e2fsck/rehash.c

   1 /*
   2  * rehash.c --- rebuild hash tree directories
   3  *
   4  * Copyright (C) 2002 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  * This algorithm is designed for simplicity of implementation and to
  12  * pack the directory as much as possible.  It however requires twice
  13  * as much memory as the size of the directory.  The maximum size
  14  * directory supported using a 4k blocksize is roughly a gigabyte, and
  15  * so there may very well be problems with machines that don't have
  16  * virtual memory, and obscenely large directories.
  17  *
  18  * An alternate algorithm which is much more disk intensive could be
  19  * written, and probably will need to be written in the future.  The
  20  * design goals of such an algorithm are: (a) use (roughly) constant
  21  * amounts of memory, no matter how large the directory, (b) the
  22  * directory must be safe at all times, even if e2fsck is interrupted
  23  * in the middle, (c) we must use minimal amounts of extra disk
  24  * blocks.  This pretty much requires an incremental approach, where
  25  * we are reading from one part of the directory, and inserting into
  26  * the front half.  So the algorithm will have to keep track of a
  27  * moving block boundary between the new tree and the old tree, and
  28  * files will need to be moved from the old directory and inserted
  29  * into the new tree.  If the new directory requires space which isn't
  30  * yet available, blocks from the beginning part of the old directory
  31  * may need to be moved to the end of the directory to make room for
  32  * the new tree:
  33  *
  34  *    --------------------------------------------------------
  35  *    |  new tree   |        | old tree                      |
  36  *    --------------------------------------------------------
  37  *                  ^ ptr    ^ptr
  38  *                tail new   head old
  39  *
  40  * This is going to be a pain in the tuckus to implement, and will
  41  * require a lot more disk accesses.  So I'm going to skip it for now;
  42  * it's only really going to be an issue for really, really big
  43  * filesystems (when we reach the level of tens of millions of files
  44  * in a single directory).  It will probably be easier to simply
  45  * require that e2fsck use VM first.
  46  */
  47
  48 #include "config.h"
  49 #include <string.h>
  50 #include <ctype.h>
  51 #include <errno.h>
  52 #include "e2fsck.h"
  53 #include "problem.h"
  54
  55 /* Schedule a dir to be rebuilt during pass 3A. */
  56 void e2fsck_rehash_dir_later(e2fsck_t ctx, ext2_ino_t ino)
  57 {
  58         if (!ctx->dirs_to_hash)
  59                 ext2fs_u32_list_create(&ctx->dirs_to_hash, 50);
  60         if (ctx->dirs_to_hash)
  61                 ext2fs_u32_list_add(ctx->dirs_to_hash, ino);
  62 }
  63
  64 /* Ask if a dir will be rebuilt during pass 3A. */
  65 int e2fsck_dir_will_be_rehashed(e2fsck_t ctx, ext2_ino_t ino)
  66 {
  67         if (ctx->options & E2F_OPT_COMPRESS_DIRS)
  68                 return 1;
  69         if (!ctx->dirs_to_hash)
  70                 return 0;
  71         return ext2fs_u32_list_test(ctx->dirs_to_hash, ino);
  72 }
  73
  74 #undef REHASH_DEBUG
  75
  76 struct fill_dir_struct {
  77         char *buf;
  78         struct ext2_inode *inode;
  79         ext2_ino_t ino;
  80         errcode_t err;
  81         e2fsck_t ctx;
  82         struct hash_entry *harray;
  83         int max_array, num_array;
  84         unsigned int dir_size;
  85         int compress;
  86         ino_t parent;
  87         ext2_ino_t dir;
  88 };
  89
  90 struct hash_entry {
  91         ext2_dirhash_t  hash;
  92         ext2_dirhash_t  minor_hash;
  93         ino_t           ino;
  94         struct ext2_dir_entry   *dir;
  95 };
  96
  97 struct out_dir {
  98         int             num;
  99         int             max;
 100         char            *buf;
 101         ext2_dirhash_t  *hashes;
 102 };
 103
 104 static int fill_dir_block(ext2_filsys fs,
 105                           blk64_t *block_nr,
 106                           e2_blkcnt_t blockcnt,
 107                           blk64_t ref_block EXT2FS_ATTR((unused)),
 108                           int ref_offset EXT2FS_ATTR((unused)),
 109                           void *priv_data)
 110 {
 111         struct fill_dir_struct  *fd = (struct fill_dir_struct *) priv_data;
 112         struct hash_entry       *new_array, *ent;
 113         struct ext2_dir_entry   *dirent;
 114         char                    *dir;
 115         unsigned int            offset, dir_offset, rec_len, name_len;
 116         int                     hash_alg, hash_flags;
 117
 118         if (blockcnt < 0)
 119                 return 0;
 120
 121         offset = blockcnt * fs->blocksize;
 122         if (offset + fs->blocksize > fd->inode->i_size) {
 123                 fd->err = EXT2_ET_DIR_CORRUPTED;
 124                 return BLOCK_ABORT;
 125         }
 126
 127         dir = (fd->buf+offset);
 128         if (*block_nr == 0) {
 129                 memset(dir, 0, fs->blocksize);
 130                 dirent = (struct ext2_dir_entry *) dir;
 131                 (void) ext2fs_set_rec_len(fs, fs->blocksize, dirent);
 132         } else {
 133                 int flags = fs->flags;
 134                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
 135                 fd->err = ext2fs_read_dir_block4(fs, *block_nr, dir, 0,
 136                                                  fd->dir);
 137                 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
 138                             (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
 139                 if (fd->err)
 140                         return BLOCK_ABORT;
 141         }
 142         hash_flags = fd->inode->i_flags & EXT4_CASEFOLD_FL;
 143         hash_alg = fs->super->s_def_hash_version;
 144         if ((hash_alg <= EXT2_HASH_TEA) &&
 145             (fs->super->s_flags & EXT2_FLAGS_UNSIGNED_HASH))
 146                 hash_alg += 3;
 147         /* While the directory block is "hot", index it. */
 148         dir_offset = 0;
 149         while (dir_offset < fs->blocksize) {
 150                 dirent = (struct ext2_dir_entry *) (dir + dir_offset);
 151                 (void) ext2fs_get_rec_len(fs, dirent, &rec_len);
 152                 name_len = ext2fs_dirent_name_len(dirent);
 153                 if (((dir_offset + rec_len) > fs->blocksize) ||
 154                     (rec_len < 8) ||
 155                     ((rec_len % 4) != 0) ||
 156                     (name_len + 8 > rec_len)) {
 157                         fd->err = EXT2_ET_DIR_CORRUPTED;
 158                         return BLOCK_ABORT;
 159                 }
 160                 dir_offset += rec_len;
 161                 if (dirent->inode == 0)
 162                         continue;
 163                 if (!fd->compress && (name_len == 1) &&
 164                     (dirent->name[0] == '.'))
 165                         continue;
 166                 if (!fd->compress && (name_len == 2) &&
 167                     (dirent->name[0] == '.') && (dirent->name[1] == '.')) {
 168                         fd->parent = dirent->inode;
 169                         continue;
 170                 }
 171                 if (fd->num_array >= fd->max_array) {
 172                         new_array = realloc(fd->harray,
 173                             sizeof(struct hash_entry) * (fd->max_array+500));
 174                         if (!new_array) {
 175                                 fd->err = ENOMEM;
 176                                 return BLOCK_ABORT;
 177                         }
 178                         fd->harray = new_array;
 179                         fd->max_array += 500;
 180                 }
 181                 ent = fd->harray + fd->num_array++;
 182                 ent->dir = dirent;
 183                 fd->dir_size += EXT2_DIR_REC_LEN(name_len);
 184                 ent->ino = dirent->inode;
 185                 if (fd->compress)
 186                         ent->hash = ent->minor_hash = 0;
 187                 else {
 188                         fd->err = ext2fs_dirhash2(hash_alg,
 189                                                   dirent->name, name_len,
 190                                                   fs->encoding, hash_flags,
 191                                                   fs->super->s_hash_seed,
 192                                                   &ent->hash, &ent->minor_hash);
 193                         if (fd->err)
 194                                 return BLOCK_ABORT;
 195                 }
 196         }
 197
 198         return 0;
 199 }
 200
 201 /* Used for sorting the hash entry */
 202 static EXT2_QSORT_TYPE ino_cmp(const void *a, const void *b)
 203 {
 204         const struct hash_entry *he_a = (const struct hash_entry *) a;
 205         const struct hash_entry *he_b = (const struct hash_entry *) b;
 206
 207         return (he_a->ino - he_b->ino);
 208 }
 209
 210 /* Used for sorting the hash entry */
 211 static EXT2_QSORT_TYPE name_cmp(const void *a, const void *b)
 212 {
 213         const struct hash_entry *he_a = (const struct hash_entry *) a;
 214         const struct hash_entry *he_b = (const struct hash_entry *) b;
 215         unsigned int he_a_len, he_b_len, min_len;
 216         int     ret;
 217
 218         he_a_len = ext2fs_dirent_name_len(he_a->dir);
 219         he_b_len = ext2fs_dirent_name_len(he_b->dir);
 220         min_len = he_a_len;
 221         if (min_len > he_b_len)
 222                 min_len = he_b_len;
 223
 224         ret = memcmp(he_a->dir->name, he_b->dir->name, min_len);
 225         if (ret == 0) {
 226                 if (he_a_len > he_b_len)
 227                         ret = 1;
 228                 else if (he_a_len < he_b_len)
 229                         ret = -1;
 230                 else
 231                         ret = he_b->dir->inode - he_a->dir->inode;
 232         }
 233         return ret;
 234 }
 235
 236 /* Used for sorting the hash entry */
 237 static EXT2_QSORT_TYPE hash_cmp(const void *a, const void *b)
 238 {
 239         const struct hash_entry *he_a = (const struct hash_entry *) a;
 240         const struct hash_entry *he_b = (const struct hash_entry *) b;
 241         int     ret;
 242
 243         if (he_a->hash > he_b->hash)
 244                 ret = 1;
 245         else if (he_a->hash < he_b->hash)
 246                 ret = -1;
 247         else {
 248                 if (he_a->minor_hash > he_b->minor_hash)
 249                         ret = 1;
 250                 else if (he_a->minor_hash < he_b->minor_hash)
 251                         ret = -1;
 252                 else
 253                         ret = name_cmp(a, b);
 254         }
 255         return ret;
 256 }
 257
 258 static errcode_t alloc_size_dir(ext2_filsys fs, struct out_dir *outdir,
 259                                 int blocks)
 260 {
 261         void                    *new_mem;
 262
 263         if (outdir->max) {
 264                 new_mem = realloc(outdir->buf, blocks * fs->blocksize);
 265                 if (!new_mem)
 266                         return ENOMEM;
 267                 outdir->buf = new_mem;
 268                 new_mem = realloc(outdir->hashes,
 269                                   blocks * sizeof(ext2_dirhash_t));
 270                 if (!new_mem)
 271                         return ENOMEM;
 272                 outdir->hashes = new_mem;
 273         } else {
 274                 outdir->buf = malloc(blocks * fs->blocksize);
 275                 outdir->hashes = malloc(blocks * sizeof(ext2_dirhash_t));
 276                 outdir->num = 0;
 277         }
 278         outdir->max = blocks;
 279         return 0;
 280 }
 281
 282 static void free_out_dir(struct out_dir *outdir)
 283 {
 284         free(outdir->buf);
 285         free(outdir->hashes);
 286         outdir->max = 0;
 287         outdir->num =0;
 288 }
 289
 290 static errcode_t get_next_block(ext2_filsys fs, struct out_dir *outdir,
 291                          char ** ret)
 292 {
 293         errcode_t       retval;
 294
 295         if (outdir->num >= outdir->max) {
 296                 retval = alloc_size_dir(fs, outdir, outdir->max + 50);
 297                 if (retval)
 298                         return retval;
 299         }
 300         *ret = outdir->buf + (outdir->num++ * fs->blocksize);
 301         memset(*ret, 0, fs->blocksize);
 302         return 0;
 303 }
 304
 305 /*
 306  * This function is used to make a unique filename.  We do this by
 307  * appending ~0, and then incrementing the number.  However, we cannot
 308  * expand the length of the filename beyond the padding available in
 309  * the directory entry.
 310  */
 311 static void mutate_name(char *str, unsigned int *len)
 312 {
 313         int i;
 314         unsigned int l = *len;
 315
 316         /*
 317          * First check to see if it looks the name has been mutated
 318          * already
 319          */
 320         for (i = l-1; i > 0; i--) {
 321                 if (!isdigit(str[i]))
 322                         break;
 323         }
 324         if ((i == (int)l - 1) || (str[i] != '~')) {
 325                 if (((l-1) & 3) < 2)
 326                         l += 2;
 327                 else
 328                         l = (l+3) & ~3;
 329                 str[l-2] = '~';
 330                 str[l-1] = '0';
 331                 *len = l;
 332                 return;
 333         }
 334         for (i = l-1; i >= 0; i--) {
 335                 if (isdigit(str[i])) {
 336                         if (str[i] == '9')
 337                                 str[i] = '0';
 338                         else {
 339                                 str[i]++;
 340                                 return;
 341                         }
 342                         continue;
 343                 }
 344                 if (i == 1) {
 345                         if (str[0] == 'z')
 346                                 str[0] = 'A';
 347                         else if (str[0] == 'Z') {
 348                                 str[0] = '~';
 349                                 str[1] = '0';
 350                         } else
 351                                 str[0]++;
 352                 } else if (i > 0) {
 353                         str[i] = '1';
 354                         str[i-1] = '~';
 355                 } else {
 356                         if (str[0] == '~')
 357                                 str[0] = 'a';
 358                         else
 359                                 str[0]++;
 360                 }
 361                 break;
 362         }
 363 }
 364
 365 static int duplicate_search_and_fix(e2fsck_t ctx, ext2_filsys fs,
 366                                     ext2_ino_t ino,
 367                                     struct fill_dir_struct *fd)
 368 {
 369         struct problem_context  pctx;
 370         struct hash_entry       *ent, *prev;
 371         int                     i, j;
 372         int                     fixed = 0;
 373         char                    new_name[256];
 374         unsigned int            new_len;
 375         int                     hash_alg;
 376         int hash_flags = fd->inode->i_flags & EXT4_CASEFOLD_FL;
 377
 378         clear_problem_context(&pctx);
 379         pctx.ino = ino;
 380
 381         hash_alg = fs->super->s_def_hash_version;
 382         if ((hash_alg <= EXT2_HASH_TEA) &&
 383             (fs->super->s_flags & EXT2_FLAGS_UNSIGNED_HASH))
 384                 hash_alg += 3;
 385
 386         for (i=1; i < fd->num_array; i++) {
 387                 ent = fd->harray + i;
 388                 prev = ent - 1;
 389                 if (!ent->dir->inode ||
 390                     (ext2fs_dirent_name_len(ent->dir) !=
 391                      ext2fs_dirent_name_len(prev->dir)) ||
 392                     memcmp(ent->dir->name, prev->dir->name,
 393                              ext2fs_dirent_name_len(ent->dir)))
 394                         continue;
 395                 pctx.dirent = ent->dir;
 396                 if ((ent->dir->inode == prev->dir->inode) &&
 397                     fix_problem(ctx, PR_2_DUPLICATE_DIRENT, &pctx)) {
 398                         e2fsck_adjust_inode_count(ctx, ent->dir->inode, -1);
 399                         ent->dir->inode = 0;
 400                         fixed++;
 401                         continue;
 402                 }
 403                 new_len = ext2fs_dirent_name_len(ent->dir);
 404                 memcpy(new_name, ent->dir->name, new_len);
 405                 mutate_name(new_name, &new_len);
 406                 for (j=0; j < fd->num_array; j++) {
 407                         if ((i==j) ||
 408                             (new_len !=
 409                              (unsigned) ext2fs_dirent_name_len(fd->harray[j].dir)) ||
 410                             memcmp(new_name, fd->harray[j].dir->name, new_len))
 411                                 continue;
 412                         mutate_name(new_name, &new_len);
 413
 414                         j = -1;
 415                 }
 416                 new_name[new_len] = 0;
 417                 pctx.str = new_name;
 418                 if (fix_problem(ctx, PR_2_NON_UNIQUE_FILE, &pctx)) {
 419                         memcpy(ent->dir->name, new_name, new_len);
 420                         ext2fs_dirent_set_name_len(ent->dir, new_len);
 421                         ext2fs_dirhash2(hash_alg, new_name, new_len,
 422                                         fs->encoding, hash_flags,
 423                                         fs->super->s_hash_seed,
 424                                         &ent->hash, &ent->minor_hash);
 425                         fixed++;
 426                 }
 427         }
 428         return fixed;
 429 }
 430
 431
 432 static errcode_t copy_dir_entries(e2fsck_t ctx,
 433                                   struct fill_dir_struct *fd,
 434                                   struct out_dir *outdir)
 435 {
 436         ext2_filsys             fs = ctx->fs;
 437         errcode_t               retval;
 438         char                    *block_start;
 439         struct hash_entry       *ent;
 440         struct ext2_dir_entry   *dirent;
 441         unsigned int            rec_len, prev_rec_len, left, slack, offset;
 442         int                     i;
 443         ext2_dirhash_t          prev_hash;
 444         int                     csum_size = 0;
 445         struct                  ext2_dir_entry_tail *t;
 446
 447         if (ctx->htree_slack_percentage == 255) {
 448                 profile_get_uint(ctx->profile, "options",
 449                                  "indexed_dir_slack_percentage",
 450                                  0, 20,
 451                                  &ctx->htree_slack_percentage);
 452                 if (ctx->htree_slack_percentage > 100)
 453                         ctx->htree_slack_percentage = 20;
 454         }
 455
 456         if (ext2fs_has_feature_metadata_csum(fs->super))
 457                 csum_size = sizeof(struct ext2_dir_entry_tail);
 458
 459         outdir->max = 0;
 460         retval = alloc_size_dir(fs, outdir,
 461                                 (fd->dir_size / fs->blocksize) + 2);
 462         if (retval)
 463                 return retval;
 464         outdir->num = fd->compress ? 0 : 1;
 465         offset = 0;
 466         outdir->hashes[0] = 0;
 467         prev_hash = 1;
 468         if ((retval = get_next_block(fs, outdir, &block_start)))
 469                 return retval;
 470         dirent = (struct ext2_dir_entry *) block_start;
 471         prev_rec_len = 0;
 472         rec_len = 0;
 473         left = fs->blocksize - csum_size;
 474         slack = fd->compress ? 12 :
 475                 ((fs->blocksize - csum_size) * ctx->htree_slack_percentage)/100;
 476         if (slack < 12)
 477                 slack = 12;
 478         for (i = 0; i < fd->num_array; i++) {
 479                 ent = fd->harray + i;
 480                 if (ent->dir->inode == 0)
 481                         continue;
 482                 rec_len = EXT2_DIR_REC_LEN(ext2fs_dirent_name_len(ent->dir));
 483                 if (rec_len > left) {
 484                         if (left) {
 485                                 left += prev_rec_len;
 486                                 retval = ext2fs_set_rec_len(fs, left, dirent);
 487                                 if (retval)
 488                                         return retval;
 489                         }
 490                         if (csum_size) {
 491                                 t = EXT2_DIRENT_TAIL(block_start,
 492                                                      fs->blocksize);
 493                                 ext2fs_initialize_dirent_tail(fs, t);
 494                         }
 495                         if ((retval = get_next_block(fs, outdir,
 496                                                       &block_start)))
 497                                 return retval;
 498                         offset = 0;
 499                 }
 500                 left = (fs->blocksize - csum_size) - offset;
 501                 dirent = (struct ext2_dir_entry *) (block_start + offset);
 502                 if (offset == 0) {
 503                         if (ent->hash == prev_hash)
 504                                 outdir->hashes[outdir->num-1] = ent->hash | 1;
 505                         else
 506                                 outdir->hashes[outdir->num-1] = ent->hash;
 507                 }
 508                 dirent->inode = ent->dir->inode;
 509                 ext2fs_dirent_set_name_len(dirent,
 510                                            ext2fs_dirent_name_len(ent->dir));
 511                 ext2fs_dirent_set_file_type(dirent,
 512                                             ext2fs_dirent_file_type(ent->dir));
 513                 retval = ext2fs_set_rec_len(fs, rec_len, dirent);
 514                 if (retval)
 515                         return retval;
 516                 prev_rec_len = rec_len;
 517                 memcpy(dirent->name, ent->dir->name,
 518                        ext2fs_dirent_name_len(dirent));
 519                 offset += rec_len;
 520                 left -= rec_len;
 521                 if (left < slack) {
 522                         prev_rec_len += left;
 523                         retval = ext2fs_set_rec_len(fs, prev_rec_len, dirent);
 524                         if (retval)
 525                                 return retval;
 526                         offset += left;
 527                         left = 0;
 528                 }
 529                 prev_hash = ent->hash;
 530         }
 531         if (left)
 532                 retval = ext2fs_set_rec_len(fs, rec_len + left, dirent);
 533         if (csum_size) {
 534                 t = EXT2_DIRENT_TAIL(block_start, fs->blocksize);
 535                 ext2fs_initialize_dirent_tail(fs, t);
 536         }
 537
 538         return retval;
 539 }
 540
 541
 542 static struct ext2_dx_root_info *set_root_node(ext2_filsys fs, char *buf,
 543                                     ext2_ino_t ino, ext2_ino_t parent)
 544 {
 545         struct ext2_dir_entry           *dir;
 546         struct ext2_dx_root_info        *root;
 547         struct ext2_dx_countlimit       *limits;
 548         int                             filetype = 0;
 549         int                             csum_size = 0;
 550
 551         if (ext2fs_has_feature_filetype(fs->super))
 552                 filetype = EXT2_FT_DIR;
 553
 554         memset(buf, 0, fs->blocksize);
 555         dir = (struct ext2_dir_entry *) buf;
 556         dir->inode = ino;
 557         dir->name[0] = '.';
 558         ext2fs_dirent_set_name_len(dir, 1);
 559         ext2fs_dirent_set_file_type(dir, filetype);
 560         dir->rec_len = 12;
 561         dir = (struct ext2_dir_entry *) (buf + 12);
 562         dir->inode = parent;
 563         dir->name[0] = '.';
 564         dir->name[1] = '.';
 565         ext2fs_dirent_set_name_len(dir, 2);
 566         ext2fs_dirent_set_file_type(dir, filetype);
 567         dir->rec_len = fs->blocksize - 12;
 568
 569         root = (struct ext2_dx_root_info *) (buf+24);
 570         root->reserved_zero = 0;
 571         root->hash_version = fs->super->s_def_hash_version;
 572         root->info_length = 8;
 573         root->indirect_levels = 0;
 574         root->unused_flags = 0;
 575
 576         if (ext2fs_has_feature_metadata_csum(fs->super))
 577                 csum_size = sizeof(struct ext2_dx_tail);
 578
 579         limits = (struct ext2_dx_countlimit *) (buf+32);
 580         limits->limit = (fs->blocksize - (32 + csum_size)) /
 581                         sizeof(struct ext2_dx_entry);
 582         limits->count = 0;
 583
 584         return root;
 585 }
 586
 587
 588 static struct ext2_dx_entry *set_int_node(ext2_filsys fs, char *buf)
 589 {
 590         struct ext2_dir_entry           *dir;
 591         struct ext2_dx_countlimit       *limits;
 592         int                             csum_size = 0;
 593
 594         memset(buf, 0, fs->blocksize);
 595         dir = (struct ext2_dir_entry *) buf;
 596         dir->inode = 0;
 597         (void) ext2fs_set_rec_len(fs, fs->blocksize, dir);
 598
 599         if (ext2fs_has_feature_metadata_csum(fs->super))
 600                 csum_size = sizeof(struct ext2_dx_tail);
 601
 602         limits = (struct ext2_dx_countlimit *) (buf+8);
 603         limits->limit = (fs->blocksize - (8 + csum_size)) /
 604                         sizeof(struct ext2_dx_entry);
 605         limits->count = 0;
 606
 607         return (struct ext2_dx_entry *) limits;
 608 }
 609
 610 static int alloc_blocks(ext2_filsys fs,
 611                         struct ext2_dx_countlimit **limit,
 612                         struct ext2_dx_entry **prev_ent,
 613                         struct ext2_dx_entry **next_ent,
 614                         int *prev_offset, int *next_offset,
 615                         struct out_dir *outdir, int i,
 616                         int *prev_count, int *next_count)
 617 {
 618         errcode_t       retval;
 619         char            *block_start;
 620
 621         if (*limit)
 622                 (*limit)->limit = (*limit)->count =
 623                         ext2fs_cpu_to_le16((*limit)->limit);
 624         *prev_ent = (struct ext2_dx_entry *) (outdir->buf + *prev_offset);
 625         (*prev_ent)->block = ext2fs_cpu_to_le32(outdir->num);
 626
 627         if (i != 1)
 628                 (*prev_ent)->hash =
 629                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 630
 631         retval = get_next_block(fs, outdir, &block_start);
 632         if (retval)
 633                 return retval;
 634
 635         *next_ent = set_int_node(fs, block_start);
 636         *limit = (struct ext2_dx_countlimit *)(*next_ent);
 637         if (next_offset)
 638                 *next_offset = ((char *) *next_ent - outdir->buf);
 639
 640         *next_count = (*limit)->limit;
 641         (*prev_offset) += sizeof(struct ext2_dx_entry);
 642         (*prev_count)--;
 643
 644         return 0;
 645 }
 646
 647 /*
 648  * This function takes the leaf nodes which have been written in
 649  * outdir, and populates the root node and any necessary interior nodes.
 650  */
 651 static errcode_t calculate_tree(ext2_filsys fs,
 652                                 struct out_dir *outdir,
 653                                 ext2_ino_t ino,
 654                                 ext2_ino_t parent)
 655 {
 656         struct ext2_dx_root_info        *root_info;
 657         struct ext2_dx_entry            *root, *int_ent, *dx_ent = 0;
 658         struct ext2_dx_countlimit       *root_limit, *int_limit, *limit;
 659         errcode_t                       retval;
 660         int                             i, c1, c2, c3, nblks;
 661         int                             limit_offset, int_offset, root_offset;
 662
 663         root_info = set_root_node(fs, outdir->buf, ino, parent);
 664         root_offset = limit_offset = ((char *) root_info - outdir->buf) +
 665                 root_info->info_length;
 666         root_limit = (struct ext2_dx_countlimit *) (outdir->buf + limit_offset);
 667         c1 = root_limit->limit;
 668         nblks = outdir->num;
 669
 670         /* Write out the pointer blocks */
 671         if (nblks - 1 <= c1) {
 672                 /* Just write out the root block, and we're done */
 673                 root = (struct ext2_dx_entry *) (outdir->buf + root_offset);
 674                 for (i=1; i < nblks; i++) {
 675                         root->block = ext2fs_cpu_to_le32(i);
 676                         if (i != 1)
 677                                 root->hash =
 678                                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 679                         root++;
 680                         c1--;
 681                 }
 682         } else if (nblks - 1 <= ext2fs_htree_intnode_maxrecs(fs, c1)) {
 683                 c2 = 0;
 684                 limit = NULL;
 685                 root_info->indirect_levels = 1;
 686                 for (i=1; i < nblks; i++) {
 687                         if (c2 == 0 && c1 == 0)
 688                                 return ENOSPC;
 689                         if (c2 == 0) {
 690                                 retval = alloc_blocks(fs, &limit, &root,
 691                                                       &dx_ent, &root_offset,
 692                                                       NULL, outdir, i, &c1,
 693                                                       &c2);
 694                                 if (retval)
 695                                         return retval;
 696                         }
 697                         dx_ent->block = ext2fs_cpu_to_le32(i);
 698                         if (c2 != limit->limit)
 699                                 dx_ent->hash =
 700                                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 701                         dx_ent++;
 702                         c2--;
 703                 }
 704                 limit->count = ext2fs_cpu_to_le16(limit->limit - c2);
 705                 limit->limit = ext2fs_cpu_to_le16(limit->limit);
 706         } else {
 707                 c2 = 0;
 708                 c3 = 0;
 709                 limit = NULL;
 710                 int_limit = 0;
 711                 root_info->indirect_levels = 2;
 712                 for (i = 1; i < nblks; i++) {
 713                         if (c3 == 0 && c2 == 0 && c1 == 0)
 714                                 return ENOSPC;
 715                         if (c3 == 0 && c2 == 0) {
 716                                 retval = alloc_blocks(fs, &int_limit, &root,
 717                                                       &int_ent, &root_offset,
 718                                                       &int_offset, outdir, i,
 719                                                       &c1, &c2);
 720                                 if (retval)
 721                                         return retval;
 722                         }
 723                         if (c3 == 0) {
 724                                 retval = alloc_blocks(fs, &limit, &int_ent,
 725                                                       &dx_ent, &int_offset,
 726                                                       NULL, outdir, i, &c2,
 727                                                       &c3);
 728                                 if (retval)
 729                                         return retval;
 730
 731                         }
 732                         dx_ent->block = ext2fs_cpu_to_le32(i);
 733                         if (c3 != limit->limit)
 734                                 dx_ent->hash =
 735                                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 736                         dx_ent++;
 737                         c3--;
 738                 }
 739                 int_limit->count = ext2fs_cpu_to_le16(limit->limit - c2);
 740                 int_limit->limit = ext2fs_cpu_to_le16(limit->limit);
 741
 742                 limit->count = ext2fs_cpu_to_le16(limit->limit - c3);
 743                 limit->limit = ext2fs_cpu_to_le16(limit->limit);
 744
 745         }
 746         root_limit = (struct ext2_dx_countlimit *) (outdir->buf + limit_offset);
 747         root_limit->count = ext2fs_cpu_to_le16(root_limit->limit - c1);
 748         root_limit->limit = ext2fs_cpu_to_le16(root_limit->limit);
 749
 750         return 0;
 751 }
 752
 753 struct write_dir_struct {
 754         struct out_dir *outdir;
 755         errcode_t       err;
 756         ext2_ino_t      ino;
 757         e2fsck_t        ctx;
 758         ext2_ino_t      dir;
 759 };
 760
 761 /*
 762  * Helper function which writes out a directory block.
 763  */
 764 static int write_dir_block(ext2_filsys fs,
 765                            blk64_t *block_nr,
 766                            e2_blkcnt_t blockcnt,
 767                            blk64_t ref_block EXT2FS_ATTR((unused)),
 768                            int ref_offset EXT2FS_ATTR((unused)),
 769                            void *priv_data)
 770 {
 771         struct write_dir_struct *wd = (struct write_dir_struct *) priv_data;
 772         char    *dir, *buf = 0;
 773
 774 #ifdef REHASH_DEBUG
 775         printf("%u: write_dir_block %lld:%lld", wd->ino, blockcnt, *block_nr);
 776 #endif
 777         if ((*block_nr == 0) || (blockcnt < 0)) {
 778 #ifdef REHASH_DEBUG
 779                 printf(" - skip\n");
 780 #endif
 781                 return 0;
 782         }
 783         if (blockcnt < wd->outdir->num)
 784                 dir = wd->outdir->buf + (blockcnt * fs->blocksize);
 785         else if (wd->ctx->lost_and_found == wd->dir) {
 786                 /* Don't release any extra directory blocks for lost+found */
 787                 wd->err = ext2fs_new_dir_block(fs, 0, 0, &buf);
 788                 if (wd->err)
 789                         return BLOCK_ABORT;
 790                 dir = buf;
 791                 wd->outdir->num++;
 792         } else {
 793                 /* Don't free blocks at the end of the directory, they
 794                  * will be truncated by the caller. */
 795 #ifdef REHASH_DEBUG
 796                 printf(" - not freed\n");
 797 #endif
 798                 return 0;
 799         }
 800         wd->err = ext2fs_write_dir_block4(fs, *block_nr, dir, 0, wd->dir);
 801         if (buf)
 802                 ext2fs_free_mem(&buf);
 803
 804 #ifdef REHASH_DEBUG
 805         printf(" - write (%d)\n", wd->err);
 806 #endif
 807         if (wd->err)
 808                 return BLOCK_ABORT;
 809         return 0;
 810 }
 811
 812 static errcode_t write_directory(e2fsck_t ctx, ext2_filsys fs,
 813                                  struct out_dir *outdir,
 814                                  ext2_ino_t ino, struct ext2_inode *inode,
 815                                  int compress)
 816 {
 817         struct write_dir_struct wd;
 818         errcode_t       retval;
 819
 820         retval = e2fsck_expand_directory(ctx, ino, -1, outdir->num);
 821         if (retval)
 822                 return retval;
 823
 824         wd.outdir = outdir;
 825         wd.err = 0;
 826         wd.ino = ino;
 827         wd.ctx = ctx;
 828         wd.dir = ino;
 829
 830         retval = ext2fs_block_iterate3(fs, ino, 0, NULL,
 831                                        write_dir_block, &wd);
 832         if (retval)
 833                 return retval;
 834         if (wd.err)
 835                 return wd.err;
 836
 837         e2fsck_read_inode(ctx, ino, inode, "rehash_dir");
 838         if (compress)
 839                 inode->i_flags &= ~EXT2_INDEX_FL;
 840         else
 841                 inode->i_flags |= EXT2_INDEX_FL;
 842 #ifdef REHASH_DEBUG
 843         printf("%u: set inode size to %u blocks = %u bytes\n",
 844                ino, outdir->num, outdir->num * fs->blocksize);
 845 #endif
 846         retval = ext2fs_inode_size_set(fs, inode, (ext2_off64_t)outdir->num *
 847                                                    fs->blocksize);
 848         if (retval)
 849                 return retval;
 850
 851         /* ext2fs_punch() calls ext2fs_write_inode() which writes the size */
 852         return ext2fs_punch(fs, ino, inode, NULL, outdir->num, ~0ULL);
 853 }
 854
 855 errcode_t e2fsck_rehash_dir(e2fsck_t ctx, ext2_ino_t ino,
 856                             struct problem_context *pctx)
 857 {
 858         ext2_filsys             fs = ctx->fs;
 859         errcode_t               retval;
 860         struct ext2_inode       inode;
 861         char                    *dir_buf = 0;
 862         struct fill_dir_struct  fd = { NULL, NULL, 0, 0, 0, NULL,
 863                                        0, 0, 0, 0, 0, 0 };
 864         struct out_dir          outdir = { 0, 0, 0, 0 };
 865
 866         e2fsck_read_inode(ctx, ino, &inode, "rehash_dir");
 867
 868         if (ext2fs_has_feature_inline_data(fs->super) &&
 869            (inode.i_flags & EXT4_INLINE_DATA_FL))
 870                 return 0;
 871
 872         retval = ENOMEM;
 873         dir_buf = malloc(inode.i_size);
 874         if (!dir_buf)
 875                 goto errout;
 876
 877         fd.max_array = inode.i_size / 32;
 878         fd.harray = malloc(fd.max_array * sizeof(struct hash_entry));
 879         if (!fd.harray)
 880                 goto errout;
 881
 882         fd.ino = ino;
 883         fd.ctx = ctx;
 884         fd.buf = dir_buf;
 885         fd.inode = &inode;
 886         fd.dir = ino;
 887         if (!ext2fs_has_feature_dir_index(fs->super) ||
 888             (inode.i_size / fs->blocksize) < 2)
 889                 fd.compress = 1;
 890         fd.parent = 0;
 891
 892 retry_nohash:
 893         /* Read in the entire directory into memory */
 894         retval = ext2fs_block_iterate3(fs, ino, 0, 0,
 895                                        fill_dir_block, &fd);
 896         if (fd.err) {
 897                 retval = fd.err;
 898                 goto errout;
 899         }
 900
 901         /*
 902          * If the entries read are less than a block, then don't index
 903          * the directory
 904          */
 905         if (!fd.compress && (fd.dir_size < (fs->blocksize - 24))) {
 906                 fd.compress = 1;
 907                 fd.dir_size = 0;
 908                 fd.num_array = 0;
 909                 goto retry_nohash;
 910         }
 911
 912 #if 0
 913         printf("%d entries (%d bytes) found in inode %d\n",
 914                fd.num_array, fd.dir_size, ino);
 915 #endif
 916
 917         /* Sort the list */
 918 resort:
 919         if (fd.compress && fd.num_array > 1)
 920                 qsort(fd.harray+2, fd.num_array-2, sizeof(struct hash_entry),
 921                       hash_cmp);
 922         else
 923                 qsort(fd.harray, fd.num_array, sizeof(struct hash_entry),
 924                       hash_cmp);
 925
 926         /*
 927          * Look for duplicates
 928          */
 929         if (duplicate_search_and_fix(ctx, fs, ino, &fd))
 930                 goto resort;
 931
 932         if (ctx->options & E2F_OPT_NO) {
 933                 retval = 0;
 934                 goto errout;
 935         }
 936
 937         /* Sort non-hashed directories by inode number */
 938         if (fd.compress && fd.num_array > 1)
 939                 qsort(fd.harray+2, fd.num_array-2,
 940                       sizeof(struct hash_entry), ino_cmp);
 941
 942         /*
 943          * Copy the directory entries.  In a htree directory these
 944          * will become the leaf nodes.
 945          */
 946         retval = copy_dir_entries(ctx, &fd, &outdir);
 947         if (retval)
 948                 goto errout;
 949
 950         free(dir_buf); dir_buf = 0;
 951
 952         if (!fd.compress) {
 953                 /* Calculate the interior nodes */
 954                 retval = calculate_tree(fs, &outdir, ino, fd.parent);
 955                 if (retval)
 956                         goto errout;
 957         }
 958
 959         retval = write_directory(ctx, fs, &outdir, ino, &inode, fd.compress);
 960         if (retval)
 961                 goto errout;
 962
 963         if (ctx->options & E2F_OPT_CONVERT_BMAP)
 964                 retval = e2fsck_rebuild_extents_later(ctx, ino);
 965         else
 966                 retval = e2fsck_check_rebuild_extents(ctx, ino, &inode, pctx);
 967 errout:
 968         free(dir_buf);
 969         free(fd.harray);
 970
 971         free_out_dir(&outdir);
 972         return retval;
 973 }
 974
 975 void e2fsck_rehash_directories(e2fsck_t ctx)
 976 {
 977         struct problem_context  pctx;
 978 #ifdef RESOURCE_TRACK
 979         struct resource_track   rtrack;
 980 #endif
 981         struct dir_info         *dir;
 982         ext2_u32_iterate        iter;
 983         struct dir_info_iter *  dirinfo_iter = 0;
 984         ext2_ino_t              ino;
 985         errcode_t               retval;
 986         int                     cur, max, all_dirs, first = 1;
 987
 988         init_resource_track(&rtrack, ctx->fs->io);
 989         all_dirs = ctx->options & E2F_OPT_COMPRESS_DIRS;
 990
 991         if (!ctx->dirs_to_hash && !all_dirs)
 992                 return;
 993
 994         (void) e2fsck_get_lost_and_found(ctx, 0);
 995
 996         clear_problem_context(&pctx);
 997
 998         cur = 0;
 999         if (all_dirs) {
1000                 dirinfo_iter = e2fsck_dir_info_iter_begin(ctx);
1001                 max = e2fsck_get_num_dirinfo(ctx);
1002         } else {
1003                 retval = ext2fs_u32_list_iterate_begin(ctx->dirs_to_hash,
1004                                                        &iter);
1005                 if (retval) {
1006                         pctx.errcode = retval;
1007                         fix_problem(ctx, PR_3A_OPTIMIZE_ITER, &pctx);
1008                         return;
1009                 }
1010                 max = ext2fs_u32_list_count(ctx->dirs_to_hash);
1011         }
1012         while (1) {
1013                 if (all_dirs) {
1014                         if ((dir = e2fsck_dir_info_iter(ctx,
1015                                                         dirinfo_iter)) == 0)
1016                                 break;
1017                         ino = dir->ino;
1018                 } else {
1019                         if (!ext2fs_u32_list_iterate(iter, &ino))
1020                                 break;
1021                 }
1022
1023                 pctx.dir = ino;
1024                 if (first) {
1025                         fix_problem(ctx, PR_3A_PASS_HEADER, &pctx);
1026                         first = 0;
1027                 }
1028 #if 0
1029                 fix_problem(ctx, PR_3A_OPTIMIZE_DIR, &pctx);
1030 #endif
1031                 pctx.errcode = e2fsck_rehash_dir(ctx, ino, &pctx);
1032                 if (pctx.errcode) {
1033                         end_problem_latch(ctx, PR_LATCH_OPTIMIZE_DIR);
1034                         fix_problem(ctx, PR_3A_OPTIMIZE_DIR_ERR, &pctx);
1035                 }
1036                 if (ctx->progress && !ctx->progress_fd)
1037                         e2fsck_simple_progress(ctx, "Rebuilding directory",
1038                                100.0 * (float) (++cur) / (float) max, ino);
1039         }
1040         end_problem_latch(ctx, PR_LATCH_OPTIMIZE_DIR);
1041         if (all_dirs)
1042                 e2fsck_dir_info_iter_end(ctx, dirinfo_iter);
1043         else
1044                 ext2fs_u32_list_iterate_end(iter);
1045
1046         if (ctx->dirs_to_hash)
1047                 ext2fs_u32_list_free(ctx->dirs_to_hash);
1048         ctx->dirs_to_hash = 0;
1049
1050         print_resource_track(ctx, "Pass 3A", &rtrack, ctx->fs->io);
1051 }