2772                              int try_repairs)
2773 {
2774         struct ext2fs_extent    extent;
2775         blk64_t                 blk, last_lblk;
2776         unsigned int            i, n;
2777         int                     is_dir, is_leaf;
2778         problem_t               problem;
2779         struct ext2_extent_info info;
2780         int                     failed_csum = 0;
2781
2782         if (pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID)
2783                 failed_csum = 1;
2784
2785         pctx->errcode = ext2fs_extent_get_info(ehandle, &info);
2786         if (pctx->errcode)
2787                 return;
2788         if (!(ctx->options & E2F_OPT_FIXES_ONLY) &&
2789             !pb->eti.force_rebuild) {
2790                 struct extent_tree_level *etl;
2791
2792                 etl = pb->eti.ext_info + info.curr_level;
2793                 etl->num_extents += info.num_entries;
2794                 etl->max_extents += info.max_entries;
2795                 /*
2796                  * Implementation wart: Splitting extent blocks when appending
2797                  * will leave the old block with one free entry.  Therefore
2798                  * unless the node is totally full, pretend that a non-root
2799                  * extent block can hold one fewer entry than it actually does,
2800                  * so that we don't repeatedly rebuild the extent tree.
2801                  */
2802                 if (info.curr_level && info.num_entries < info.max_entries)
2803                         etl->max_extents--;
2804         }
2805
2806         pctx->errcode = ext2fs_extent_get(ehandle, EXT2_EXTENT_FIRST_SIB,
2807                                           &extent);
2808         while ((pctx->errcode == 0 ||
2809                 pctx->errcode == EXT2_ET_EXTENT_CSUM_INVALID) &&
2810                info.num_entries-- > 0) {
2811                 is_leaf = extent.e_flags & EXT2_EXTENT_FLAGS_LEAF;
2812                 is_dir = LINUX_S_ISDIR(pctx->inode->i_mode);
2813                 last_lblk = extent.e_lblk + extent.e_len - 1;
2814
2815                 problem = 0;
2816                 pctx->blk = extent.e_pblk;
2817                 pctx->blk2 = extent.e_lblk;
2818                 pctx->num = extent.e_len;
2819                 pctx->blkcount = extent.e_lblk + extent.e_len;
2820
2821                 if (extent.e_pblk == 0 ||
2822                     extent.e_pblk < ctx->fs->super->s_first_data_block ||
2823                     extent.e_pblk >= ext2fs_blocks_count(ctx->fs->super))
2824                         problem = PR_1_EXTENT_BAD_START_BLK;
2825                 else if (extent.e_lblk < start_block)
2826                         problem = PR_1_OUT_OF_ORDER_EXTENTS;
2827                 else if ((end_block && last_lblk > end_block) &&
2828                          !(last_lblk > eof_block &&
2829                            ((extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) ||
2830                             (pctx->inode->i_flags & EXT4_VERITY_FL))))
2831                         problem = PR_1_EXTENT_END_OUT_OF_BOUNDS;
2832                 else if (is_leaf && extent.e_len == 0)
2833                         problem = PR_1_EXTENT_LENGTH_ZERO;
2834                 else if (is_leaf &&
2835                          (extent.e_pblk + extent.e_len) >
2836                          ext2fs_blocks_count(ctx->fs->super))
2837                         problem = PR_1_EXTENT_ENDS_BEYOND;
2838                 else if (is_leaf && is_dir &&
2839                          ((extent.e_lblk + extent.e_len) >
2840                           (1U << (21 - ctx->fs->super->s_log_block_size))))
2841                         problem = PR_1_TOOBIG_DIR;
2842
2843                 if (is_leaf && problem == 0 && extent.e_len > 0) {
2844 #if 0
2845                         printf("extent_region(ino=%u, expect=%llu, "
2846                                "lblk=%llu, len=%u)\n",
2847                                pb->ino, pb->next_lblock,
2848                                extent.e_lblk, extent.e_len);
2849 #endif
2850                         if (extent.e_lblk < pb->next_lblock)
2851                                 problem = PR_1_EXTENT_COLLISION;
2852                         else if (extent.e_lblk + extent.e_len > pb->next_lblock)
2853                                 pb->next_lblock = extent.e_lblk + extent.e_len;
2854                 }
2855
2856                 /*
2857                  * Uninitialized blocks in a directory?  Clear the flag and
2858                  * we'll interpret the blocks later.
2859                  */
2860                 if (try_repairs && is_dir && problem == 0 &&
2861                     (extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT) &&
2862                     fix_problem(ctx, PR_1_UNINIT_DBLOCK, pctx)) {
2863                         extent.e_flags &= ~EXT2_EXTENT_FLAGS_UNINIT;
2864                         pb->inode_modified = 1;
2865                         pctx->errcode = ext2fs_extent_replace(ehandle, 0,
2866                                                               &extent);
2867                         if (pctx->errcode)
2868                                 return;
2869                         failed_csum = 0;
2870                 }
2871
2872                 if (try_repairs && problem) {
2873 report_problem:
2874                         if (fix_problem(ctx, problem, pctx)) {
2875                                 if (ctx->invalid_bitmaps) {
2876                                         /*
2877                                          * If fsck knows the bitmaps are bad,
2878                                          * skip to the next extent and
2879                                          * try to clear this extent again
2880                                          * after fixing the bitmaps, by
2881                                          * restarting fsck.
2882                                          */
2883                                         pctx->errcode = ext2fs_extent_get(
2884                                                           ehandle,
2885                                                           EXT2_EXTENT_NEXT_SIB,
2886                                                           &extent);
2887                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2888                                         if (pctx->errcode ==
2889                                                     EXT2_ET_NO_CURRENT_NODE) {
2890                                                 pctx->errcode = 0;
2891                                                 break;
2892                                         }
2893                                         continue;
2894                                 }
2895                                 e2fsck_read_bitmaps(ctx);
2896                                 pb->inode_modified = 1;
2897                                 pctx->errcode =
2898                                         ext2fs_extent_delete(ehandle, 0);
2899                                 if (pctx->errcode) {
2900                                         pctx->str = "ext2fs_extent_delete";
2901                                         return;
2902                                 }
2903                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
2904                                 if (pctx->errcode &&
2905                                     pctx->errcode != EXT2_ET_NO_CURRENT_NODE) {
2906                                         pctx->str = "ext2fs_extent_fix_parents";
2907                                         return;
2908                                 }
2909                                 pctx->errcode = ext2fs_extent_get(ehandle,
2910                                                                   EXT2_EXTENT_CURRENT,
2911                                                                   &extent);
2912                                 if (pctx->errcode == EXT2_ET_NO_CURRENT_NODE) {
2913                                         pctx->errcode = 0;
2914                                         break;
2915                                 }
2916                                 failed_csum = 0;
2917                                 continue;
2918                         }
2919                         goto next;
2920                 }
2921
2922                 if (!is_leaf) {
2923                         blk64_t lblk = extent.e_lblk;
2924                         int next_try_repairs = 1;
2925
2926                         blk = extent.e_pblk;
2927
2928                         /*
2929                          * If this lower extent block collides with critical
2930                          * metadata, don't try to repair the damage.  Pass 1b
2931                          * will reallocate the block; then we can try again.
2932                          */
2933                         if (pb->ino != EXT2_RESIZE_INO &&
2934                             extent.e_pblk < ctx->fs->super->s_blocks_count &&
2935                             ext2fs_test_block_bitmap2(ctx->block_metadata_map,
2936                                                       extent.e_pblk)) {
2937                                 next_try_repairs = 0;
2938                                 pctx->blk = blk;
2939                                 fix_problem(ctx,
2940                                             PR_1_CRITICAL_METADATA_COLLISION,
2941                                             pctx);
2942                                 if ((ctx->options & E2F_OPT_NO) == 0)
2943                                         ctx->flags |= E2F_FLAG_RESTART_LATER;
2944                         }
2945                         pctx->errcode = ext2fs_extent_get(ehandle,
2946                                                   EXT2_EXTENT_DOWN, &extent);
2947                         if (pctx->errcode &&
2948                             pctx->errcode != EXT2_ET_EXTENT_CSUM_INVALID) {
2949                                 pctx->str = "EXT2_EXTENT_DOWN";
2950                                 problem = PR_1_EXTENT_HEADER_INVALID;
2951                                 if (!next_try_repairs)
2952                                         return;
2953                                 if (pctx->errcode == EXT2_ET_EXTENT_HEADER_BAD)
2954                                         goto report_problem;
2955                                 return;
2956                         }
2957                         /* The next extent should match this index's logical start */
2958                         if (extent.e_lblk != lblk) {
2959                                 struct ext2_extent_info e_info;
2960
2961                                 ext2fs_extent_get_info(ehandle, &e_info);
2962                                 pctx->blk = lblk;
2963                                 pctx->blk2 = extent.e_lblk;
2964                                 pctx->num = e_info.curr_level - 1;
2965                                 problem = PR_1_EXTENT_INDEX_START_INVALID;
2966                                 if (fix_problem(ctx, problem, pctx)) {
2967                                         pb->inode_modified = 1;
2968                                         pctx->errcode =
2969                                                 ext2fs_extent_fix_parents(ehandle);
2970                                         if (pctx->errcode) {
2971                                                 pctx->str = "ext2fs_extent_fix_parents";
2972                                                 return;
2973                                         }
2974                                 }
2975                         }
2976                         scan_extent_node(ctx, pctx, pb, extent.e_lblk,
2977                                          last_lblk, eof_block, ehandle,
2978                                          next_try_repairs);
2979                         if (pctx->errcode)
2980                                 return;
2981                         pctx->errcode = ext2fs_extent_get(ehandle,
2982                                                   EXT2_EXTENT_UP, &extent);
2983                         if (pctx->errcode) {
2984                                 pctx->str = "EXT2_EXTENT_UP";
2985                                 return;
2986                         }
2987                         mark_block_used(ctx, blk);
2988                         pb->num_blocks++;
2989                         goto next;
2990                 }
2991
2992                 if ((pb->previous_block != 0) &&
2993                     (pb->previous_block+1 != extent.e_pblk)) {
2994                         if (ctx->options & E2F_OPT_FRAGCHECK) {
2995                                 char type = '?';
2996
2997                                 if (pb->is_dir)
2998                                         type = 'd';
2999                                 else if (pb->is_reg)
3000                                         type = 'f';
3001
3002                                 printf(("%6lu(%c): expecting %6lu "
3003                                         "actual extent "
3004                                         "phys %6lu log %lu len %lu\n"),
3005                                        (unsigned long) pctx->ino, type,
3006                                        (unsigned long) pb->previous_block+1,
3007                                        (unsigned long) extent.e_pblk,
3008                                        (unsigned long) extent.e_lblk,
3009                                        (unsigned long) extent.e_len);
3010                         }
3011                         pb->fragmented = 1;
3012                 }
3013                 /*
3014                  * If we notice a gap in the logical block mappings of an
3015                  * extent-mapped directory, offer to close the hole by
3016                  * moving the logical block down, otherwise we'll go mad in
3017                  * pass 3 allocating empty directory blocks to fill the hole.
3018                  */
3019                 if (try_repairs && is_dir &&
3020                     pb->last_block + 1 < extent.e_lblk) {
3021                         blk64_t new_lblk;
3022
3023                         new_lblk = pb->last_block + 1;
3024                         if (EXT2FS_CLUSTER_RATIO(ctx->fs) > 1)
3025                                 new_lblk = ((new_lblk +
3026                                              EXT2FS_CLUSTER_RATIO(ctx->fs) - 1) &
3027                                             ~EXT2FS_CLUSTER_MASK(ctx->fs)) |
3028                                            (extent.e_pblk &
3029                                             EXT2FS_CLUSTER_MASK(ctx->fs));
3030                         pctx->blk = extent.e_lblk;
3031                         pctx->blk2 = new_lblk;
3032                         if (fix_problem(ctx, PR_1_COLLAPSE_DBLOCK, pctx)) {
3033                                 extent.e_lblk = new_lblk;
3034                                 pb->inode_modified = 1;
3035                                 pctx->errcode = ext2fs_extent_replace(ehandle,
3036                                                                 0, &extent);
3037                                 if (pctx->errcode) {
3038                                         pctx->errcode = 0;
3039                                         goto alloc_later;
3040                                 }
3041                                 pctx->errcode = ext2fs_extent_fix_parents(ehandle);
3042                                 if (pctx->errcode)
3043                                         goto failed_add_dir_block;
3044                                 pctx->errcode = ext2fs_extent_goto(ehandle,
3045                                                                 extent.e_lblk);
3046                                 if (pctx->errcode)
3047                                         goto failed_add_dir_block;
3048                                 last_lblk = extent.e_lblk + extent.e_len - 1;
3049                                 failed_csum = 0;
3050                         }
3051                 }
3052 alloc_later:
3053                 if (is_dir) {
3054                         while (++pb->last_db_block <
3055                                (e2_blkcnt_t) extent.e_lblk) {
3056                                 pctx->errcode = ext2fs_add_dir_block2(
3057                                                         ctx->fs->dblist,
3058                                                         pb->ino, 0,
3059                                                         pb->last_db_block);
3060                                 if (pctx->errcode) {
3061                                         pctx->blk = 0;
3062                                         pctx->num = pb->last_db_block;
3063                                         goto failed_add_dir_block;
3064                                 }
3065                         }
3066
3067                         for (i = 0; i < extent.e_len; i++) {
3068                                 pctx->errcode = ext2fs_add_dir_block2(
3069                                                         ctx->fs->dblist,
3070                                                         pctx->ino,
3071                                                         extent.e_pblk + i,
3072                                                         extent.e_lblk + i);
3073                                 if (pctx->errcode) {
3074                                         pctx->blk = extent.e_pblk + i;
3075                                         pctx->num = extent.e_lblk + i;
3076                                 failed_add_dir_block:
3077                                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3078                                         /* Should never get here */
3079                                         ctx->flags |= E2F_FLAG_ABORT;
3080                                         return;
3081                                 }
3082                         }
3083                         if (extent.e_len > 0)
3084                                 pb->last_db_block = extent.e_lblk + extent.e_len - 1;
3085                 }
3086                 if (has_unaligned_cluster_map(ctx, pb->previous_block,
3087                                               pb->last_block,
3088                                               extent.e_pblk,
3089                                               extent.e_lblk)) {
3090                         for (i = 0; i < extent.e_len; i++) {
3091                                 pctx->blk = extent.e_lblk + i;
3092                                 pctx->blk2 = extent.e_pblk + i;
3093                                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3094                                 mark_block_used(ctx, extent.e_pblk + i);
3095                                 mark_block_used(ctx, extent.e_pblk + i);
3096                         }
3097                 }
3098
3099                 /*
3100                  * Check whether first cluster got marked in previous iteration.
3101                  */
3102                 if (ctx->fs->cluster_ratio_bits &&
3103                     pb->previous_block &&
3104                     (EXT2FS_B2C(ctx->fs, extent.e_pblk) ==
3105                      EXT2FS_B2C(ctx->fs, pb->previous_block)))
3106                         /* Set blk to the beginning of next cluster. */
3107                         blk = EXT2FS_C2B(
3108                                 ctx->fs,
3109                                 EXT2FS_B2C(ctx->fs, extent.e_pblk) + 1);
3110                 else
3111                         /* Set blk to the beginning of current cluster. */
3112                         blk = EXT2FS_C2B(ctx->fs,
3113                                          EXT2FS_B2C(ctx->fs, extent.e_pblk));
3114
3115                 if (blk < extent.e_pblk + extent.e_len) {
3116                         mark_blocks_used(ctx, blk,
3117                                          extent.e_pblk + extent.e_len - blk);
3118                         n = DIV_ROUND_UP(extent.e_pblk + extent.e_len - blk,
3119                                          EXT2FS_CLUSTER_RATIO(ctx->fs));
3120                         pb->num_blocks += n;
3121                 }
3122                 pb->last_block = extent.e_lblk + extent.e_len - 1;
3123                 pb->previous_block = extent.e_pblk + extent.e_len - 1;
3124                 start_block = pb->last_block = last_lblk;
3125                 if (is_leaf && !is_dir &&
3126                     !(extent.e_flags & EXT2_EXTENT_FLAGS_UNINIT))
3127                         pb->last_init_lblock = last_lblk;
3128         next:
3129                 pctx->errcode = ext2fs_extent_get(ehandle,
3130                                                   EXT2_EXTENT_NEXT_SIB,
3131                                                   &extent);
3132         }
3133
3134         /* Failed csum but passes checks?  Ask to fix checksum. */
3135         if (failed_csum &&
3136             fix_problem(ctx, PR_1_EXTENT_ONLY_CSUM_INVALID, pctx)) {
3137                 pb->inode_modified = 1;
3138                 pctx->errcode = ext2fs_extent_replace(ehandle, 0, &extent);
3139                 if (pctx->errcode)
3140                         return;
3141         }
3142
3143         if (pctx->errcode == EXT2_ET_EXTENT_NO_NEXT)
3144                 pctx->errcode = 0;
3145 }
3146
3147 static void check_blocks_extents(e2fsck_t ctx, struct problem_context *pctx,
3148                                  struct process_block_struct *pb)
3149 {
3150         struct ext2_extent_info info;
3151         struct ext2_inode       *inode = pctx->inode;
3152         ext2_extent_handle_t    ehandle;
3153         ext2_filsys             fs = ctx->fs;
3154         ext2_ino_t              ino = pctx->ino;
3155         errcode_t               retval;
3156         blk64_t                 eof_lblk;
3157         struct ext3_extent_header       *eh;
3158
3159         /* Check for a proper extent header... */
3160         eh = (struct ext3_extent_header *) &inode->i_block[0];
3161         retval = ext2fs_extent_header_verify(eh, sizeof(inode->i_block));
3162         if (retval) {
3163                 if (fix_problem(ctx, PR_1_MISSING_EXTENT_HEADER, pctx))
3164                         e2fsck_clear_inode(ctx, ino, inode, 0,
3165                                            "check_blocks_extents");
3166                 pctx->errcode = 0;
3167                 return;
3168         }
3169
3170         /* ...since this function doesn't fail if i_block is zeroed. */
3171         pctx->errcode = ext2fs_extent_open2(fs, ino, inode, &ehandle);
3172         if (pctx->errcode) {
3173                 if (fix_problem(ctx, PR_1_READ_EXTENT, pctx))
3174                         e2fsck_clear_inode(ctx, ino, inode, 0,
3175                                            "check_blocks_extents");
3176                 pctx->errcode = 0;
3177                 return;
3178         }
3179
3180         retval = ext2fs_extent_get_info(ehandle, &info);
3181         if (retval == 0) {
3182                 int max_depth = info.max_depth;
3183
3184                 if (max_depth >= MAX_EXTENT_DEPTH_COUNT)
3185                         max_depth = MAX_EXTENT_DEPTH_COUNT-1;
3186                 ctx->extent_depth_count[max_depth]++;
3187         }
3188
3189         /* Check maximum extent depth */
3190         pctx->blk = info.max_depth;
3191         pctx->blk2 = ext2fs_max_extent_depth(ehandle);
3192         if (pctx->blk2 < pctx->blk &&
3193             fix_problem(ctx, PR_1_EXTENT_BAD_MAX_DEPTH, pctx))
3194                 pb->eti.force_rebuild = 1;
3195
3196         /* Can we collect extent tree level stats? */
3197         pctx->blk = MAX_EXTENT_DEPTH_COUNT;
3198         if (pctx->blk2 > pctx->blk)
3199                 fix_problem(ctx, PR_1E_MAX_EXTENT_TREE_DEPTH, pctx);
3200         memset(pb->eti.ext_info, 0, sizeof(pb->eti.ext_info));
3201         pb->eti.ino = pb->ino;
3202
3203         pb->next_lblock = 0;
3204
3205         eof_lblk = ((EXT2_I_SIZE(inode) + fs->blocksize - 1) >>
3206                 EXT2_BLOCK_SIZE_BITS(fs->super)) - 1;
3207         scan_extent_node(ctx, pctx, pb, 0, 0, eof_lblk, ehandle, 1);
3208         if (pctx->errcode &&
3209             fix_problem(ctx, PR_1_EXTENT_ITERATE_FAILURE, pctx)) {
3210                 pb->num_blocks = 0;
3211                 inode->i_blocks = 0;
3212                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3213                                    "check_blocks_extents");
3214                 pctx->errcode = 0;
3215         }
3216         ext2fs_extent_free(ehandle);
3217
3218         /* Rebuild unless it's a dir and we're rehashing it */
3219         if (LINUX_S_ISDIR(inode->i_mode) &&
3220             e2fsck_dir_will_be_rehashed(ctx, ino))
3221                 return;
3222
3223         if (ctx->options & E2F_OPT_CONVERT_BMAP)
3224                 e2fsck_rebuild_extents_later(ctx, ino);
3225         else
3226                 e2fsck_should_rebuild_extents(ctx, pctx, &pb->eti, &info);
3227 }
3228
3229 /*
3230  * In fact we don't need to check blocks for an inode with inline data
3231  * because this inode doesn't have any blocks.  In this function all
3232  * we need to do is add this inode into dblist when it is a directory.
3233  */
3234 static void check_blocks_inline_data(e2fsck_t ctx, struct problem_context *pctx,
3235                                      struct process_block_struct *pb)
3236 {
3237         int     flags;
3238         size_t  inline_data_size = 0;
3239
3240         if (!pb->is_dir) {
3241                 pctx->errcode = 0;
3242                 return;
3243         }
3244
3245         /* Process the dirents in i_block[] as the "first" block. */
3246         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 0);
3247         if (pctx->errcode)
3248                 goto err;
3249
3250         /* Process the dirents in the EA as a "second" block. */
3251         flags = ctx->fs->flags;
3252         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3253         pctx->errcode = ext2fs_inline_data_size(ctx->fs, pb->ino,
3254                                                 &inline_data_size);
3255         ctx->fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3256                          (ctx->fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3257         if (pctx->errcode) {
3258                 pctx->errcode = 0;
3259                 return;
3260         }
3261
3262         if (inline_data_size <= EXT4_MIN_INLINE_DATA_SIZE)
3263                 return;
3264
3265         pctx->errcode = ext2fs_add_dir_block2(ctx->fs->dblist, pb->ino, 0, 1);
3266         if (pctx->errcode)
3267                 goto err;
3268
3269         return;
3270 err:
3271         pctx->blk = 0;
3272         pctx->num = 0;
3273         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3274         ctx->flags |= E2F_FLAG_ABORT;
3275 }
3276
3277 /*
3278  * This subroutine is called on each inode to account for all of the
3279  * blocks used by that inode.
3280  */
3281 static void check_blocks(e2fsck_t ctx, struct problem_context *pctx,
3282                          char *block_buf, const struct ea_quota *ea_ibody_quota)
3283 {
3284         ext2_filsys fs = ctx->fs;
3285         struct process_block_struct pb;
3286         ext2_ino_t      ino = pctx->ino;
3287         struct ext2_inode *inode = pctx->inode;
3288         unsigned        bad_size = 0;
3289         int             dirty_inode = 0;
3290         int             extent_fs;
3291         int             inlinedata_fs;
3292         __u64           size;
3293         struct ea_quota ea_block_quota;
3294
3295         pb.ino = ino;
3296         pb.num_blocks = EXT2FS_B2C(ctx->fs,
3297                                    ea_ibody_quota ? ea_ibody_quota->blocks : 0);
3298         pb.last_block = ~0;
3299         pb.last_init_lblock = -1;
3300         pb.last_db_block = -1;
3301         pb.num_illegal_blocks = 0;
3302         pb.suppress = 0; pb.clear = 0;
3303         pb.fragmented = 0;
3304         pb.compressed = 0;
3305         pb.previous_block = 0;
3306         pb.is_dir = LINUX_S_ISDIR(inode->i_mode);
3307         pb.is_reg = LINUX_S_ISREG(inode->i_mode);
3308         pb.max_blocks = 1U << (31 - fs->super->s_log_block_size);
3309         pb.inode = inode;
3310         pb.pctx = pctx;
3311         pb.ctx = ctx;
3312         pb.inode_modified = 0;
3313         pb.eti.force_rebuild = 0;
3314         pctx->ino = ino;
3315         pctx->errcode = 0;
3316
3317         extent_fs = ext2fs_has_feature_extents(ctx->fs->super);
3318         inlinedata_fs = ext2fs_has_feature_inline_data(ctx->fs->super);
3319
3320         if (check_ext_attr(ctx, pctx, block_buf, &ea_block_quota)) {
3321                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3322                         goto out;
3323                 pb.num_blocks += EXT2FS_B2C(ctx->fs, ea_block_quota.blocks);
3324         }
3325
3326         if (inlinedata_fs && (inode->i_flags & EXT4_INLINE_DATA_FL))
3327                 check_blocks_inline_data(ctx, pctx, &pb);
3328         else if (ext2fs_inode_has_valid_blocks2(fs, inode)) {
3329                 if (extent_fs && (inode->i_flags & EXT4_EXTENTS_FL))
3330                         check_blocks_extents(ctx, pctx, &pb);
3331                 else {
3332                         int flags;
3333                         /*
3334                          * If we've modified the inode, write it out before
3335                          * iterate() tries to use it.
3336                          */
3337                         if (dirty_inode) {
3338                                 e2fsck_write_inode(ctx, ino, inode,
3339                                                    "check_blocks");
3340                                 dirty_inode = 0;
3341                         }
3342                         flags = fs->flags;
3343                         fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3344                         pctx->errcode = ext2fs_block_iterate3(fs, ino,
3345                                                 pb.is_dir ? BLOCK_FLAG_HOLE : 0,
3346                                                 block_buf, process_block, &pb);
3347                         /*
3348                          * We do not have uninitialized extents in non extent
3349                          * files.
3350                          */
3351                         pb.last_init_lblock = pb.last_block;
3352                         /*
3353                          * If iterate() changed a block mapping, we have to
3354                          * re-read the inode.  If we decide to clear the
3355                          * inode after clearing some stuff, we'll re-write the
3356                          * bad mappings into the inode!
3357                          */
3358                         if (pb.inode_modified)
3359                                 e2fsck_read_inode(ctx, ino, inode,
3360                                                   "check_blocks");
3361                         fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3362                                     (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3363
3364                         if (ctx->options & E2F_OPT_CONVERT_BMAP) {
3365 #ifdef DEBUG
3366                                 printf("bmap rebuild ino=%d\n", ino);
3367 #endif
3368                                 if (!LINUX_S_ISDIR(inode->i_mode) ||
3369                                     !e2fsck_dir_will_be_rehashed(ctx, ino))
3370                                         e2fsck_rebuild_extents_later(ctx, ino);
3371                         }
3372                 }
3373         }
3374         end_problem_latch(ctx, PR_LATCH_BLOCK);
3375         end_problem_latch(ctx, PR_LATCH_TOOBIG);
3376         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3377                 goto out;
3378         if (pctx->errcode)
3379                 fix_problem(ctx, PR_1_BLOCK_ITERATE, pctx);
3380
3381         if (pb.fragmented && pb.num_blocks < fs->super->s_blocks_per_group) {
3382                 if (LINUX_S_ISDIR(inode->i_mode))
3383                         ctx->fs_fragmented_dir++;
3384                 else
3385                         ctx->fs_fragmented++;
3386         }
3387
3388         if (pb.clear) {
3389                 e2fsck_clear_inode(ctx, ino, inode, E2F_FLAG_RESTART,
3390                                    "check_blocks");
3391                 return;
3392         }
3393
3394         if (inode->i_flags & EXT2_INDEX_FL) {
3395                 if (handle_htree(ctx, pctx, ino, inode, block_buf)) {
3396                         inode->i_flags &= ~EXT2_INDEX_FL;
3397                         dirty_inode++;
3398                 } else {
3399                         e2fsck_add_dx_dir(ctx, ino, inode, pb.last_block+1);
3400                 }
3401         }
3402
3403         if (!pb.num_blocks && pb.is_dir &&
3404             !(inode->i_flags & EXT4_INLINE_DATA_FL)) {
3405                 if (fix_problem(ctx, PR_1_ZERO_LENGTH_DIR, pctx)) {
3406                         e2fsck_clear_inode(ctx, ino, inode, 0, "check_blocks");
3407                         ctx->fs_directory_count--;
3408                         return;
3409                 }
3410         }
3411
3412         if (ino != quota_type2inum(PRJQUOTA, fs->super) &&
3413             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INODE(ctx->fs->super)) &&
3414             !(inode->i_flags & EXT4_EA_INODE_FL)) {
3415                 quota_data_add(ctx->qctx, (struct ext2_inode_large *) inode,
3416                                ino,
3417                                pb.num_blocks * EXT2_CLUSTER_SIZE(fs->super));
3418                 quota_data_inodes(ctx->qctx, (struct ext2_inode_large *) inode,
3419                                   ino, (ea_ibody_quota ?
3420                                         ea_ibody_quota->inodes : 0) +
3421                                                 ea_block_quota.inodes + 1);
3422         }
3423
3424         if (!ext2fs_has_feature_huge_file(fs->super) ||
3425             !(inode->i_flags & EXT4_HUGE_FILE_FL))
3426                 pb.num_blocks *= (fs->blocksize / 512);
3427         pb.num_blocks *= EXT2FS_CLUSTER_RATIO(fs);
3428 #if 0
3429         printf("inode %u, i_size = %u, last_block = %llu, i_blocks=%llu, num_blocks = %llu\n",
3430                ino, inode->i_size, pb.last_block, ext2fs_inode_i_blocks(fs, inode),
3431                pb.num_blocks);
3432 #endif
3433         if (pb.is_dir) {
3434                 unsigned nblock = inode->i_size >> EXT2_BLOCK_SIZE_BITS(fs->super);
3435                 if (inode->i_flags & EXT4_INLINE_DATA_FL) {
3436                         int flags;
3437                         size_t sz = 0;
3438                         errcode_t err;
3439
3440                         flags = ctx->fs->flags;
3441                         ctx->fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
3442                         err = ext2fs_inline_data_size(ctx->fs, pctx->ino,
3443                                                       &sz);
3444                         ctx->fs->flags = (flags &
3445                                           EXT2_FLAG_IGNORE_CSUM_ERRORS) |
3446                                          (ctx->fs->flags &
3447                                           ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
3448                         if (err || sz != inode->i_size) {
3449                                 bad_size = 7;
3450                                 pctx->num = sz;
3451                         }
3452                 } else if (inode->i_size & (fs->blocksize - 1))
3453                         bad_size = 5;
3454                 else if (nblock > (pb.last_block + 1))
3455                         bad_size = 1;
3456                 else if (nblock < (pb.last_block + 1)) {
3457                         if (((pb.last_block + 1) - nblock) >
3458                             fs->super->s_prealloc_dir_blocks)
3459                                 bad_size = 2;
3460                 }
3461         } else {
3462                 size = EXT2_I_SIZE(inode);
3463                 if ((pb.last_init_lblock >= 0) &&
3464                     /* Do not allow initialized allocated blocks past i_size*/
3465                     (size < (__u64)pb.last_init_lblock * fs->blocksize) &&
3466                     !(inode->i_flags & EXT4_VERITY_FL))
3467                         bad_size = 3;
3468                 else if (!(extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3469                          size > ext2_max_sizes[fs->super->s_log_block_size])
3470                         /* too big for a direct/indirect-mapped file */
3471                         bad_size = 4;
3472                 else if ((extent_fs && (inode->i_flags & EXT4_EXTENTS_FL)) &&
3473                          size >
3474                          ((1ULL << (32 + EXT2_BLOCK_SIZE_BITS(fs->super))) - 1))
3475                         /* too big for an extent-based file - 32bit ee_block */
3476                         bad_size = 6;
3477         }
3478         /* i_size for symlinks is checked elsewhere */
3479         if (bad_size && !LINUX_S_ISLNK(inode->i_mode)) {
3480                 /* Did inline_data set pctx->num earlier? */
3481                 if (bad_size != 7)
3482                         pctx->num = (pb.last_block + 1) * fs->blocksize;
3483                 pctx->group = bad_size;
3484                 if (fix_problem(ctx, PR_1_BAD_I_SIZE, pctx)) {
3485                         if (LINUX_S_ISDIR(inode->i_mode))
3486                                 pctx->num &= 0xFFFFFFFFULL;
3487                         ext2fs_inode_size_set(fs, inode, pctx->num);
3488                         if (EXT2_I_SIZE(inode) == 0 &&
3489                             (inode->i_flags & EXT4_INLINE_DATA_FL)) {
3490                                 memset(inode->i_block, 0,
3491                                        sizeof(inode->i_block));
3492                                 inode->i_flags &= ~EXT4_INLINE_DATA_FL;
3493                         }
3494                         dirty_inode++;
3495                 }
3496                 pctx->num = 0;
3497         }
3498         if (LINUX_S_ISREG(inode->i_mode) &&
3499             ext2fs_needs_large_file_feature(EXT2_I_SIZE(inode)))
3500                 ctx->large_files++;
3501         if ((fs->super->s_creator_os != EXT2_OS_HURD) &&
3502             ((pb.num_blocks != ext2fs_inode_i_blocks(fs, inode)) ||
3503              (ext2fs_has_feature_huge_file(fs->super) &&
3504               (inode->i_flags & EXT4_HUGE_FILE_FL) &&
3505               (inode->osd2.linux2.l_i_blocks_hi != 0)))) {
3506                 pctx->num = pb.num_blocks;
3507                 if (fix_problem(ctx, PR_1_BAD_I_BLOCKS, pctx)) {
3508                         inode->i_blocks = pb.num_blocks;
3509                         inode->osd2.linux2.l_i_blocks_hi = pb.num_blocks >> 32;
3510                         dirty_inode++;
3511                 }
3512                 pctx->num = 0;
3513         }
3514
3515         /*
3516          * The kernel gets mad if we ask it to allocate bigalloc clusters to
3517          * a block mapped file, so rebuild it as an extent file.  We can skip
3518          * symlinks because they're never rewritten.
3519          */
3520         if (ext2fs_has_feature_bigalloc(fs->super) &&
3521             (LINUX_S_ISREG(inode->i_mode) || LINUX_S_ISDIR(inode->i_mode)) &&
3522             ext2fs_inode_data_blocks2(fs, inode) > 0 &&
3523             (ino == EXT2_ROOT_INO || ino >= EXT2_FIRST_INO(fs->super)) &&
3524             !(inode->i_flags & (EXT4_EXTENTS_FL | EXT4_INLINE_DATA_FL)) &&
3525             fix_problem(ctx, PR_1_NO_BIGALLOC_BLOCKMAP_FILES, pctx)) {
3526                 pctx->errcode = e2fsck_rebuild_extents_later(ctx, ino);
3527                 if (pctx->errcode)
3528                         goto out;
3529         }
3530
3531         if (ctx->dirs_to_hash && pb.is_dir &&
3532             !(ctx->lost_and_found && ctx->lost_and_found == ino) &&
3533             !(inode->i_flags & EXT2_INDEX_FL) &&
3534             ((inode->i_size / fs->blocksize) >= 3))
3535                 e2fsck_rehash_dir_later(ctx, ino);
3536
3537 out:
3538         if (dirty_inode)
3539                 e2fsck_write_inode(ctx, ino, inode, "check_blocks");
3540 }
3541
3542 #if 0
3543 /*
3544  * Helper function called by process block when an illegal block is
3545  * found.  It returns a description about why the block is illegal
3546  */
3547 static char *describe_illegal_block(ext2_filsys fs, blk64_t block)
3548 {
3549         blk64_t super;
3550         int     i;
3551         static char     problem[80];
3552
3553         super = fs->super->s_first_data_block;
3554         strcpy(problem, "PROGRAMMING ERROR: Unknown reason for illegal block");
3555         if (block < super) {
3556                 sprintf(problem, "< FIRSTBLOCK (%u)", super);
3557                 return(problem);
3558         } else if (block >= ext2fs_blocks_count(fs->super)) {
3559                 sprintf(problem, "> BLOCKS (%u)", ext2fs_blocks_count(fs->super));
3560                 return(problem);
3561         }
3562         for (i = 0; i < fs->group_desc_count; i++) {
3563                 if (block == super) {
3564                         sprintf(problem, "is the superblock in group %d", i);
3565                         break;
3566                 }
3567                 if (block > super &&
3568                     block <= (super + fs->desc_blocks)) {
3569                         sprintf(problem, "is in the group descriptors "
3570                                 "of group %d", i);
3571                         break;
3572                 }
3573                 if (block == ext2fs_block_bitmap_loc(fs, i)) {
3574                         sprintf(problem, "is the block bitmap of group %d", i);
3575                         break;
3576                 }
3577                 if (block == ext2fs_inode_bitmap_loc(fs, i)) {
3578                         sprintf(problem, "is the inode bitmap of group %d", i);
3579                         break;
3580                 }
3581                 if (block >= ext2fs_inode_table_loc(fs, i) &&
3582                     (block < ext2fs_inode_table_loc(fs, i)
3583                      + fs->inode_blocks_per_group)) {
3584                         sprintf(problem, "is in the inode table of group %d",
3585                                 i);
3586                         break;
3587                 }
3588                 super += fs->super->s_blocks_per_group;
3589         }
3590         return(problem);
3591 }
3592 #endif
3593
3594 /*
3595  * This is a helper function for check_blocks().
3596  */
3597 static int process_block(ext2_filsys fs,
3598                   blk64_t       *block_nr,
3599                   e2_blkcnt_t blockcnt,
3600                   blk64_t ref_block EXT2FS_ATTR((unused)),
3601                   int ref_offset EXT2FS_ATTR((unused)),
3602                   void *priv_data)
3603 {
3604         struct process_block_struct *p;
3605         struct problem_context *pctx;
3606         blk64_t blk = *block_nr;
3607         int     ret_code = 0;
3608         problem_t       problem = 0;
3609         e2fsck_t        ctx;
3610
3611         p = (struct process_block_struct *) priv_data;
3612         pctx = p->pctx;
3613         ctx = p->ctx;
3614
3615         /*
3616          * For a directory, add logical block zero for processing even if it's
3617          * not mapped or we'll be perennially stuck with broken "." and ".."
3618          * entries.
3619          */
3620         if (p->is_dir && blockcnt == 0 && blk == 0) {
3621                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino, 0, 0);
3622                 if (pctx->errcode) {
3623                         pctx->blk = blk;
3624                         pctx->num = blockcnt;
3625                         goto failed_add_dir_block;
3626                 }
3627                 p->last_db_block++;
3628         }
3629
3630         if (blk == 0)
3631                 return 0;
3632
3633 #if 0
3634         printf("Process_block, inode %lu, block %u, #%d\n", p->ino, blk,
3635                blockcnt);
3636 #endif
3637
3638         /*
3639          * Simplistic fragmentation check.  We merely require that the
3640          * file be contiguous.  (Which can never be true for really
3641          * big files that are greater than a block group.)
3642          */
3643         if (p->previous_block && p->ino != EXT2_RESIZE_INO) {
3644                 if (p->previous_block+1 != blk) {
3645                         if (ctx->options & E2F_OPT_FRAGCHECK) {
3646                                 char type = '?';
3647
3648                                 if (p->is_dir)
3649                                         type = 'd';
3650                                 else if (p->is_reg)
3651                                         type = 'f';
3652
3653                                 printf(_("%6lu(%c): expecting %6lu "
3654                                          "got phys %6lu (blkcnt %lld)\n"),
3655                                        (unsigned long) pctx->ino, type,
3656                                        (unsigned long) p->previous_block+1,
3657                                        (unsigned long) blk,
3658                                        blockcnt);
3659                         }
3660                         p->fragmented = 1;
3661                 }
3662         }
3663
3664         if (p->is_dir && !ext2fs_has_feature_largedir(fs->super) &&
3665             blockcnt > (1 << (21 - fs->super->s_log_block_size)))
3666                 problem = PR_1_TOOBIG_DIR;
3667         if (p->is_dir && p->num_blocks + 1 >= p->max_blocks)
3668                 problem = PR_1_TOOBIG_DIR;
3669         if (p->is_reg && p->num_blocks + 1 >= p->max_blocks)
3670                 problem = PR_1_TOOBIG_REG;
3671         if (!p->is_dir && !p->is_reg && blockcnt > 0)
3672                 problem = PR_1_TOOBIG_SYMLINK;
3673
3674         if (blk < fs->super->s_first_data_block ||
3675             blk >= ext2fs_blocks_count(fs->super))
3676                 problem = PR_1_ILLEGAL_BLOCK_NUM;
3677
3678         /*
3679          * If this IND/DIND/TIND block is squatting atop some critical metadata
3680          * (group descriptors, superblock, bitmap, inode table), any write to
3681          * "fix" mapping problems will destroy the metadata.  We'll let pass 1b
3682          * fix that and restart fsck.
3683          */
3684         if (blockcnt < 0 &&
3685             p->ino != EXT2_RESIZE_INO &&
3686             blk < ctx->fs->super->s_blocks_count &&
3687             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk)) {
3688                 pctx->blk = blk;
3689                 fix_problem(ctx, PR_1_CRITICAL_METADATA_COLLISION, pctx);
3690                 if ((ctx->options & E2F_OPT_NO) == 0)
3691                         ctx->flags |= E2F_FLAG_RESTART_LATER;
3692         }
3693
3694         if (problem) {
3695                 p->num_illegal_blocks++;
3696                 /*
3697                  * A bit of subterfuge here -- we're trying to fix a block
3698                  * mapping, but the IND/DIND/TIND block could have collided
3699                  * with some critical metadata.  So, fix the in-core mapping so
3700                  * iterate won't go insane, but return 0 instead of
3701                  * BLOCK_CHANGED so that it won't write the remapping out to
3702                  * our multiply linked block.
3703                  *
3704                  * Even if we previously determined that an *IND block
3705                  * conflicts with critical metadata, we must still try to
3706                  * iterate the *IND block as if it is an *IND block to find and
3707                  * mark the blocks it points to.  Better to be overly cautious
3708                  * with the used_blocks map so that we don't move the *IND
3709                  * block to a block that's really in use!
3710                  */
3711                 if (p->ino != EXT2_RESIZE_INO &&
3712                     ref_block != 0 &&
3713                     ext2fs_test_block_bitmap2(ctx->block_metadata_map,
3714                                               ref_block)) {
3715                         *block_nr = 0;
3716                         return 0;
3717                 }
3718                 if (!p->suppress && (p->num_illegal_blocks % 12) == 0) {
3719                         if (fix_problem(ctx, PR_1_TOO_MANY_BAD_BLOCKS, pctx)) {
3720                                 p->clear = 1;
3721                                 return BLOCK_ABORT;
3722                         }
3723                         if (fix_problem(ctx, PR_1_SUPPRESS_MESSAGES, pctx)) {
3724                                 p->suppress = 1;
3725                                 set_latch_flags(PR_LATCH_BLOCK,
3726                                                 PRL_SUPPRESS, 0);
3727                         }
3728                 }
3729                 pctx->blk = blk;
3730                 pctx->blkcount = blockcnt;
3731                 if (fix_problem(ctx, problem, pctx)) {
3732                         blk = *block_nr = 0;
3733                         ret_code = BLOCK_CHANGED;
3734                         p->inode_modified = 1;
3735                         /*
3736                          * If the directory block is too big and is beyond the
3737                          * end of the FS, don't bother trying to add it for
3738                          * processing -- the kernel would never have created a
3739                          * directory this large, and we risk an ENOMEM abort.
3740                          * In any case, the toobig handler for extent-based
3741                          * directories also doesn't feed toobig blocks to
3742                          * pass 2.
3743                          */
3744                         if (problem == PR_1_TOOBIG_DIR)
3745                                 return ret_code;
3746                         goto mark_dir;
3747                 } else
3748                         return 0;
3749         }
3750
3751         if (p->ino == EXT2_RESIZE_INO) {
3752                 /*
3753                  * The resize inode has already be sanity checked
3754                  * during pass #0 (the superblock checks).  All we
3755                  * have to do is mark the double indirect block as
3756                  * being in use; all of the other blocks are handled
3757                  * by mark_table_blocks()).
3758                  */
3759                 if (blockcnt == BLOCK_COUNT_DIND)
3760                         mark_block_used(ctx, blk);
3761                 p->num_blocks++;
3762         } else if (!(ctx->fs->cluster_ratio_bits &&
3763                      p->previous_block &&
3764                      (EXT2FS_B2C(ctx->fs, blk) ==
3765                       EXT2FS_B2C(ctx->fs, p->previous_block)) &&
3766                      (blk & EXT2FS_CLUSTER_MASK(ctx->fs)) ==
3767                      ((unsigned) blockcnt & EXT2FS_CLUSTER_MASK(ctx->fs)))) {
3768                 mark_block_used(ctx, blk);
3769                 p->num_blocks++;
3770         } else if (has_unaligned_cluster_map(ctx, p->previous_block,
3771                                              p->last_block, blk, blockcnt)) {
3772                 pctx->blk = blockcnt;
3773                 pctx->blk2 = blk;
3774                 fix_problem(ctx, PR_1_MISALIGNED_CLUSTER, pctx);
3775                 mark_block_used(ctx, blk);
3776                 mark_block_used(ctx, blk);
3777         }
3778         if (blockcnt >= 0)
3779                 p->last_block = blockcnt;
3780         p->previous_block = blk;
3781 mark_dir:
3782         if (p->is_dir && (blockcnt >= 0)) {
3783                 while (++p->last_db_block < blockcnt) {
3784                         pctx->errcode = ext2fs_add_dir_block2(fs->dblist,
3785                                                               p->ino, 0,
3786                                                               p->last_db_block);
3787                         if (pctx->errcode) {
3788                                 pctx->blk = 0;
3789                                 pctx->num = p->last_db_block;
3790                                 goto failed_add_dir_block;
3791                         }
3792                 }
3793                 pctx->errcode = ext2fs_add_dir_block2(fs->dblist, p->ino,
3794                                                       blk, blockcnt);
3795                 if (pctx->errcode) {
3796                         pctx->blk = blk;
3797                         pctx->num = blockcnt;
3798                 failed_add_dir_block:
3799                         fix_problem(ctx, PR_1_ADD_DBLOCK, pctx);
3800                         /* Should never get here */
3801                         ctx->flags |= E2F_FLAG_ABORT;
3802                         return BLOCK_ABORT;
3803                 }
3804         }
3805         return ret_code;
3806 }
3807
3808 static int process_bad_block(ext2_filsys fs,
3809                       blk64_t *block_nr,
3810                       e2_blkcnt_t blockcnt,
3811                       blk64_t ref_block EXT2FS_ATTR((unused)),
3812                       int ref_offset EXT2FS_ATTR((unused)),
3813                       void *priv_data)
3814 {
3815         struct process_block_struct *p;
3816         blk64_t         blk = *block_nr;
3817         blk64_t         first_block;
3818         dgrp_t          i;
3819         struct problem_context *pctx;
3820         e2fsck_t        ctx;
3821
3822         if (!blk)
3823                 return 0;
3824
3825         p = (struct process_block_struct *) priv_data;
3826         ctx = p->ctx;
3827         pctx = p->pctx;
3828
3829         pctx->ino = EXT2_BAD_INO;
3830         pctx->blk = blk;
3831         pctx->blkcount = blockcnt;
3832
3833         if ((blk < fs->super->s_first_data_block) ||
3834             (blk >= ext2fs_blocks_count(fs->super))) {
3835                 if (fix_problem(ctx, PR_1_BB_ILLEGAL_BLOCK_NUM, pctx)) {
3836                         *block_nr = 0;
3837                         return BLOCK_CHANGED;
3838                 } else
3839                         return 0;
3840         }
3841
3842         if (blockcnt < 0) {
3843                 if (ext2fs_test_block_bitmap2(p->fs_meta_blocks, blk)) {
3844                         p->bbcheck = 1;
3845                         if (fix_problem(ctx, PR_1_BB_FS_BLOCK, pctx)) {
3846                                 *block_nr = 0;
3847                                 return BLOCK_CHANGED;
3848                         }
3849                 } else if (ext2fs_test_block_bitmap2(ctx->block_found_map,
3850                                                     blk)) {
3851                         p->bbcheck = 1;
3852                         if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK,
3853                                         pctx)) {
3854                                 *block_nr = 0;
3855                                 return BLOCK_CHANGED;
3856                         }
3857                         if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3858                                 return BLOCK_ABORT;
3859                 } else
3860                         mark_block_used(ctx, blk);
3861                 return 0;
3862         }
3863 #if 0
3864         printf ("DEBUG: Marking %u as bad.\n", blk);
3865 #endif
3866         ctx->fs_badblocks_count++;
3867         /*
3868          * If the block is not used, then mark it as used and return.
3869          * If it is already marked as found, this must mean that
3870          * there's an overlap between the filesystem table blocks
3871          * (bitmaps and inode table) and the bad block list.
3872          */
3873         if (!ext2fs_test_block_bitmap2(ctx->block_found_map, blk)) {
3874                 ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
3875                 return 0;
3876         }
3877         /*
3878          * Try to find the where the filesystem block was used...
3879          */
3880         first_block = fs->super->s_first_data_block;
3881
3882         for (i = 0; i < fs->group_desc_count; i++ ) {
3883                 pctx->group = i;
3884                 pctx->blk = blk;
3885                 if (!ext2fs_bg_has_super(fs, i))
3886                         goto skip_super;
3887                 if (blk == first_block) {
3888                         if (i == 0) {
3889                                 if (fix_problem(ctx,
3890                                                 PR_1_BAD_PRIMARY_SUPERBLOCK,
3891                                                 pctx)) {
3892                                         *block_nr = 0;
3893                                         return BLOCK_CHANGED;
3894                                 }
3895                                 return 0;
3896                         }
3897                         fix_problem(ctx, PR_1_BAD_SUPERBLOCK, pctx);
3898                         return 0;
3899                 }
3900                 if ((blk > first_block) &&
3901                     (blk <= first_block + fs->desc_blocks)) {
3902                         if (i == 0) {
3903                                 pctx->blk = *block_nr;
3904                                 if (fix_problem(ctx,
3905                         PR_1_BAD_PRIMARY_GROUP_DESCRIPTOR, pctx)) {
3906                                         *block_nr = 0;
3907                                         return BLOCK_CHANGED;
3908                                 }
3909                                 return 0;
3910                         }
3911                         fix_problem(ctx, PR_1_BAD_GROUP_DESCRIPTORS, pctx);
3912                         return 0;
3913                 }
3914         skip_super:
3915                 if (blk == ext2fs_block_bitmap_loc(fs, i)) {
3916                         if (fix_problem(ctx, PR_1_BB_BAD_BLOCK, pctx)) {
3917                                 ctx->invalid_block_bitmap_flag[i]++;
3918                                 ctx->invalid_bitmaps++;
3919                         }
3920                         return 0;
3921                 }
3922                 if (blk == ext2fs_inode_bitmap_loc(fs, i)) {
3923                         if (fix_problem(ctx, PR_1_IB_BAD_BLOCK, pctx)) {
3924                                 ctx->invalid_inode_bitmap_flag[i]++;
3925                                 ctx->invalid_bitmaps++;
3926                         }
3927                         return 0;
3928                 }
3929                 if ((blk >= ext2fs_inode_table_loc(fs, i)) &&
3930                     (blk < (ext2fs_inode_table_loc(fs, i) +
3931                             fs->inode_blocks_per_group))) {
3932                         /*
3933                          * If there are bad blocks in the inode table,
3934                          * the inode scan code will try to do
3935                          * something reasonable automatically.
3936                          */
3937                         return 0;
3938                 }
3939                 first_block += fs->super->s_blocks_per_group;
3940         }
3941         /*
3942          * If we've gotten to this point, then the only
3943          * possibility is that the bad block inode meta data
3944          * is using a bad block.
3945          */
3946         if ((blk == p->inode->i_block[EXT2_IND_BLOCK]) ||
3947             (blk == p->inode->i_block[EXT2_DIND_BLOCK]) ||
3948             (blk == p->inode->i_block[EXT2_TIND_BLOCK])) {
3949                 p->bbcheck = 1;
3950                 if (fix_problem(ctx, PR_1_BBINODE_BAD_METABLOCK, pctx)) {
3951                         *block_nr = 0;
3952                         return BLOCK_CHANGED;
3953                 }
3954                 if (ctx->flags & E2F_FLAG_SIGNAL_MASK)
3955                         return BLOCK_ABORT;
3956                 return 0;
3957         }
3958
3959         pctx->group = -1;
3960
3961         /* Warn user that the block wasn't claimed */
3962         fix_problem(ctx, PR_1_PROGERR_CLAIMED_BLOCK, pctx);
3963
3964         return 0;
3965 }
3966
3967 static void new_table_block(e2fsck_t ctx, blk64_t first_block, dgrp_t group,
3968                             const char *name, int num, blk64_t *new_block)
3969 {
3970         ext2_filsys fs = ctx->fs;
3971         dgrp_t          last_grp;
3972         blk64_t         old_block = *new_block;
3973         blk64_t         last_block;
3974         dgrp_t          flexbg;
3975         unsigned        flexbg_size;
3976         int             i, is_flexbg;
3977         char            *buf;
3978         struct problem_context  pctx;
3979
3980         clear_problem_context(&pctx);
3981
3982         pctx.group = group;
3983         pctx.blk = old_block;
3984         pctx.str = name;
3985
3986         /*
3987          * For flex_bg filesystems, first try to allocate the metadata
3988          * within the flex_bg, and if that fails then try finding the
3989          * space anywhere in the filesystem.
3990          */
3991         is_flexbg = ext2fs_has_feature_flex_bg(fs->super);
3992         if (is_flexbg) {
3993                 flexbg_size = 1 << fs->super->s_log_groups_per_flex;
3994                 flexbg = group / flexbg_size;
3995                 first_block = ext2fs_group_first_block2(fs,
3996                                                         flexbg_size * flexbg);
3997                 last_grp = group | (flexbg_size - 1);
3998                 if (last_grp >= fs->group_desc_count)
3999                         last_grp = fs->group_desc_count - 1;
4000                 last_block = ext2fs_group_last_block2(fs, last_grp);
4001         } else
4002                 last_block = ext2fs_group_last_block2(fs, group);
4003         pctx.errcode = ext2fs_get_free_blocks2(fs, first_block, last_block,
4004                                                num, ctx->block_found_map,
4005                                                new_block);
4006         if (is_flexbg && (pctx.errcode == EXT2_ET_BLOCK_ALLOC_FAIL))
4007                 pctx.errcode = ext2fs_get_free_blocks2(fs,
4008                                 fs->super->s_first_data_block,
4009                                 ext2fs_blocks_count(fs->super),
4010                                 num, ctx->block_found_map, new_block);
4011         if (pctx.errcode) {
4012                 pctx.num = num;
4013                 fix_problem(ctx, PR_1_RELOC_BLOCK_ALLOCATE, &pctx);
4014                 ext2fs_unmark_valid(fs);
4015                 ctx->flags |= E2F_FLAG_ABORT;
4016                 return;
4017         }
4018         pctx.errcode = ext2fs_get_mem(fs->blocksize, &buf);
4019         if (pctx.errcode) {
4020                 fix_problem(ctx, PR_1_RELOC_MEMORY_ALLOCATE, &pctx);
4021                 ext2fs_unmark_valid(fs);
4022                 ctx->flags |= E2F_FLAG_ABORT;
4023                 return;
4024         }
4025         ext2fs_mark_super_dirty(fs);
4026         fs->flags &= ~EXT2_FLAG_MASTER_SB_ONLY;
4027         pctx.blk2 = *new_block;
4028         fix_problem(ctx, (old_block ? PR_1_RELOC_FROM_TO :
4029                           PR_1_RELOC_TO), &pctx);
4030         pctx.blk2 = 0;
4031         for (i = 0; i < num; i++) {
4032                 pctx.blk = i;
4033                 ext2fs_mark_block_bitmap2(ctx->block_found_map, (*new_block)+i);
4034                 if (old_block) {
4035                         pctx.errcode = io_channel_read_blk64(fs->io,
4036                                    old_block + i, 1, buf);
4037                         if (pctx.errcode)
4038                                 fix_problem(ctx, PR_1_RELOC_READ_ERR, &pctx);
4039                         pctx.blk = (*new_block) + i;
4040                         pctx.errcode = io_channel_write_blk64(fs->io, pctx.blk,
4041                                                               1, buf);
4042                 } else {
4043                         pctx.blk = (*new_block) + i;
4044                         pctx.errcode = ext2fs_zero_blocks2(fs, pctx.blk, 1,
4045                                                            NULL, NULL);
4046                 }
4047
4048                 if (pctx.errcode)
4049                         fix_problem(ctx, PR_1_RELOC_WRITE_ERR, &pctx);
4050         }
4051         ext2fs_free_mem(&buf);
4052 }
4053
4054 /*
4055  * This routine gets called at the end of pass 1 if bad blocks are
4056  * detected in the superblock, group descriptors, inode_bitmaps, or
4057  * block bitmaps.  At this point, all of the blocks have been mapped
4058  * out, so we can try to allocate new block(s) to replace the bad
4059  * blocks.
4060  */
4061 static void handle_fs_bad_blocks(e2fsck_t ctx)
4062 {
4063         ext2_filsys fs = ctx->fs;
4064         dgrp_t          i;
4065         blk64_t         first_block;
4066         blk64_t         new_blk;
4067
4068         for (i = 0; i < fs->group_desc_count; i++) {
4069                 first_block = ext2fs_group_first_block2(fs, i);
4070
4071                 if (ctx->invalid_block_bitmap_flag[i]) {
4072                         new_blk = ext2fs_block_bitmap_loc(fs, i);
4073                         new_table_block(ctx, first_block, i, _("block bitmap"),
4074                                         1, &new_blk);
4075                         ext2fs_block_bitmap_loc_set(fs, i, new_blk);
4076                 }
4077                 if (ctx->invalid_inode_bitmap_flag[i]) {
4078                         new_blk = ext2fs_inode_bitmap_loc(fs, i);
4079                         new_table_block(ctx, first_block, i, _("inode bitmap"),
4080                                         1, &new_blk);
4081                         ext2fs_inode_bitmap_loc_set(fs, i, new_blk);
4082                 }
4083                 if (ctx->invalid_inode_table_flag[i]) {
4084                         new_blk = ext2fs_inode_table_loc(fs, i);
4085                         new_table_block(ctx, first_block, i, _("inode table"),
4086                                         fs->inode_blocks_per_group,
4087                                         &new_blk);
4088                         ext2fs_inode_table_loc_set(fs, i, new_blk);
4089                         ctx->flags |= E2F_FLAG_RESTART;
4090                 }
4091         }
4092         ctx->invalid_bitmaps = 0;
4093 }
4094
4095 /*
4096  * This routine marks all blocks which are used by the superblock,
4097  * group descriptors, inode bitmaps, and block bitmaps.
4098  */
4099 static void mark_table_blocks(e2fsck_t ctx)
4100 {
4101         ext2_filsys fs = ctx->fs;
4102         blk64_t b;
4103         dgrp_t  i;
4104         unsigned int    j;
4105         struct problem_context pctx;
4106
4107         clear_problem_context(&pctx);
4108
4109         for (i = 0; i < fs->group_desc_count; i++) {
4110                 pctx.group = i;
4111
4112                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_found_map);
4113                 ext2fs_reserve_super_and_bgd(fs, i, ctx->block_metadata_map);
4114
4115                 /*
4116                  * Mark the blocks used for the inode table
4117                  */
4118                 if (ext2fs_inode_table_loc(fs, i)) {
4119                         for (j = 0, b = ext2fs_inode_table_loc(fs, i);
4120                              j < fs->inode_blocks_per_group;
4121                              j++, b++) {
4122                                 if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4123                                                              b)) {
4124                                         pctx.blk = b;
4125                                         if (!ctx->invalid_inode_table_flag[i] &&
4126                                             fix_problem(ctx,
4127                                                 PR_1_ITABLE_CONFLICT, &pctx)) {
4128                                                 ctx->invalid_inode_table_flag[i]++;
4129                                                 ctx->invalid_bitmaps++;
4130                                         }
4131                                 } else {
4132                                     ext2fs_mark_block_bitmap2(
4133                                                 ctx->block_found_map, b);
4134                                     ext2fs_mark_block_bitmap2(
4135                                                 ctx->block_metadata_map, b);
4136                                 }
4137                         }
4138                 }
4139
4140                 /*
4141                  * Mark block used for the block bitmap
4142                  */
4143                 if (ext2fs_block_bitmap_loc(fs, i)) {
4144                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4145                                      ext2fs_block_bitmap_loc(fs, i))) {
4146                                 pctx.blk = ext2fs_block_bitmap_loc(fs, i);
4147                                 if (fix_problem(ctx, PR_1_BB_CONFLICT, &pctx)) {
4148                                         ctx->invalid_block_bitmap_flag[i]++;
4149                                         ctx->invalid_bitmaps++;
4150                                 }
4151                         } else {
4152                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4153                                      ext2fs_block_bitmap_loc(fs, i));
4154                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4155                                      ext2fs_block_bitmap_loc(fs, i));
4156                         }
4157                 }
4158                 /*
4159                  * Mark block used for the inode bitmap
4160                  */
4161                 if (ext2fs_inode_bitmap_loc(fs, i)) {
4162                         if (ext2fs_test_block_bitmap2(ctx->block_found_map,
4163                                      ext2fs_inode_bitmap_loc(fs, i))) {
4164                                 pctx.blk = ext2fs_inode_bitmap_loc(fs, i);
4165                                 if (fix_problem(ctx, PR_1_IB_CONFLICT, &pctx)) {
4166                                         ctx->invalid_inode_bitmap_flag[i]++;
4167                                         ctx->invalid_bitmaps++;
4168                                 }
4169                         } else {
4170                             ext2fs_mark_block_bitmap2(ctx->block_metadata_map,
4171                                      ext2fs_inode_bitmap_loc(fs, i));
4172                             ext2fs_mark_block_bitmap2(ctx->block_found_map,
4173                                      ext2fs_inode_bitmap_loc(fs, i));
4174                         }
4175                 }
4176         }
4177 }
4178
4179 /*
4180  * These subroutines short circuits ext2fs_get_blocks and
4181  * ext2fs_check_directory; we use them since we already have the inode
4182  * structure, so there's no point in letting the ext2fs library read
4183  * the inode again.
4184  */
4185 static errcode_t pass1_get_blocks(ext2_filsys fs, ext2_ino_t ino,
4186                                   blk_t *blocks)
4187 {
4188         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4189         int     i;
4190
4191         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4192                 return EXT2_ET_CALLBACK_NOTHANDLED;
4193
4194         for (i=0; i < EXT2_N_BLOCKS; i++)
4195                 blocks[i] = ctx->stashed_inode->i_block[i];
4196         return 0;
4197 }
4198
4199 static errcode_t pass1_read_inode(ext2_filsys fs, ext2_ino_t ino,
4200                                   struct ext2_inode *inode)
4201 {
4202         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4203
4204         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4205                 return EXT2_ET_CALLBACK_NOTHANDLED;
4206         *inode = *ctx->stashed_inode;
4207         return 0;
4208 }
4209
4210 static errcode_t pass1_write_inode(ext2_filsys fs, ext2_ino_t ino,
4211                             struct ext2_inode *inode)
4212 {
4213         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4214
4215         if ((ino == ctx->stashed_ino) && ctx->stashed_inode &&
4216                 (inode != ctx->stashed_inode))
4217                 *ctx->stashed_inode = *inode;
4218         return EXT2_ET_CALLBACK_NOTHANDLED;
4219 }
4220
4221 static errcode_t pass1_check_directory(ext2_filsys fs, ext2_ino_t ino)
4222 {
4223         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4224
4225         if ((ino != ctx->stashed_ino) || !ctx->stashed_inode)
4226                 return EXT2_ET_CALLBACK_NOTHANDLED;
4227
4228         if (!LINUX_S_ISDIR(ctx->stashed_inode->i_mode))
4229                 return EXT2_ET_NO_DIRECTORY;
4230         return 0;
4231 }
4232
4233 static errcode_t e2fsck_get_alloc_block(ext2_filsys fs, blk64_t goal,
4234                                         blk64_t *ret)
4235 {
4236         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4237         errcode_t       retval;
4238         blk64_t         new_block;
4239
4240         if (ctx->block_found_map) {
4241                 retval = ext2fs_new_block2(fs, goal, ctx->block_found_map,
4242                                            &new_block);
4243                 if (retval)
4244                         return retval;
4245                 if (fs->block_map) {
4246                         ext2fs_mark_block_bitmap2(fs->block_map, new_block);
4247                         ext2fs_mark_bb_dirty(fs);
4248                 }
4249         } else {
4250                 if (!fs->block_map) {
4251                         retval = ext2fs_read_block_bitmap(fs);
4252                         if (retval)
4253                                 return retval;
4254                 }
4255
4256                 retval = ext2fs_new_block2(fs, goal, fs->block_map, &new_block);
4257                 if (retval)
4258                         return retval;
4259         }
4260
4261         *ret = new_block;
4262         return (0);
4263 }
4264
4265 static errcode_t e2fsck_new_range(ext2_filsys fs, int flags, blk64_t goal,
4266                                   blk64_t len, blk64_t *pblk, blk64_t *plen)
4267 {
4268         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4269         errcode_t       retval;
4270
4271         if (ctx->block_found_map)
4272                 return ext2fs_new_range(fs, flags, goal, len,
4273                                         ctx->block_found_map, pblk, plen);
4274
4275         if (!fs->block_map) {
4276                 retval = ext2fs_read_block_bitmap(fs);
4277                 if (retval)
4278                         return retval;
4279         }
4280
4281         return ext2fs_new_range(fs, flags, goal, len, fs->block_map,
4282                                 pblk, plen);
4283 }
4284
4285 static void e2fsck_block_alloc_stats(ext2_filsys fs, blk64_t blk, int inuse)
4286 {
4287         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4288
4289         /* Never free a critical metadata block */
4290         if (ctx->block_found_map &&
4291             ctx->block_metadata_map &&
4292             inuse < 0 &&
4293             ext2fs_test_block_bitmap2(ctx->block_metadata_map, blk))
4294                 return;
4295
4296         if (ctx->block_found_map) {
4297                 if (inuse > 0)
4298                         ext2fs_mark_block_bitmap2(ctx->block_found_map, blk);
4299                 else
4300                         ext2fs_unmark_block_bitmap2(ctx->block_found_map, blk);
4301         }
4302 }
4303
4304 static void e2fsck_block_alloc_stats_range(ext2_filsys fs, blk64_t blk,
4305                                            blk_t num, int inuse)
4306 {
4307         e2fsck_t ctx = (e2fsck_t) fs->priv_data;
4308
4309         /* Never free a critical metadata block */
4310         if (ctx->block_found_map &&
4311             ctx->block_metadata_map &&
4312             inuse < 0 &&
4313             ext2fs_test_block_bitmap_range2(ctx->block_metadata_map, blk, num))
4314                 return;
4315
4316         if (ctx->block_found_map) {
4317                 if (inuse > 0)
4318                         ext2fs_mark_block_bitmap_range2(ctx->block_found_map,
4319                                                         blk, num);
4320                 else
4321                         ext2fs_unmark_block_bitmap_range2(ctx->block_found_map,
4322                                                         blk, num);
4323         }
4324 }
4325
4326 void e2fsck_use_inode_shortcuts(e2fsck_t ctx, int use_shortcuts)
4327 {
4328         ext2_filsys fs = ctx->fs;
4329
4330         if (use_shortcuts) {
4331                 fs->get_blocks = pass1_get_blocks;
4332                 fs->check_directory = pass1_check_directory;
4333                 fs->read_inode = pass1_read_inode;
4334                 fs->write_inode = pass1_write_inode;
4335                 ctx->stashed_ino = 0;
4336         } else {
4337                 fs->get_blocks = 0;
4338                 fs->check_directory = 0;
4339                 fs->read_inode = 0;
4340                 fs->write_inode = 0;
4341         }
4342 }
4343
4344 void e2fsck_intercept_block_allocations(e2fsck_t ctx)
4345 {
4346         ext2fs_set_alloc_block_callback(ctx->fs, e2fsck_get_alloc_block, 0);
4347         ext2fs_set_block_alloc_stats_callback(ctx->fs,
4348                                                 e2fsck_block_alloc_stats, 0);
4349         ext2fs_set_new_range_callback(ctx->fs, e2fsck_new_range, NULL);
4350         ext2fs_set_block_alloc_stats_range_callback(ctx->fs,
4351                                         e2fsck_block_alloc_stats_range, NULL);
4352 }