fs/ext3/ialloc.c

   1 /*
   2  *  linux/fs/ext3/ialloc.c
   3  *
   4  * Copyright (C) 1992, 1993, 1994, 1995
   5  * Remy Card (card@masi.ibp.fr)
   6  * Laboratoire MASI - Institut Blaise Pascal
   7  * Universite Pierre et Marie Curie (Paris VI)
   8  *
   9  *  BSD ufs-inspired inode and directory allocation by
  10  *  Stephen Tweedie (sct@redhat.com), 1993
  11  *  Big-endian to little-endian byte-swapping/bitmaps by
  12  *        David S. Miller (davem@caip.rutgers.edu), 1995
  13  */
  14
  15 #include <linux/time.h>
  16 #include <linux/fs.h>
  17 #include <linux/jbd.h>
  18 #include <linux/ext3_fs.h>
  19 #include <linux/ext3_jbd.h>
  20 #include <linux/stat.h>
  21 #include <linux/string.h>
  22 #include <linux/quotaops.h>
  23 #include <linux/buffer_head.h>
  24 #include <linux/random.h>
  25 #include <linux/bitops.h>
  26 #include <trace/events/ext3.h>
  27
  28 #include <asm/byteorder.h>
  29
  30 #include "xattr.h"
  31 #include "acl.h"
  32
  33 /*
  34  * ialloc.c contains the inodes allocation and deallocation routines
  35  */
  36
  37 /*
  38  * The free inodes are managed by bitmaps.  A file system contains several
  39  * blocks groups.  Each group contains 1 bitmap block for blocks, 1 bitmap
  40  * block for inodes, N blocks for the inode table and data blocks.
  41  *
  42  * The file system contains group descriptors which are located after the
  43  * super block.  Each descriptor contains the number of the bitmap block and
  44  * the free blocks count in the block.
  45  */
  46
  47
  48 /*
  49  * Read the inode allocation bitmap for a given block_group, reading
  50  * into the specified slot in the superblock's bitmap cache.
  51  *
  52  * Return buffer_head of bitmap on success or NULL.
  53  */
  54 static struct buffer_head *
  55 read_inode_bitmap(struct super_block * sb, unsigned long block_group)
  56 {
  57         struct ext3_group_desc *desc;
  58         struct buffer_head *bh = NULL;
  59
  60         desc = ext3_get_group_desc(sb, block_group, NULL);
  61         if (!desc)
  62                 goto error_out;
  63
  64         bh = sb_bread(sb, le32_to_cpu(desc->bg_inode_bitmap));
  65         if (!bh)
  66                 ext3_error(sb, "read_inode_bitmap",
  67                             "Cannot read inode bitmap - "
  68                             "block_group = %lu, inode_bitmap = %u",
  69                             block_group, le32_to_cpu(desc->bg_inode_bitmap));
  70 error_out:
  71         return bh;
  72 }
  73
  74 /*
  75  * NOTE! When we get the inode, we're the only people
  76  * that have access to it, and as such there are no
  77  * race conditions we have to worry about. The inode
  78  * is not on the hash-lists, and it cannot be reached
  79  * through the filesystem because the directory entry
  80  * has been deleted earlier.
  81  *
  82  * HOWEVER: we must make sure that we get no aliases,
  83  * which means that we have to call "clear_inode()"
  84  * _before_ we mark the inode not in use in the inode
  85  * bitmaps. Otherwise a newly created file might use
  86  * the same inode number (not actually the same pointer
  87  * though), and then we'd have two inodes sharing the
  88  * same inode number and space on the harddisk.
  89  */
  90 void ext3_free_inode (handle_t *handle, struct inode * inode)
  91 {
  92         struct super_block * sb = inode->i_sb;
  93         int is_directory;
  94         unsigned long ino;
  95         struct buffer_head *bitmap_bh = NULL;
  96         struct buffer_head *bh2;
  97         unsigned long block_group;
  98         unsigned long bit;
  99         struct ext3_group_desc * gdp;
 100         struct ext3_super_block * es;
 101         struct ext3_sb_info *sbi;
 102         int fatal = 0, err;
 103
 104         if (atomic_read(&inode->i_count) > 1) {
 105                 printk ("ext3_free_inode: inode has count=%d\n",
 106                                         atomic_read(&inode->i_count));
 107                 return;
 108         }
 109         if (inode->i_nlink) {
 110                 printk ("ext3_free_inode: inode has nlink=%d\n",
 111                         inode->i_nlink);
 112                 return;
 113         }
 114         if (!sb) {
 115                 printk("ext3_free_inode: inode on nonexistent device\n");
 116                 return;
 117         }
 118         sbi = EXT3_SB(sb);
 119
 120         ino = inode->i_ino;
 121         ext3_debug ("freeing inode %lu\n", ino);
 122         trace_ext3_free_inode(inode);
 123
 124         is_directory = S_ISDIR(inode->i_mode);
 125
 126         es = EXT3_SB(sb)->s_es;
 127         if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
 128                 ext3_error (sb, "ext3_free_inode",
 129                             "reserved or nonexistent inode %lu", ino);
 130                 goto error_return;
 131         }
 132         block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
 133         bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
 134         bitmap_bh = read_inode_bitmap(sb, block_group);
 135         if (!bitmap_bh)
 136                 goto error_return;
 137
 138         BUFFER_TRACE(bitmap_bh, "get_write_access");
 139         fatal = ext3_journal_get_write_access(handle, bitmap_bh);
 140         if (fatal)
 141                 goto error_return;
 142
 143         /* Ok, now we can actually update the inode bitmaps.. */
 144         if (!ext3_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
 145                                         bit, bitmap_bh->b_data))
 146                 ext3_error (sb, "ext3_free_inode",
 147                               "bit already cleared for inode %lu", ino);
 148         else {
 149                 gdp = ext3_get_group_desc (sb, block_group, &bh2);
 150
 151                 BUFFER_TRACE(bh2, "get_write_access");
 152                 fatal = ext3_journal_get_write_access(handle, bh2);
 153                 if (fatal) goto error_return;
 154
 155                 if (gdp) {
 156                         spin_lock(sb_bgl_lock(sbi, block_group));
 157                         le16_add_cpu(&gdp->bg_free_inodes_count, 1);
 158                         if (is_directory)
 159                                 le16_add_cpu(&gdp->bg_used_dirs_count, -1);
 160                         spin_unlock(sb_bgl_lock(sbi, block_group));
 161                         percpu_counter_inc(&sbi->s_freeinodes_counter);
 162                         if (is_directory)
 163                                 percpu_counter_dec(&sbi->s_dirs_counter);
 164
 165                 }
 166                 BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
 167                 err = ext3_journal_dirty_metadata(handle, bh2);
 168                 if (!fatal) fatal = err;
 169         }
 170         BUFFER_TRACE(bitmap_bh, "call ext3_journal_dirty_metadata");
 171         err = ext3_journal_dirty_metadata(handle, bitmap_bh);
 172         if (!fatal)
 173                 fatal = err;
 174
 175 error_return:
 176         brelse(bitmap_bh);
 177         ext3_std_error(sb, fatal);
 178 }
 179
 180 /*
 181  * Orlov's allocator for directories.
 182  *
 183  * We always try to spread first-level directories.
 184  *
 185  * If there are blockgroups with both free inodes and free blocks counts
 186  * not worse than average we return one with smallest directory count.
 187  * Otherwise we simply return a random group.
 188  *
 189  * For the rest rules look so:
 190  *
 191  * It's OK to put directory into a group unless
 192  * it has too many directories already (max_dirs) or
 193  * it has too few free inodes left (min_inodes) or
 194  * it has too few free blocks left (min_blocks) or
 195  * it's already running too large debt (max_debt).
 196  * Parent's group is preferred, if it doesn't satisfy these
 197  * conditions we search cyclically through the rest. If none
 198  * of the groups look good we just look for a group with more
 199  * free inodes than average (starting at parent's group).
 200  *
 201  * Debt is incremented each time we allocate a directory and decremented
 202  * when we allocate an inode, within 0--255.
 203  */
 204
 205 #define INODE_COST 64
 206 #define BLOCK_COST 256
 207
 208 static int find_group_orlov(struct super_block *sb, struct inode *parent)
 209 {
 210         int parent_group = EXT3_I(parent)->i_block_group;
 211         struct ext3_sb_info *sbi = EXT3_SB(sb);
 212         struct ext3_super_block *es = sbi->s_es;
 213         int ngroups = sbi->s_groups_count;
 214         int inodes_per_group = EXT3_INODES_PER_GROUP(sb);
 215         unsigned int freei, avefreei;
 216         ext3_fsblk_t freeb, avefreeb;
 217         ext3_fsblk_t blocks_per_dir;
 218         unsigned int ndirs;
 219         int max_debt, max_dirs, min_inodes;
 220         ext3_grpblk_t min_blocks;
 221         int group = -1, i;
 222         struct ext3_group_desc *desc;
 223
 224         freei = percpu_counter_read_positive(&sbi->s_freeinodes_counter);
 225         avefreei = freei / ngroups;
 226         freeb = percpu_counter_read_positive(&sbi->s_freeblocks_counter);
 227         avefreeb = freeb / ngroups;
 228         ndirs = percpu_counter_read_positive(&sbi->s_dirs_counter);
 229
 230         if ((parent == sb->s_root->d_inode) ||
 231             (EXT3_I(parent)->i_flags & EXT3_TOPDIR_FL)) {
 232                 int best_ndir = inodes_per_group;
 233                 int best_group = -1;
 234
 235                 get_random_bytes(&group, sizeof(group));
 236                 parent_group = (unsigned)group % ngroups;
 237                 for (i = 0; i < ngroups; i++) {
 238                         group = (parent_group + i) % ngroups;
 239                         desc = ext3_get_group_desc (sb, group, NULL);
 240                         if (!desc || !desc->bg_free_inodes_count)
 241                                 continue;
 242                         if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir)
 243                                 continue;
 244                         if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei)
 245                                 continue;
 246                         if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb)
 247                                 continue;
 248                         best_group = group;
 249                         best_ndir = le16_to_cpu(desc->bg_used_dirs_count);
 250                 }
 251                 if (best_group >= 0)
 252                         return best_group;
 253                 goto fallback;
 254         }
 255
 256         blocks_per_dir = (le32_to_cpu(es->s_blocks_count) - freeb) / ndirs;
 257
 258         max_dirs = ndirs / ngroups + inodes_per_group / 16;
 259         min_inodes = avefreei - inodes_per_group / 4;
 260         min_blocks = avefreeb - EXT3_BLOCKS_PER_GROUP(sb) / 4;
 261
 262         max_debt = EXT3_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, (ext3_fsblk_t)BLOCK_COST);
 263         if (max_debt * INODE_COST > inodes_per_group)
 264                 max_debt = inodes_per_group / INODE_COST;
 265         if (max_debt > 255)
 266                 max_debt = 255;
 267         if (max_debt == 0)
 268                 max_debt = 1;
 269
 270         for (i = 0; i < ngroups; i++) {
 271                 group = (parent_group + i) % ngroups;
 272                 desc = ext3_get_group_desc (sb, group, NULL);
 273                 if (!desc || !desc->bg_free_inodes_count)
 274                         continue;
 275                 if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs)
 276                         continue;
 277                 if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes)
 278                         continue;
 279                 if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks)
 280                         continue;
 281                 return group;
 282         }
 283
 284 fallback:
 285         for (i = 0; i < ngroups; i++) {
 286                 group = (parent_group + i) % ngroups;
 287                 desc = ext3_get_group_desc (sb, group, NULL);
 288                 if (!desc || !desc->bg_free_inodes_count)
 289                         continue;
 290                 if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei)
 291                         return group;
 292         }
 293
 294         if (avefreei) {
 295                 /*
 296                  * The free-inodes counter is approximate, and for really small
 297                  * filesystems the above test can fail to find any blockgroups
 298                  */
 299                 avefreei = 0;
 300                 goto fallback;
 301         }
 302
 303         return -1;
 304 }
 305
 306 static int find_group_other(struct super_block *sb, struct inode *parent)
 307 {
 308         int parent_group = EXT3_I(parent)->i_block_group;
 309         int ngroups = EXT3_SB(sb)->s_groups_count;
 310         struct ext3_group_desc *desc;
 311         int group, i;
 312
 313         /*
 314          * Try to place the inode in its parent directory
 315          */
 316         group = parent_group;
 317         desc = ext3_get_group_desc (sb, group, NULL);
 318         if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
 319                         le16_to_cpu(desc->bg_free_blocks_count))
 320                 return group;
 321
 322         /*
 323          * We're going to place this inode in a different blockgroup from its
 324          * parent.  We want to cause files in a common directory to all land in
 325          * the same blockgroup.  But we want files which are in a different
 326          * directory which shares a blockgroup with our parent to land in a
 327          * different blockgroup.
 328          *
 329          * So add our directory's i_ino into the starting point for the hash.
 330          */
 331         group = (group + parent->i_ino) % ngroups;
 332
 333         /*
 334          * Use a quadratic hash to find a group with a free inode and some free
 335          * blocks.
 336          */
 337         for (i = 1; i < ngroups; i <<= 1) {
 338                 group += i;
 339                 if (group >= ngroups)
 340                         group -= ngroups;
 341                 desc = ext3_get_group_desc (sb, group, NULL);
 342                 if (desc && le16_to_cpu(desc->bg_free_inodes_count) &&
 343                                 le16_to_cpu(desc->bg_free_blocks_count))
 344                         return group;
 345         }
 346
 347         /*
 348          * That failed: try linear search for a free inode, even if that group
 349          * has no free blocks.
 350          */
 351         group = parent_group;
 352         for (i = 0; i < ngroups; i++) {
 353                 if (++group >= ngroups)
 354                         group = 0;
 355                 desc = ext3_get_group_desc (sb, group, NULL);
 356                 if (desc && le16_to_cpu(desc->bg_free_inodes_count))
 357                         return group;
 358         }
 359
 360         return -1;
 361 }
 362
 363 /*
 364  * There are two policies for allocating an inode.  If the new inode is
 365  * a directory, then a forward search is made for a block group with both
 366  * free space and a low directory-to-inode ratio; if that fails, then of
 367  * the groups with above-average free space, that group with the fewest
 368  * directories already is chosen.
 369  *
 370  * For other inodes, search forward from the parent directory's block
 371  * group to find a free inode.
 372  */
 373 struct inode *ext3_new_inode(handle_t *handle, struct inode * dir,
 374                              const struct qstr *qstr, int mode)
 375 {
 376         struct super_block *sb;
 377         struct buffer_head *bitmap_bh = NULL;
 378         struct buffer_head *bh2;
 379         int group;
 380         unsigned long ino = 0;
 381         struct inode * inode;
 382         struct ext3_group_desc * gdp = NULL;
 383         struct ext3_super_block * es;
 384         struct ext3_inode_info *ei;
 385         struct ext3_sb_info *sbi;
 386         int err = 0;
 387         struct inode *ret;
 388         int i;
 389
 390         /* Cannot create files in a deleted directory */
 391         if (!dir || !dir->i_nlink)
 392                 return ERR_PTR(-EPERM);
 393
 394         sb = dir->i_sb;
 395         trace_ext3_request_inode(dir, mode);
 396         inode = new_inode(sb);
 397         if (!inode)
 398                 return ERR_PTR(-ENOMEM);
 399         ei = EXT3_I(inode);
 400
 401         sbi = EXT3_SB(sb);
 402         es = sbi->s_es;
 403         if (S_ISDIR(mode))
 404                 group = find_group_orlov(sb, dir);
 405         else
 406                 group = find_group_other(sb, dir);
 407
 408         err = -ENOSPC;
 409         if (group == -1)
 410                 goto out;
 411
 412         for (i = 0; i < sbi->s_groups_count; i++) {
 413                 err = -EIO;
 414
 415                 gdp = ext3_get_group_desc(sb, group, &bh2);
 416                 if (!gdp)
 417                         goto fail;
 418
 419                 brelse(bitmap_bh);
 420                 bitmap_bh = read_inode_bitmap(sb, group);
 421                 if (!bitmap_bh)
 422                         goto fail;
 423
 424                 ino = 0;
 425
 426 repeat_in_this_group:
 427                 ino = ext3_find_next_zero_bit((unsigned long *)
 428                                 bitmap_bh->b_data, EXT3_INODES_PER_GROUP(sb), ino);
 429                 if (ino < EXT3_INODES_PER_GROUP(sb)) {
 430
 431                         BUFFER_TRACE(bitmap_bh, "get_write_access");
 432                         err = ext3_journal_get_write_access(handle, bitmap_bh);
 433                         if (err)
 434                                 goto fail;
 435
 436                         if (!ext3_set_bit_atomic(sb_bgl_lock(sbi, group),
 437                                                 ino, bitmap_bh->b_data)) {
 438                                 /* we won it */
 439                                 BUFFER_TRACE(bitmap_bh,
 440                                         "call ext3_journal_dirty_metadata");
 441                                 err = ext3_journal_dirty_metadata(handle,
 442                                                                 bitmap_bh);
 443                                 if (err)
 444                                         goto fail;
 445                                 goto got;
 446                         }
 447                         /* we lost it */
 448                         journal_release_buffer(handle, bitmap_bh);
 449
 450                         if (++ino < EXT3_INODES_PER_GROUP(sb))
 451                                 goto repeat_in_this_group;
 452                 }
 453
 454                 /*
 455                  * This case is possible in concurrent environment.  It is very
 456                  * rare.  We cannot repeat the find_group_xxx() call because
 457                  * that will simply return the same blockgroup, because the
 458                  * group descriptor metadata has not yet been updated.
 459                  * So we just go onto the next blockgroup.
 460                  */
 461                 if (++group == sbi->s_groups_count)
 462                         group = 0;
 463         }
 464         err = -ENOSPC;
 465         goto out;
 466
 467 got:
 468         ino += group * EXT3_INODES_PER_GROUP(sb) + 1;
 469         if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
 470                 ext3_error (sb, "ext3_new_inode",
 471                             "reserved inode or inode > inodes count - "
 472                             "block_group = %d, inode=%lu", group, ino);
 473                 err = -EIO;
 474                 goto fail;
 475         }
 476
 477         BUFFER_TRACE(bh2, "get_write_access");
 478         err = ext3_journal_get_write_access(handle, bh2);
 479         if (err) goto fail;
 480         spin_lock(sb_bgl_lock(sbi, group));
 481         le16_add_cpu(&gdp->bg_free_inodes_count, -1);
 482         if (S_ISDIR(mode)) {
 483                 le16_add_cpu(&gdp->bg_used_dirs_count, 1);
 484         }
 485         spin_unlock(sb_bgl_lock(sbi, group));
 486         BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata");
 487         err = ext3_journal_dirty_metadata(handle, bh2);
 488         if (err) goto fail;
 489
 490         percpu_counter_dec(&sbi->s_freeinodes_counter);
 491         if (S_ISDIR(mode))
 492                 percpu_counter_inc(&sbi->s_dirs_counter);
 493
 494
 495         if (test_opt(sb, GRPID)) {
 496                 inode->i_mode = mode;
 497                 inode->i_uid = current_fsuid();
 498                 inode->i_gid = dir->i_gid;
 499         } else
 500                 inode_init_owner(inode, dir, mode);
 501
 502         inode->i_ino = ino;
 503         /* This is the optimal IO size (for stat), not the fs block size */
 504         inode->i_blocks = 0;
 505         inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
 506
 507         memset(ei->i_data, 0, sizeof(ei->i_data));
 508         ei->i_dir_start_lookup = 0;
 509         ei->i_disksize = 0;
 510
 511         ei->i_flags =
 512                 ext3_mask_flags(mode, EXT3_I(dir)->i_flags & EXT3_FL_INHERITED);
 513 #ifdef EXT3_FRAGMENTS
 514         ei->i_faddr = 0;
 515         ei->i_frag_no = 0;
 516         ei->i_frag_size = 0;
 517 #endif
 518         ei->i_file_acl = 0;
 519         ei->i_dir_acl = 0;
 520         ei->i_dtime = 0;
 521         ei->i_block_alloc_info = NULL;
 522         ei->i_block_group = group;
 523
 524         ext3_set_inode_flags(inode);
 525         if (IS_DIRSYNC(inode))
 526                 handle->h_sync = 1;
 527         if (insert_inode_locked(inode) < 0) {
 528                 err = -EINVAL;
 529                 goto fail_drop;
 530         }
 531         spin_lock(&sbi->s_next_gen_lock);
 532         inode->i_generation = sbi->s_next_generation++;
 533         spin_unlock(&sbi->s_next_gen_lock);
 534
 535         ei->i_state_flags = 0;
 536         ext3_set_inode_state(inode, EXT3_STATE_NEW);
 537
 538         /* See comment in ext3_iget for explanation */
 539         if (ino >= EXT3_FIRST_INO(sb) + 1 &&
 540             EXT3_INODE_SIZE(sb) > EXT3_GOOD_OLD_INODE_SIZE) {
 541                 ei->i_extra_isize =
 542                         sizeof(struct ext3_inode) - EXT3_GOOD_OLD_INODE_SIZE;
 543         } else {
 544                 ei->i_extra_isize = 0;
 545         }
 546
 547         ret = inode;
 548         dquot_initialize(inode);
 549         err = dquot_alloc_inode(inode);
 550         if (err)
 551                 goto fail_drop;
 552
 553         err = ext3_init_acl(handle, inode, dir);
 554         if (err)
 555                 goto fail_free_drop;
 556
 557         err = ext3_init_security(handle, inode, dir, qstr);
 558         if (err)
 559                 goto fail_free_drop;
 560
 561         err = ext3_mark_inode_dirty(handle, inode);
 562         if (err) {
 563                 ext3_std_error(sb, err);
 564                 goto fail_free_drop;
 565         }
 566
 567         ext3_debug("allocating inode %lu\n", inode->i_ino);
 568         trace_ext3_allocate_inode(inode, dir, mode);
 569         goto really_out;
 570 fail:
 571         ext3_std_error(sb, err);
 572 out:
 573         iput(inode);
 574         ret = ERR_PTR(err);
 575 really_out:
 576         brelse(bitmap_bh);
 577         return ret;
 578
 579 fail_free_drop:
 580         dquot_free_inode(inode);
 581
 582 fail_drop:
 583         dquot_drop(inode);
 584         inode->i_flags |= S_NOQUOTA;
 585         clear_nlink(inode);
 586         unlock_new_inode(inode);
 587         iput(inode);
 588         brelse(bitmap_bh);
 589         return ERR_PTR(err);
 590 }
 591
 592 /* Verify that we are loading a valid orphan from disk */
 593 struct inode *ext3_orphan_get(struct super_block *sb, unsigned long ino)
 594 {
 595         unsigned long max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count);
 596         unsigned long block_group;
 597         int bit;
 598         struct buffer_head *bitmap_bh;
 599         struct inode *inode = NULL;
 600         long err = -EIO;
 601
 602         /* Error cases - e2fsck has already cleaned up for us */
 603         if (ino > max_ino) {
 604                 ext3_warning(sb, __func__,
 605                              "bad orphan ino %lu!  e2fsck was run?", ino);
 606                 goto error;
 607         }
 608
 609         block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb);
 610         bit = (ino - 1) % EXT3_INODES_PER_GROUP(sb);
 611         bitmap_bh = read_inode_bitmap(sb, block_group);
 612         if (!bitmap_bh) {
 613                 ext3_warning(sb, __func__,
 614                              "inode bitmap error for orphan %lu", ino);
 615                 goto error;
 616         }
 617
 618         /* Having the inode bit set should be a 100% indicator that this
 619          * is a valid orphan (no e2fsck run on fs).  Orphans also include
 620          * inodes that were being truncated, so we can't check i_nlink==0.
 621          */
 622         if (!ext3_test_bit(bit, bitmap_bh->b_data))
 623                 goto bad_orphan;
 624
 625         inode = ext3_iget(sb, ino);
 626         if (IS_ERR(inode))
 627                 goto iget_failed;
 628
 629         /*
 630          * If the orphans has i_nlinks > 0 then it should be able to be
 631          * truncated, otherwise it won't be removed from the orphan list
 632          * during processing and an infinite loop will result.
 633          */
 634         if (inode->i_nlink && !ext3_can_truncate(inode))
 635                 goto bad_orphan;
 636
 637         if (NEXT_ORPHAN(inode) > max_ino)
 638                 goto bad_orphan;
 639         brelse(bitmap_bh);
 640         return inode;
 641
 642 iget_failed:
 643         err = PTR_ERR(inode);
 644         inode = NULL;
 645 bad_orphan:
 646         ext3_warning(sb, __func__,
 647                      "bad orphan inode %lu!  e2fsck was run?", ino);
 648         printk(KERN_NOTICE "ext3_test_bit(bit=%d, block=%llu) = %d\n",
 649                bit, (unsigned long long)bitmap_bh->b_blocknr,
 650                ext3_test_bit(bit, bitmap_bh->b_data));
 651         printk(KERN_NOTICE "inode=%p\n", inode);
 652         if (inode) {
 653                 printk(KERN_NOTICE "is_bad_inode(inode)=%d\n",
 654                        is_bad_inode(inode));
 655                 printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n",
 656                        NEXT_ORPHAN(inode));
 657                 printk(KERN_NOTICE "max_ino=%lu\n", max_ino);
 658                 printk(KERN_NOTICE "i_nlink=%u\n", inode->i_nlink);
 659                 /* Avoid freeing blocks if we got a bad deleted inode */
 660                 if (inode->i_nlink == 0)
 661                         inode->i_blocks = 0;
 662                 iput(inode);
 663         }
 664         brelse(bitmap_bh);
 665 error:
 666         return ERR_PTR(err);
 667 }
 668
 669 unsigned long ext3_count_free_inodes (struct super_block * sb)
 670 {
 671         unsigned long desc_count;
 672         struct ext3_group_desc *gdp;
 673         int i;
 674 #ifdef EXT3FS_DEBUG
 675         struct ext3_super_block *es;
 676         unsigned long bitmap_count, x;
 677         struct buffer_head *bitmap_bh = NULL;
 678
 679         es = EXT3_SB(sb)->s_es;
 680         desc_count = 0;
 681         bitmap_count = 0;
 682         gdp = NULL;
 683         for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
 684                 gdp = ext3_get_group_desc (sb, i, NULL);
 685                 if (!gdp)
 686                         continue;
 687                 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
 688                 brelse(bitmap_bh);
 689                 bitmap_bh = read_inode_bitmap(sb, i);
 690                 if (!bitmap_bh)
 691                         continue;
 692
 693                 x = ext3_count_free(bitmap_bh, EXT3_INODES_PER_GROUP(sb) / 8);
 694                 printk("group %d: stored = %d, counted = %lu\n",
 695                         i, le16_to_cpu(gdp->bg_free_inodes_count), x);
 696                 bitmap_count += x;
 697         }
 698         brelse(bitmap_bh);
 699         printk("ext3_count_free_inodes: stored = %u, computed = %lu, %lu\n",
 700                 le32_to_cpu(es->s_free_inodes_count), desc_count, bitmap_count);
 701         return desc_count;
 702 #else
 703         desc_count = 0;
 704         for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
 705                 gdp = ext3_get_group_desc (sb, i, NULL);
 706                 if (!gdp)
 707                         continue;
 708                 desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
 709                 cond_resched();
 710         }
 711         return desc_count;
 712 #endif
 713 }
 714
 715 /* Called at mount-time, super-block is locked */
 716 unsigned long ext3_count_dirs (struct super_block * sb)
 717 {
 718         unsigned long count = 0;
 719         int i;
 720
 721         for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) {
 722                 struct ext3_group_desc *gdp = ext3_get_group_desc (sb, i, NULL);
 723                 if (!gdp)
 724                         continue;
 725                 count += le16_to_cpu(gdp->bg_used_dirs_count);
 726         }
 727         return count;
 728 }
 729