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 #include "support/sort_r.h"
  55
  56 /* Schedule a dir to be rebuilt during pass 3A. */
  57 void e2fsck_rehash_dir_later(e2fsck_t ctx, ext2_ino_t ino)
  58 {
  59         if (!ctx->dirs_to_hash)
  60                 ext2fs_u32_list_create(&ctx->dirs_to_hash, 50);
  61         if (ctx->dirs_to_hash)
  62                 ext2fs_u32_list_add(ctx->dirs_to_hash, ino);
  63 }
  64
  65 /* Ask if a dir will be rebuilt during pass 3A. */
  66 int e2fsck_dir_will_be_rehashed(e2fsck_t ctx, ext2_ino_t ino)
  67 {
  68         if (ctx->options & E2F_OPT_COMPRESS_DIRS)
  69                 return 1;
  70         if (!ctx->dirs_to_hash)
  71                 return 0;
  72         return ext2fs_u32_list_test(ctx->dirs_to_hash, ino);
  73 }
  74
  75 #undef REHASH_DEBUG
  76
  77 struct fill_dir_struct {
  78         char *buf;
  79         struct ext2_inode *inode;
  80         ext2_ino_t ino;
  81         errcode_t err;
  82         e2fsck_t ctx;
  83         struct hash_entry *harray;
  84         blk_t max_array, num_array;
  85         ext2_off64_t dir_size;
  86         int compress;
  87         ext2_ino_t parent;
  88         ext2_ino_t dir;
  89 };
  90
  91 struct hash_entry {
  92         ext2_dirhash_t  hash;
  93         ext2_dirhash_t  minor_hash;
  94         ino_t           ino;
  95         struct ext2_dir_entry   *dir;
  96 };
  97
  98 struct out_dir {
  99         blk_t           num;
 100         blk_t           max;
 101         char            *buf;
 102         ext2_dirhash_t  *hashes;
 103 };
 104
 105 #define DOTDOT_OFFSET 12
 106
 107 static int is_fake_entry(ext2_filsys fs, int lblk, unsigned int offset)
 108 {
 109         /* Entries in the first block before this value refer to . or .. */
 110         if (lblk == 0 && offset <= DOTDOT_OFFSET)
 111                 return 1;
 112         /* Check if this is likely the csum entry */
 113         if (ext2fs_has_feature_metadata_csum(fs->super) &&
 114             (offset & (fs->blocksize - 1)) ==
 115                             fs->blocksize - sizeof(struct ext2_dir_entry_tail))
 116                 return 1;
 117         return 0;
 118 }
 119
 120 static int fill_dir_block(ext2_filsys fs,
 121                           blk64_t *block_nr,
 122                           e2_blkcnt_t blockcnt,
 123                           blk64_t ref_block EXT2FS_ATTR((unused)),
 124                           int ref_offset EXT2FS_ATTR((unused)),
 125                           void *priv_data)
 126 {
 127         struct fill_dir_struct  *fd = (struct fill_dir_struct *) priv_data;
 128         struct hash_entry       *ent;
 129         struct ext2_dir_entry   *dirent;
 130         char                    *dir;
 131         unsigned int            offset, dir_offset, rec_len, name_len;
 132         int                     hash_alg, hash_flags, hash_in_entry;
 133
 134         if (blockcnt < 0)
 135                 return 0;
 136
 137         offset = blockcnt * fs->blocksize;
 138         if (offset + fs->blocksize > fd->inode->i_size) {
 139                 fd->err = EXT2_ET_DIR_CORRUPTED;
 140                 return BLOCK_ABORT;
 141         }
 142
 143         dir = (fd->buf+offset);
 144         if (*block_nr == 0) {
 145                 memset(dir, 0, fs->blocksize);
 146                 dirent = (struct ext2_dir_entry *) dir;
 147                 (void) ext2fs_set_rec_len(fs, fs->blocksize, dirent);
 148         } else {
 149                 int flags = fs->flags;
 150                 fs->flags |= EXT2_FLAG_IGNORE_CSUM_ERRORS;
 151                 fd->err = ext2fs_read_dir_block4(fs, *block_nr, dir, 0,
 152                                                  fd->dir);
 153                 fs->flags = (flags & EXT2_FLAG_IGNORE_CSUM_ERRORS) |
 154                             (fs->flags & ~EXT2_FLAG_IGNORE_CSUM_ERRORS);
 155                 if (fd->err)
 156                         return BLOCK_ABORT;
 157         }
 158         hash_flags = fd->inode->i_flags & EXT4_CASEFOLD_FL;
 159         hash_in_entry = ext4_hash_in_dirent(fd->inode);
 160         hash_alg = fs->super->s_def_hash_version;
 161         if ((hash_alg <= EXT2_HASH_TEA) &&
 162             (fs->super->s_flags & EXT2_FLAGS_UNSIGNED_HASH))
 163                 hash_alg += 3;
 164         /* While the directory block is "hot", index it. */
 165         dir_offset = 0;
 166         while (dir_offset < fs->blocksize) {
 167                 int min_rec = EXT2_DIR_ENTRY_HEADER_LEN;
 168                 int extended = hash_in_entry && !is_fake_entry(fs, blockcnt, dir_offset);
 169
 170                 if (extended)
 171                         min_rec += EXT2_DIR_ENTRY_HASH_LEN;
 172                 dirent = (struct ext2_dir_entry *) (dir + dir_offset);
 173                 (void) ext2fs_get_rec_len(fs, dirent, &rec_len);
 174                 name_len = ext2fs_dirent_name_len(dirent);
 175                 if (((dir_offset + rec_len) > fs->blocksize) ||
 176                     (rec_len < min_rec) ||
 177                     ((rec_len % 4) != 0) ||
 178                     (name_len + min_rec > rec_len)) {
 179                         fd->err = EXT2_ET_DIR_CORRUPTED;
 180                         return BLOCK_ABORT;
 181                 }
 182                 dir_offset += rec_len;
 183                 if (dirent->inode == 0)
 184                         continue;
 185                 if ((name_len) == 0) {
 186                         fd->err = EXT2_ET_DIR_CORRUPTED;
 187                         return BLOCK_ABORT;
 188                 }
 189                 if (!fd->compress && (name_len == 1) &&
 190                     (dirent->name[0] == '.'))
 191                         continue;
 192                 if (!fd->compress && (name_len == 2) &&
 193                     (dirent->name[0] == '.') && (dirent->name[1] == '.')) {
 194                         fd->parent = dirent->inode;
 195                         continue;
 196                 }
 197                 if (fd->num_array >= fd->max_array) {
 198                         errcode_t retval;
 199
 200                         retval = ext2fs_resize_array(sizeof(struct hash_entry),
 201                                                      fd->max_array,
 202                                                      fd->max_array + 500,
 203                                                      &fd->harray);
 204                         if (retval) {
 205                                 fd->err = retval;
 206                                 return BLOCK_ABORT;
 207                         }
 208                         fd->max_array += 500;
 209                 }
 210                 ent = fd->harray + fd->num_array++;
 211                 ent->dir = dirent;
 212                 fd->dir_size += ext2fs_dir_rec_len(name_len, extended);
 213                 ent->ino = dirent->inode;
 214                 if (extended) {
 215                         ent->hash = EXT2_DIRENT_HASH(dirent);
 216                         ent->minor_hash = EXT2_DIRENT_MINOR_HASH(dirent);
 217                 } else if (fd->compress) {
 218                         ent->hash = ent->minor_hash = 0;
 219                 } else {
 220                         fd->err = ext2fs_dirhash2(hash_alg,
 221                                                   dirent->name, name_len,
 222                                                   fs->encoding, hash_flags,
 223                                                   fs->super->s_hash_seed,
 224                                                   &ent->hash, &ent->minor_hash);
 225                         if (fd->err)
 226                                 return BLOCK_ABORT;
 227                 }
 228         }
 229
 230         return 0;
 231 }
 232
 233 /* Used for sorting the hash entry */
 234 static EXT2_QSORT_TYPE ino_cmp(const void *a, const void *b)
 235 {
 236         const struct hash_entry *he_a = (const struct hash_entry *) a;
 237         const struct hash_entry *he_b = (const struct hash_entry *) b;
 238
 239         return (he_a->ino - he_b->ino);
 240 }
 241
 242 struct name_cmp_ctx
 243 {
 244         int casefold;
 245         const struct ext2fs_nls_table *tbl;
 246 };
 247
 248 static int same_name(const struct name_cmp_ctx *cmp_ctx, char *s1,
 249                      int len1, char *s2, int len2)
 250 {
 251         if (!cmp_ctx->casefold)
 252                 return (len1 == len2 && !memcmp(s1, s2, len1));
 253         else
 254                 return !ext2fs_casefold_cmp(cmp_ctx->tbl,
 255                                             (unsigned char *) s1, len1,
 256                                             (unsigned char *) s2, len2);
 257 }
 258
 259 /* Used for sorting the hash entry */
 260 static EXT2_QSORT_TYPE name_cmp(const void *a, const void *b)
 261 {
 262         const struct hash_entry *he_a = (const struct hash_entry *) a;
 263         const struct hash_entry *he_b = (const struct hash_entry *) b;
 264         unsigned int he_a_len, he_b_len, min_len;
 265         int     ret;
 266
 267         he_a_len = ext2fs_dirent_name_len(he_a->dir);
 268         he_b_len = ext2fs_dirent_name_len(he_b->dir);
 269         min_len = he_a_len;
 270         if (min_len > he_b_len)
 271                 min_len = he_b_len;
 272
 273         ret = memcmp(he_a->dir->name, he_b->dir->name, min_len);
 274         if (ret == 0) {
 275                 if (he_a_len > he_b_len)
 276                         ret = 1;
 277                 else if (he_a_len < he_b_len)
 278                         ret = -1;
 279                 else
 280                         ret = he_b->dir->inode - he_a->dir->inode;
 281         }
 282         return ret;
 283 }
 284
 285 static EXT2_QSORT_TYPE name_cf_cmp(const struct name_cmp_ctx *ctx,
 286                                    const void *a, const void *b)
 287 {
 288         const struct hash_entry *he_a = (const struct hash_entry *) a;
 289         const struct hash_entry *he_b = (const struct hash_entry *) b;
 290         unsigned int he_a_len, he_b_len;
 291         int ret;
 292
 293         he_a_len = ext2fs_dirent_name_len(he_a->dir);
 294         he_b_len = ext2fs_dirent_name_len(he_b->dir);
 295
 296         ret = ext2fs_casefold_cmp(ctx->tbl,
 297                                   (unsigned char *) he_a->dir->name, he_a_len,
 298                                   (unsigned char *) he_b->dir->name, he_b_len);
 299         if (ret == 0) {
 300                 if (he_a_len > he_b_len)
 301                         ret = 1;
 302                 else if (he_a_len < he_b_len)
 303                         ret = -1;
 304                 else
 305                         ret = he_b->dir->inode - he_a->dir->inode;
 306         }
 307         return ret;
 308 }
 309
 310 /* Used for sorting the hash entry */
 311 static EXT2_QSORT_TYPE hash_cmp(const void *a, const void *b, void *arg)
 312 {
 313         const struct name_cmp_ctx *ctx = (struct name_cmp_ctx *) arg;
 314         const struct hash_entry *he_a = (const struct hash_entry *) a;
 315         const struct hash_entry *he_b = (const struct hash_entry *) b;
 316         int     ret;
 317
 318         if (he_a->hash > he_b->hash)
 319                 ret = 1;
 320         else if (he_a->hash < he_b->hash)
 321                 ret = -1;
 322         else {
 323                 if (he_a->minor_hash > he_b->minor_hash)
 324                         ret = 1;
 325                 else if (he_a->minor_hash < he_b->minor_hash)
 326                         ret = -1;
 327                 else {
 328                         if (ctx->casefold)
 329                                 ret = name_cf_cmp(ctx, a, b);
 330                         else
 331                                 ret = name_cmp(a, b);
 332                 }
 333         }
 334         return ret;
 335 }
 336
 337 static errcode_t alloc_size_dir(ext2_filsys fs, struct out_dir *outdir,
 338                                 blk_t blocks)
 339 {
 340         errcode_t retval;
 341
 342         if (outdir->max) {
 343                 retval = ext2fs_resize_array(fs->blocksize, outdir->max, blocks,
 344                                              &outdir->buf);
 345                 if (retval)
 346                         return retval;
 347                 retval = ext2fs_resize_array(sizeof(ext2_dirhash_t),
 348                                              outdir->max, blocks,
 349                                              &outdir->hashes);
 350                 if (retval)
 351                         return retval;
 352         } else {
 353                 retval = ext2fs_get_array(fs->blocksize, blocks, &outdir->buf);
 354                 if (retval)
 355                         return retval;
 356                 retval = ext2fs_get_array(sizeof(ext2_dirhash_t), blocks,
 357                                           &outdir->hashes);
 358                 if (retval)
 359                         return retval;
 360                 outdir->num = 0;
 361         }
 362         outdir->max = blocks;
 363         return 0;
 364 }
 365
 366 static void free_out_dir(struct out_dir *outdir)
 367 {
 368         free(outdir->buf);
 369         free(outdir->hashes);
 370         outdir->max = 0;
 371         outdir->num =0;
 372 }
 373
 374 static errcode_t get_next_block(ext2_filsys fs, struct out_dir *outdir,
 375                          char ** ret)
 376 {
 377         errcode_t       retval;
 378
 379         if (outdir->num >= outdir->max) {
 380                 int increment = outdir->max / 10;
 381
 382                 if (increment < 50)
 383                         increment = 50;
 384                 retval = alloc_size_dir(fs, outdir, outdir->max + increment);
 385                 if (retval)
 386                         return retval;
 387         }
 388         *ret = outdir->buf + (size_t)outdir->num++ * fs->blocksize;
 389         memset(*ret, 0, fs->blocksize);
 390         return 0;
 391 }
 392
 393 /*
 394  * This function is used to make a unique filename.  We do this by
 395  * appending ~0, and then incrementing the number.  However, we cannot
 396  * expand the length of the filename beyond the padding available in
 397  * the directory entry.
 398  */
 399 static void mutate_name(char *str, unsigned int *len)
 400 {
 401         int i;
 402         unsigned int l = *len;
 403
 404         /*
 405          * First check to see if it looks the name has been mutated
 406          * already
 407          */
 408         for (i = l-1; i > 0; i--) {
 409                 if (!isdigit(str[i]))
 410                         break;
 411         }
 412         if ((i == (int)l - 1) || (str[i] != '~')) {
 413                 if (((l-1) & 3) < 2)
 414                         l += 2;
 415                 else
 416                         l = (l+3) & ~3;
 417                 str[l-2] = '~';
 418                 str[l-1] = '0';
 419                 *len = l;
 420                 return;
 421         }
 422         for (i = l-1; i >= 0; i--) {
 423                 if (isdigit(str[i])) {
 424                         if (str[i] == '9')
 425                                 str[i] = '0';
 426                         else {
 427                                 str[i]++;
 428                                 return;
 429                         }
 430                         continue;
 431                 }
 432                 if (i == 1) {
 433                         if (str[0] == 'z')
 434                                 str[0] = 'A';
 435                         else if (str[0] == 'Z') {
 436                                 str[0] = '~';
 437                                 str[1] = '0';
 438                         } else
 439                                 str[0]++;
 440                 } else if (i > 0) {
 441                         str[i] = '1';
 442                         str[i-1] = '~';
 443                 } else {
 444                         if (str[0] == '~')
 445                                 str[0] = 'a';
 446                         else
 447                                 str[0]++;
 448                 }
 449                 break;
 450         }
 451 }
 452
 453 static int duplicate_search_and_fix(e2fsck_t ctx, ext2_filsys fs,
 454                                     ext2_ino_t ino,
 455                                     struct fill_dir_struct *fd,
 456                                     const struct name_cmp_ctx *cmp_ctx)
 457 {
 458         struct problem_context  pctx;
 459         struct hash_entry       *ent, *prev;
 460         blk_t                   i, j;
 461         int                     fixed = 0;
 462         char                    new_name[256];
 463         unsigned int            new_len;
 464         int                     hash_alg;
 465         int hash_flags = fd->inode->i_flags & EXT4_CASEFOLD_FL;
 466
 467         clear_problem_context(&pctx);
 468         pctx.ino = ino;
 469
 470         hash_alg = fs->super->s_def_hash_version;
 471         if ((hash_alg <= EXT2_HASH_TEA) &&
 472             (fs->super->s_flags & EXT2_FLAGS_UNSIGNED_HASH))
 473                 hash_alg += 3;
 474
 475         for (i=1; i < fd->num_array; i++) {
 476                 ent = fd->harray + i;
 477                 prev = ent - 1;
 478                 if (!ent->dir->inode ||
 479                     !same_name(cmp_ctx, ent->dir->name,
 480                                ext2fs_dirent_name_len(ent->dir),
 481                                prev->dir->name,
 482                                ext2fs_dirent_name_len(prev->dir)))
 483                         continue;
 484                 pctx.dirent = ent->dir;
 485                 if ((ent->dir->inode == prev->dir->inode) &&
 486                     fix_problem(ctx, PR_2_DUPLICATE_DIRENT, &pctx)) {
 487                         e2fsck_adjust_inode_count(ctx, ent->dir->inode, -1);
 488                         ent->dir->inode = 0;
 489                         fixed++;
 490                         continue;
 491                 }
 492                 /* Can't alter encrypted name without key, so just drop it */
 493                 if (fd->inode->i_flags & EXT4_ENCRYPT_FL) {
 494                         if (fix_problem(ctx, PR_2_NON_UNIQUE_FILE_NO_RENAME, &pctx)) {
 495                                 e2fsck_adjust_inode_count(ctx, ent->dir->inode, -1);
 496                                 ent->dir->inode = 0;
 497                                 fixed++;
 498                                 continue;
 499                         }
 500                 }
 501                 new_len = ext2fs_dirent_name_len(ent->dir);
 502                 if (new_len == 0) {
 503                          /* should never happen */
 504                         ext2fs_unmark_valid(fs);
 505                         continue;
 506                 }
 507                 memcpy(new_name, ent->dir->name, new_len);
 508                 mutate_name(new_name, &new_len);
 509                 for (j=0; j < fd->num_array; j++) {
 510                         if ((i==j) ||
 511                             !same_name(cmp_ctx, new_name, new_len,
 512                                        fd->harray[j].dir->name,
 513                                        ext2fs_dirent_name_len(fd->harray[j].dir))) {
 514                                 continue;
 515                         }
 516                         mutate_name(new_name, &new_len);
 517
 518                         j = -1;
 519                 }
 520                 new_name[new_len] = 0;
 521                 pctx.str = new_name;
 522                 if (fix_problem(ctx, PR_2_NON_UNIQUE_FILE, &pctx)) {
 523                         memcpy(ent->dir->name, new_name, new_len);
 524                         ext2fs_dirent_set_name_len(ent->dir, new_len);
 525                         ext2fs_dirhash2(hash_alg, new_name, new_len,
 526                                         fs->encoding, hash_flags,
 527                                         fs->super->s_hash_seed,
 528                                         &ent->hash, &ent->minor_hash);
 529                         fixed++;
 530                 }
 531         }
 532         return fixed;
 533 }
 534
 535
 536 static errcode_t copy_dir_entries(e2fsck_t ctx,
 537                                   struct fill_dir_struct *fd,
 538                                   struct out_dir *outdir)
 539 {
 540         ext2_filsys             fs = ctx->fs;
 541         errcode_t               retval;
 542         char                    *block_start;
 543         struct hash_entry       *ent;
 544         struct ext2_dir_entry   *dirent;
 545         unsigned int            rec_len, prev_rec_len, left, slack, offset;
 546         blk_t                   i;
 547         ext2_dirhash_t          prev_hash;
 548         int                     csum_size = 0;
 549         struct                  ext2_dir_entry_tail *t;
 550         int hash_in_entry = ext4_hash_in_dirent(fd->inode);
 551         int min_rec_len = ext2fs_dir_rec_len(1, hash_in_entry);
 552
 553         if (ctx->htree_slack_percentage == 255) {
 554                 profile_get_uint(ctx->profile, "options",
 555                                  "indexed_dir_slack_percentage",
 556                                  0, 20,
 557                                  &ctx->htree_slack_percentage);
 558                 if (ctx->htree_slack_percentage > 100)
 559                         ctx->htree_slack_percentage = 20;
 560         }
 561
 562         if (ext2fs_has_feature_metadata_csum(fs->super))
 563                 csum_size = sizeof(struct ext2_dir_entry_tail);
 564
 565         outdir->max = 0;
 566         retval = alloc_size_dir(fs, outdir,
 567                                 (fd->dir_size / fs->blocksize) + 2);
 568         if (retval)
 569                 return retval;
 570         outdir->num = fd->compress ? 0 : 1;
 571         offset = 0;
 572         outdir->hashes[0] = 0;
 573         prev_hash = 1;
 574         if ((retval = get_next_block(fs, outdir, &block_start)))
 575                 return retval;
 576         dirent = (struct ext2_dir_entry *) block_start;
 577         prev_rec_len = 0;
 578         rec_len = 0;
 579         left = fs->blocksize - csum_size;
 580         slack = fd->compress ? min_rec_len :
 581                 ((fs->blocksize - csum_size) * ctx->htree_slack_percentage)/100;
 582         if (slack < min_rec_len)
 583                 slack = min_rec_len;
 584         for (i = 0; i < fd->num_array; i++) {
 585                 ent = fd->harray + i;
 586                 if (ent->dir->inode == 0)
 587                         continue;
 588                 rec_len = ext2fs_dir_rec_len(ext2fs_dirent_name_len(ent->dir),
 589                                              hash_in_entry);
 590                 if (rec_len > left) {
 591                         if (left) {
 592                                 left += prev_rec_len;
 593                                 retval = ext2fs_set_rec_len(fs, left, dirent);
 594                                 if (retval)
 595                                         return retval;
 596                         }
 597                         if (csum_size) {
 598                                 t = EXT2_DIRENT_TAIL(block_start,
 599                                                      fs->blocksize);
 600                                 ext2fs_initialize_dirent_tail(fs, t);
 601                         }
 602                         if ((retval = get_next_block(fs, outdir,
 603                                                       &block_start)))
 604                                 return retval;
 605                         offset = 0;
 606                 }
 607                 left = (fs->blocksize - csum_size) - offset;
 608                 dirent = (struct ext2_dir_entry *) (block_start + offset);
 609                 if (offset == 0) {
 610                         if (ent->hash == prev_hash)
 611                                 outdir->hashes[outdir->num-1] = ent->hash | 1;
 612                         else
 613                                 outdir->hashes[outdir->num-1] = ent->hash;
 614                 }
 615                 dirent->inode = ent->dir->inode;
 616                 ext2fs_dirent_set_name_len(dirent,
 617                                            ext2fs_dirent_name_len(ent->dir));
 618                 ext2fs_dirent_set_file_type(dirent,
 619                                             ext2fs_dirent_file_type(ent->dir));
 620                 retval = ext2fs_set_rec_len(fs, rec_len, dirent);
 621                 if (retval)
 622                         return retval;
 623                 prev_rec_len = rec_len;
 624                 memcpy(dirent->name, ent->dir->name,
 625                        ext2fs_dirent_name_len(dirent));
 626                 if (hash_in_entry) {
 627                         EXT2_DIRENT_HASHES(dirent)->hash = ext2fs_cpu_to_le32(ent->hash);
 628                         EXT2_DIRENT_HASHES(dirent)->minor_hash =
 629                                                         ext2fs_cpu_to_le32(ent->minor_hash);
 630                 }
 631                 offset += rec_len;
 632                 left -= rec_len;
 633                 if (left < slack) {
 634                         prev_rec_len += left;
 635                         retval = ext2fs_set_rec_len(fs, prev_rec_len, dirent);
 636                         if (retval)
 637                                 return retval;
 638                         offset += left;
 639                         left = 0;
 640                 }
 641                 prev_hash = ent->hash;
 642         }
 643         if (left)
 644                 retval = ext2fs_set_rec_len(fs, rec_len + left, dirent);
 645         if (csum_size) {
 646                 t = EXT2_DIRENT_TAIL(block_start, fs->blocksize);
 647                 ext2fs_initialize_dirent_tail(fs, t);
 648         }
 649
 650         return retval;
 651 }
 652
 653
 654 static struct ext2_dx_root_info *set_root_node(ext2_filsys fs, char *buf,
 655                                     ext2_ino_t ino, ext2_ino_t parent,
 656                                     struct ext2_inode *inode)
 657 {
 658         struct ext2_dir_entry           *dir;
 659         struct ext2_dx_root_info        *root;
 660         struct ext2_dx_countlimit       *limits;
 661         int                             filetype = 0;
 662         int                             csum_size = 0;
 663
 664         if (ext2fs_has_feature_filetype(fs->super))
 665                 filetype = EXT2_FT_DIR;
 666
 667         memset(buf, 0, fs->blocksize);
 668         dir = (struct ext2_dir_entry *) buf;
 669         dir->inode = ino;
 670         dir->name[0] = '.';
 671         ext2fs_dirent_set_name_len(dir, 1);
 672         ext2fs_dirent_set_file_type(dir, filetype);
 673         dir->rec_len = 12;
 674         dir = (struct ext2_dir_entry *) (buf + 12);
 675         dir->inode = parent;
 676         dir->name[0] = '.';
 677         dir->name[1] = '.';
 678         ext2fs_dirent_set_name_len(dir, 2);
 679         ext2fs_dirent_set_file_type(dir, filetype);
 680         dir->rec_len = fs->blocksize - 12;
 681
 682         root = (struct ext2_dx_root_info *) (buf+24);
 683         root->reserved_zero = 0;
 684         if (ext4_hash_in_dirent(inode))
 685                 root->hash_version = EXT2_HASH_SIPHASH;
 686         else
 687                 root->hash_version = fs->super->s_def_hash_version;
 688         root->info_length = 8;
 689         root->indirect_levels = 0;
 690         root->unused_flags = 0;
 691
 692         if (ext2fs_has_feature_metadata_csum(fs->super))
 693                 csum_size = sizeof(struct ext2_dx_tail);
 694
 695         limits = (struct ext2_dx_countlimit *) (buf+32);
 696         limits->limit = (fs->blocksize - (32 + csum_size)) /
 697                         sizeof(struct ext2_dx_entry);
 698         limits->count = 0;
 699
 700         return root;
 701 }
 702
 703
 704 static struct ext2_dx_entry *set_int_node(ext2_filsys fs, char *buf)
 705 {
 706         struct ext2_dir_entry           *dir;
 707         struct ext2_dx_countlimit       *limits;
 708         int                             csum_size = 0;
 709
 710         memset(buf, 0, fs->blocksize);
 711         dir = (struct ext2_dir_entry *) buf;
 712         dir->inode = 0;
 713         (void) ext2fs_set_rec_len(fs, fs->blocksize, dir);
 714
 715         if (ext2fs_has_feature_metadata_csum(fs->super))
 716                 csum_size = sizeof(struct ext2_dx_tail);
 717
 718         limits = (struct ext2_dx_countlimit *) (buf+8);
 719         limits->limit = (fs->blocksize - (8 + csum_size)) /
 720                         sizeof(struct ext2_dx_entry);
 721         limits->count = 0;
 722
 723         return (struct ext2_dx_entry *) limits;
 724 }
 725
 726 static int alloc_blocks(ext2_filsys fs,
 727                         struct ext2_dx_countlimit **limit,
 728                         struct ext2_dx_entry **prev_ent,
 729                         struct ext2_dx_entry **next_ent,
 730                         int *prev_offset, int *next_offset,
 731                         struct out_dir *outdir, int i,
 732                         int *prev_count, int *next_count)
 733 {
 734         errcode_t       retval;
 735         char            *block_start;
 736
 737         if (*limit)
 738                 (*limit)->limit = (*limit)->count =
 739                         ext2fs_cpu_to_le16((*limit)->limit);
 740         *prev_ent = (struct ext2_dx_entry *) (outdir->buf + *prev_offset);
 741         (*prev_ent)->block = ext2fs_cpu_to_le32(outdir->num);
 742
 743         if (i != 1)
 744                 (*prev_ent)->hash =
 745                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 746
 747         retval = get_next_block(fs, outdir, &block_start);
 748         if (retval)
 749                 return retval;
 750
 751         /* outdir->buf might be reallocated */
 752         *prev_ent = (struct ext2_dx_entry *) (outdir->buf + *prev_offset);
 753
 754         *next_ent = set_int_node(fs, block_start);
 755         *limit = (struct ext2_dx_countlimit *)(*next_ent);
 756         if (next_offset)
 757                 *next_offset = ((char *) *next_ent - outdir->buf);
 758
 759         *next_count = (*limit)->limit;
 760         (*prev_offset) += sizeof(struct ext2_dx_entry);
 761         (*prev_count)--;
 762
 763         return 0;
 764 }
 765
 766 /*
 767  * This function takes the leaf nodes which have been written in
 768  * outdir, and populates the root node and any necessary interior nodes.
 769  */
 770 static errcode_t calculate_tree(ext2_filsys fs,
 771                                 struct out_dir *outdir,
 772                                 ext2_ino_t ino,
 773                                 ext2_ino_t parent,
 774                                 struct ext2_inode *inode)
 775 {
 776         struct ext2_dx_root_info        *root_info;
 777         struct ext2_dx_entry            *root, *int_ent, *dx_ent = 0;
 778         struct ext2_dx_countlimit       *root_limit, *int_limit, *limit;
 779         errcode_t                       retval;
 780         int                             i, c1, c2, c3, nblks;
 781         int                             limit_offset, int_offset, root_offset;
 782
 783         root_info = set_root_node(fs, outdir->buf, ino, parent, inode);
 784         root_offset = limit_offset = ((char *) root_info - outdir->buf) +
 785                 root_info->info_length;
 786         root_limit = (struct ext2_dx_countlimit *) (outdir->buf + limit_offset);
 787         c1 = root_limit->limit;
 788         nblks = outdir->num;
 789
 790         /* Write out the pointer blocks */
 791         if (nblks - 1 <= c1) {
 792                 /* Just write out the root block, and we're done */
 793                 root = (struct ext2_dx_entry *) (outdir->buf + root_offset);
 794                 for (i=1; i < nblks; i++) {
 795                         root->block = ext2fs_cpu_to_le32(i);
 796                         if (i != 1)
 797                                 root->hash =
 798                                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 799                         root++;
 800                         c1--;
 801                 }
 802         } else if (nblks - 1 <= ext2fs_htree_intnode_maxrecs(fs, c1)) {
 803                 c2 = 0;
 804                 limit = NULL;
 805                 root_info->indirect_levels = 1;
 806                 for (i=1; i < nblks; i++) {
 807                         if (c2 == 0 && c1 == 0)
 808                                 return ENOSPC;
 809                         if (c2 == 0) {
 810                                 retval = alloc_blocks(fs, &limit, &root,
 811                                                       &dx_ent, &root_offset,
 812                                                       NULL, outdir, i, &c1,
 813                                                       &c2);
 814                                 if (retval)
 815                                         return retval;
 816                         }
 817                         dx_ent->block = ext2fs_cpu_to_le32(i);
 818                         if (c2 != limit->limit)
 819                                 dx_ent->hash =
 820                                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 821                         dx_ent++;
 822                         c2--;
 823                 }
 824                 limit->count = ext2fs_cpu_to_le16(limit->limit - c2);
 825                 limit->limit = ext2fs_cpu_to_le16(limit->limit);
 826         } else {
 827                 c2 = 0;
 828                 c3 = 0;
 829                 limit = NULL;
 830                 int_limit = 0;
 831                 root_info->indirect_levels = 2;
 832                 for (i = 1; i < nblks; i++) {
 833                         if (c3 == 0 && c2 == 0 && c1 == 0)
 834                                 return ENOSPC;
 835                         if (c3 == 0 && c2 == 0) {
 836                                 retval = alloc_blocks(fs, &int_limit, &root,
 837                                                       &int_ent, &root_offset,
 838                                                       &int_offset, outdir, i,
 839                                                       &c1, &c2);
 840                                 if (retval)
 841                                         return retval;
 842                         }
 843                         if (c3 == 0) {
 844                                 int delta1 = (char *)int_limit - outdir->buf;
 845                                 int delta2 = (char *)root - outdir->buf;
 846
 847                                 retval = alloc_blocks(fs, &limit, &int_ent,
 848                                                       &dx_ent, &int_offset,
 849                                                       NULL, outdir, i, &c2,
 850                                                       &c3);
 851                                 if (retval)
 852                                         return retval;
 853
 854                                 /* outdir->buf might be reallocated */
 855                                 int_limit = (struct ext2_dx_countlimit *)
 856                                         (outdir->buf + delta1);
 857                                 root = (struct ext2_dx_entry *)
 858                                         (outdir->buf + delta2);
 859                         }
 860                         dx_ent->block = ext2fs_cpu_to_le32(i);
 861                         if (c3 != limit->limit)
 862                                 dx_ent->hash =
 863                                         ext2fs_cpu_to_le32(outdir->hashes[i]);
 864                         dx_ent++;
 865                         c3--;
 866                 }
 867                 int_limit->count = ext2fs_cpu_to_le16(limit->limit - c2);
 868                 int_limit->limit = ext2fs_cpu_to_le16(limit->limit);
 869
 870                 limit->count = ext2fs_cpu_to_le16(limit->limit - c3);
 871                 limit->limit = ext2fs_cpu_to_le16(limit->limit);
 872
 873         }
 874         root_limit = (struct ext2_dx_countlimit *) (outdir->buf + limit_offset);
 875         root_limit->count = ext2fs_cpu_to_le16(root_limit->limit - c1);
 876         root_limit->limit = ext2fs_cpu_to_le16(root_limit->limit);
 877
 878         return 0;
 879 }
 880
 881 struct write_dir_struct {
 882         struct out_dir *outdir;
 883         errcode_t       err;
 884         ext2_ino_t      ino;
 885         e2fsck_t        ctx;
 886         ext2_ino_t      dir;
 887 };
 888
 889 /*
 890  * Helper function which writes out a directory block.
 891  */
 892 static int write_dir_block(ext2_filsys fs,
 893                            blk64_t *block_nr,
 894                            e2_blkcnt_t blockcnt,
 895                            blk64_t ref_block EXT2FS_ATTR((unused)),
 896                            int ref_offset EXT2FS_ATTR((unused)),
 897                            void *priv_data)
 898 {
 899         struct write_dir_struct *wd = (struct write_dir_struct *) priv_data;
 900         char    *dir, *buf = 0;
 901
 902 #ifdef REHASH_DEBUG
 903         printf("%u: write_dir_block %lld:%lld", wd->ino, blockcnt, *block_nr);
 904 #endif
 905         if ((*block_nr == 0) || (blockcnt < 0)) {
 906 #ifdef REHASH_DEBUG
 907                 printf(" - skip\n");
 908 #endif
 909                 return 0;
 910         }
 911         if (blockcnt < wd->outdir->num)
 912                 dir = wd->outdir->buf + (blockcnt * fs->blocksize);
 913         else if (wd->ctx->lost_and_found == wd->dir) {
 914                 /* Don't release any extra directory blocks for lost+found */
 915                 wd->err = ext2fs_new_dir_block(fs, 0, 0, &buf);
 916                 if (wd->err)
 917                         return BLOCK_ABORT;
 918                 dir = buf;
 919                 wd->outdir->num++;
 920         } else {
 921                 /* Don't free blocks at the end of the directory, they
 922                  * will be truncated by the caller. */
 923 #ifdef REHASH_DEBUG
 924                 printf(" - not freed\n");
 925 #endif
 926                 return 0;
 927         }
 928         wd->err = ext2fs_write_dir_block4(fs, *block_nr, dir, 0, wd->dir);
 929         if (buf)
 930                 ext2fs_free_mem(&buf);
 931
 932 #ifdef REHASH_DEBUG
 933         printf(" - write (%d)\n", wd->err);
 934 #endif
 935         if (wd->err)
 936                 return BLOCK_ABORT;
 937         return 0;
 938 }
 939
 940 static errcode_t write_directory(e2fsck_t ctx, ext2_filsys fs,
 941                                  struct out_dir *outdir,
 942                                  ext2_ino_t ino, struct ext2_inode *inode,
 943                                  int compress)
 944 {
 945         struct write_dir_struct wd;
 946         errcode_t       retval;
 947
 948         retval = e2fsck_expand_directory(ctx, ino, -1, outdir->num);
 949         if (retval)
 950                 return retval;
 951
 952         wd.outdir = outdir;
 953         wd.err = 0;
 954         wd.ino = ino;
 955         wd.ctx = ctx;
 956         wd.dir = ino;
 957
 958         retval = ext2fs_block_iterate3(fs, ino, 0, NULL,
 959                                        write_dir_block, &wd);
 960         if (retval)
 961                 return retval;
 962         if (wd.err)
 963                 return wd.err;
 964
 965         e2fsck_read_inode(ctx, ino, inode, "rehash_dir");
 966         if (compress)
 967                 inode->i_flags &= ~EXT2_INDEX_FL;
 968         else
 969                 inode->i_flags |= EXT2_INDEX_FL;
 970 #ifdef REHASH_DEBUG
 971         printf("%u: set inode size to %u blocks = %u bytes\n",
 972                ino, outdir->num, outdir->num * fs->blocksize);
 973 #endif
 974         retval = ext2fs_inode_size_set(fs, inode, (ext2_off64_t)outdir->num *
 975                                                    fs->blocksize);
 976         if (retval)
 977                 return retval;
 978
 979         /* ext2fs_punch() calls ext2fs_write_inode() which writes the size */
 980         return ext2fs_punch(fs, ino, inode, NULL, outdir->num, ~0ULL);
 981 }
 982
 983 errcode_t e2fsck_rehash_dir(e2fsck_t ctx, ext2_ino_t ino,
 984                             struct problem_context *pctx)
 985 {
 986         ext2_filsys             fs = ctx->fs;
 987         errcode_t               retval;
 988         struct ext2_inode       inode;
 989         char                    *dir_buf = 0;
 990         struct fill_dir_struct  fd = { NULL, NULL, 0, 0, 0, NULL,
 991                                        0, 0, 0, 0, 0, 0 };
 992         struct out_dir          outdir = { 0, 0, 0, 0 };
 993         struct name_cmp_ctx name_cmp_ctx = {0, NULL};
 994
 995         e2fsck_read_inode(ctx, ino, &inode, "rehash_dir");
 996
 997         if (ext2fs_has_feature_inline_data(fs->super) &&
 998            (inode.i_flags & EXT4_INLINE_DATA_FL))
 999                 return 0;
1000
1001         retval = ext2fs_get_mem(inode.i_size, &dir_buf);
1002         if (retval)
1003                 goto errout;
1004
1005         fd.max_array = inode.i_size / 32;
1006         retval = ext2fs_get_array(sizeof(struct hash_entry),
1007                                   fd.max_array, &fd.harray);
1008         if (retval)
1009                 goto errout;
1010
1011         fd.ino = ino;
1012         fd.ctx = ctx;
1013         fd.buf = dir_buf;
1014         fd.inode = &inode;
1015         fd.dir = ino;
1016         if (!ext2fs_has_feature_dir_index(fs->super) ||
1017             (inode.i_size / fs->blocksize) < 2)
1018                 fd.compress = 1;
1019         fd.parent = 0;
1020
1021         if (fs->encoding && (inode.i_flags & EXT4_CASEFOLD_FL)) {
1022                 name_cmp_ctx.casefold = 1;
1023                 name_cmp_ctx.tbl = fs->encoding;
1024         }
1025
1026 retry_nohash:
1027         /* Read in the entire directory into memory */
1028         retval = ext2fs_block_iterate3(fs, ino, 0, 0,
1029                                        fill_dir_block, &fd);
1030         if (fd.err) {
1031                 retval = fd.err;
1032                 goto errout;
1033         }
1034
1035         /*
1036          * If the entries read are less than a block, then don't index
1037          * the directory
1038          */
1039         if (!fd.compress && (fd.dir_size < (fs->blocksize - 24))) {
1040                 fd.compress = 1;
1041                 fd.dir_size = 0;
1042                 fd.num_array = 0;
1043                 goto retry_nohash;
1044         }
1045
1046 #if 0
1047         printf("%d entries (%d bytes) found in inode %d\n",
1048                fd.num_array, fd.dir_size, ino);
1049 #endif
1050
1051         /* Sort the list */
1052 resort:
1053         if (fd.compress && fd.num_array > 1)
1054                 sort_r_simple(fd.harray+2, fd.num_array-2,
1055                               sizeof(struct hash_entry),
1056                               hash_cmp, &name_cmp_ctx);
1057         else
1058                 sort_r_simple(fd.harray, fd.num_array,
1059                               sizeof(struct hash_entry),
1060                               hash_cmp, &name_cmp_ctx);
1061
1062         /*
1063          * Look for duplicates
1064          */
1065         if (duplicate_search_and_fix(ctx, fs, ino, &fd, &name_cmp_ctx))
1066                 goto resort;
1067
1068         if (ctx->options & E2F_OPT_NO) {
1069                 retval = 0;
1070                 goto errout;
1071         }
1072
1073         /* Sort non-hashed directories by inode number */
1074         if (fd.compress && fd.num_array > 1)
1075                 qsort(fd.harray+2, fd.num_array-2,
1076                       sizeof(struct hash_entry), ino_cmp);
1077
1078         /*
1079          * Copy the directory entries.  In a htree directory these
1080          * will become the leaf nodes.
1081          */
1082         retval = copy_dir_entries(ctx, &fd, &outdir);
1083         if (retval)
1084                 goto errout;
1085
1086         free(dir_buf); dir_buf = 0;
1087
1088         if (!fd.compress) {
1089                 /* Calculate the interior nodes */
1090                 retval = calculate_tree(fs, &outdir, ino, fd.parent, fd.inode);
1091                 if (retval)
1092                         goto errout;
1093         }
1094
1095         retval = write_directory(ctx, fs, &outdir, ino, &inode, fd.compress);
1096         if (retval)
1097                 goto errout;
1098
1099         if (ctx->options & E2F_OPT_CONVERT_BMAP)
1100                 retval = e2fsck_rebuild_extents_later(ctx, ino);
1101         else
1102                 retval = e2fsck_check_rebuild_extents(ctx, ino, &inode, pctx);
1103 errout:
1104         ext2fs_free_mem(&dir_buf);
1105         ext2fs_free_mem(&fd.harray);
1106
1107         free_out_dir(&outdir);
1108         return retval;
1109 }
1110
1111 void e2fsck_rehash_directories(e2fsck_t ctx)
1112 {
1113         struct problem_context  pctx;
1114 #ifdef RESOURCE_TRACK
1115         struct resource_track   rtrack;
1116 #endif
1117         struct dir_info         *dir;
1118         ext2_u32_iterate        iter;
1119         struct dir_info_iter *  dirinfo_iter = 0;
1120         ext2_ino_t              ino;
1121         errcode_t               retval;
1122         int                     cur, max, all_dirs, first = 1;
1123
1124         init_resource_track(&rtrack, ctx->fs->io);
1125         all_dirs = ctx->options & E2F_OPT_COMPRESS_DIRS;
1126
1127         if (!ctx->dirs_to_hash && !all_dirs)
1128                 return;
1129
1130         (void) e2fsck_get_lost_and_found(ctx, 0);
1131
1132         clear_problem_context(&pctx);
1133
1134         cur = 0;
1135         if (all_dirs) {
1136                 dirinfo_iter = e2fsck_dir_info_iter_begin(ctx);
1137                 max = e2fsck_get_num_dirinfo(ctx);
1138         } else {
1139                 retval = ext2fs_u32_list_iterate_begin(ctx->dirs_to_hash,
1140                                                        &iter);
1141                 if (retval) {
1142                         pctx.errcode = retval;
1143                         fix_problem(ctx, PR_3A_OPTIMIZE_ITER, &pctx);
1144                         return;
1145                 }
1146                 max = ext2fs_u32_list_count(ctx->dirs_to_hash);
1147         }
1148         while (1) {
1149                 if (all_dirs) {
1150                         if ((dir = e2fsck_dir_info_iter(ctx,
1151                                                         dirinfo_iter)) == 0)
1152                                 break;
1153                         ino = dir->ino;
1154                 } else {
1155                         if (!ext2fs_u32_list_iterate(iter, &ino))
1156                                 break;
1157                 }
1158                 if (!ext2fs_test_inode_bitmap2(ctx->inode_dir_map, ino))
1159                         continue;
1160
1161                 pctx.dir = ino;
1162                 if (first) {
1163                         fix_problem(ctx, PR_3A_PASS_HEADER, &pctx);
1164                         first = 0;
1165                 }
1166 #if 0
1167                 fix_problem(ctx, PR_3A_OPTIMIZE_DIR, &pctx);
1168 #endif
1169                 pctx.errcode = e2fsck_rehash_dir(ctx, ino, &pctx);
1170                 if (pctx.errcode) {
1171                         end_problem_latch(ctx, PR_LATCH_OPTIMIZE_DIR);
1172                         fix_problem(ctx, PR_3A_OPTIMIZE_DIR_ERR, &pctx);
1173                 }
1174                 if (ctx->progress && !ctx->progress_fd)
1175                         e2fsck_simple_progress(ctx, "Rebuilding directory",
1176                                100.0 * (float) (++cur) / (float) max, ino);
1177         }
1178         end_problem_latch(ctx, PR_LATCH_OPTIMIZE_DIR);
1179         if (all_dirs)
1180                 e2fsck_dir_info_iter_end(ctx, dirinfo_iter);
1181         else
1182                 ext2fs_u32_list_iterate_end(iter);
1183
1184         if (ctx->dirs_to_hash)
1185                 ext2fs_u32_list_free(ctx->dirs_to_hash);
1186         ctx->dirs_to_hash = 0;
1187
1188         print_resource_track(ctx, "Pass 3A", &rtrack, ctx->fs->io);
1189 }