fs/reiserfs/bitmap.c

   1 /*
   2  * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
   3  */
   4 /* Reiserfs block (de)allocator, bitmap-based. */
   5
   6 #include <linux/time.h>
   7 #include "reiserfs.h"
   8 #include <linux/errno.h>
   9 #include <linux/buffer_head.h>
  10 #include <linux/kernel.h>
  11 #include <linux/pagemap.h>
  12 #include <linux/vmalloc.h>
  13 #include <linux/quotaops.h>
  14 #include <linux/seq_file.h>
  15
  16 #define PREALLOCATION_SIZE 9
  17
  18 /* different reiserfs block allocator options */
  19
  20 #define SB_ALLOC_OPTS(s) (REISERFS_SB(s)->s_alloc_options.bits)
  21
  22 #define  _ALLOC_concentrating_formatted_nodes 0
  23 #define  _ALLOC_displacing_large_files 1
  24 #define  _ALLOC_displacing_new_packing_localities 2
  25 #define  _ALLOC_old_hashed_relocation 3
  26 #define  _ALLOC_new_hashed_relocation 4
  27 #define  _ALLOC_skip_busy 5
  28 #define  _ALLOC_displace_based_on_dirid 6
  29 #define  _ALLOC_hashed_formatted_nodes 7
  30 #define  _ALLOC_old_way 8
  31 #define  _ALLOC_hundredth_slices 9
  32 #define  _ALLOC_dirid_groups 10
  33 #define  _ALLOC_oid_groups 11
  34 #define  _ALLOC_packing_groups 12
  35
  36 #define  concentrating_formatted_nodes(s)       test_bit(_ALLOC_concentrating_formatted_nodes, &SB_ALLOC_OPTS(s))
  37 #define  displacing_large_files(s)              test_bit(_ALLOC_displacing_large_files, &SB_ALLOC_OPTS(s))
  38 #define  displacing_new_packing_localities(s)   test_bit(_ALLOC_displacing_new_packing_localities, &SB_ALLOC_OPTS(s))
  39
  40 #define SET_OPTION(optname) \
  41    do { \
  42         reiserfs_info(s, "block allocator option \"%s\" is set", #optname); \
  43         set_bit(_ALLOC_ ## optname , &SB_ALLOC_OPTS(s)); \
  44     } while(0)
  45 #define TEST_OPTION(optname, s) \
  46     test_bit(_ALLOC_ ## optname , &SB_ALLOC_OPTS(s))
  47
  48 static inline void get_bit_address(struct super_block *s,
  49                                    b_blocknr_t block,
  50                                    unsigned int *bmap_nr,
  51                                    unsigned int *offset)
  52 {
  53         /*
  54          * It is in the bitmap block number equal to the block
  55          * number divided by the number of bits in a block.
  56          */
  57         *bmap_nr = block >> (s->s_blocksize_bits + 3);
  58         /* Within that bitmap block it is located at bit offset *offset. */
  59         *offset = block & ((s->s_blocksize << 3) - 1);
  60 }
  61
  62 int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value)
  63 {
  64         unsigned int bmap, offset;
  65         unsigned int bmap_count = reiserfs_bmap_count(s);
  66
  67         if (block == 0 || block >= SB_BLOCK_COUNT(s)) {
  68                 reiserfs_error(s, "vs-4010",
  69                                "block number is out of range %lu (%u)",
  70                                block, SB_BLOCK_COUNT(s));
  71                 return 0;
  72         }
  73
  74         get_bit_address(s, block, &bmap, &offset);
  75
  76         /*
  77          * Old format filesystem? Unlikely, but the bitmaps are all
  78          * up front so we need to account for it.
  79          */
  80         if (unlikely(test_bit(REISERFS_OLD_FORMAT,
  81                               &REISERFS_SB(s)->s_properties))) {
  82                 b_blocknr_t bmap1 = REISERFS_SB(s)->s_sbh->b_blocknr + 1;
  83                 if (block >= bmap1 &&
  84                     block <= bmap1 + bmap_count) {
  85                         reiserfs_error(s, "vs-4019", "bitmap block %lu(%u) "
  86                                        "can't be freed or reused",
  87                                        block, bmap_count);
  88                         return 0;
  89                 }
  90         } else {
  91                 if (offset == 0) {
  92                         reiserfs_error(s, "vs-4020", "bitmap block %lu(%u) "
  93                                        "can't be freed or reused",
  94                                        block, bmap_count);
  95                         return 0;
  96                 }
  97         }
  98
  99         if (bmap >= bmap_count) {
 100                 reiserfs_error(s, "vs-4030", "bitmap for requested block "
 101                                "is out of range: block=%lu, bitmap_nr=%u",
 102                                block, bmap);
 103                 return 0;
 104         }
 105
 106         if (bit_value == 0 && block == SB_ROOT_BLOCK(s)) {
 107                 reiserfs_error(s, "vs-4050", "this is root block (%u), "
 108                                "it must be busy", SB_ROOT_BLOCK(s));
 109                 return 0;
 110         }
 111
 112         return 1;
 113 }
 114
 115 /*
 116  * Searches in journal structures for a given block number (bmap, off).
 117  * If block is found in reiserfs journal it suggests next free block
 118  * candidate to test.
 119  */
 120 static inline int is_block_in_journal(struct super_block *s, unsigned int bmap,
 121                                       int off, int *next)
 122 {
 123         b_blocknr_t tmp;
 124
 125         if (reiserfs_in_journal(s, bmap, off, 1, &tmp)) {
 126                 if (tmp) {      /* hint supplied */
 127                         *next = tmp;
 128                         PROC_INFO_INC(s, scan_bitmap.in_journal_hint);
 129                 } else {
 130                         (*next) = off + 1;  /* inc offset to avoid looping. */
 131                         PROC_INFO_INC(s, scan_bitmap.in_journal_nohint);
 132                 }
 133                 PROC_INFO_INC(s, scan_bitmap.retry);
 134                 return 1;
 135         }
 136         return 0;
 137 }
 138
 139 /*
 140  * Searches for a window of zero bits with given minimum and maximum
 141  * lengths in one bitmap block
 142  */
 143 static int scan_bitmap_block(struct reiserfs_transaction_handle *th,
 144                              unsigned int bmap_n, int *beg, int boundary,
 145                              int min, int max, int unfm)
 146 {
 147         struct super_block *s = th->t_super;
 148         struct reiserfs_bitmap_info *bi = &SB_AP_BITMAP(s)[bmap_n];
 149         struct buffer_head *bh;
 150         int end, next;
 151         int org = *beg;
 152
 153         BUG_ON(!th->t_trans_id);
 154         RFALSE(bmap_n >= reiserfs_bmap_count(s), "Bitmap %u is out of "
 155                "range (0..%u)", bmap_n, reiserfs_bmap_count(s) - 1);
 156         PROC_INFO_INC(s, scan_bitmap.bmap);
 157
 158         if (!bi) {
 159                 reiserfs_error(s, "jdm-4055", "NULL bitmap info pointer "
 160                                "for bitmap %d", bmap_n);
 161                 return 0;
 162         }
 163
 164         bh = reiserfs_read_bitmap_block(s, bmap_n);
 165         if (bh == NULL)
 166                 return 0;
 167
 168         while (1) {
 169 cont:
 170                 if (bi->free_count < min) {
 171                         brelse(bh);
 172                         return 0;       /* No free blocks in this bitmap */
 173                 }
 174
 175                 /* search for a first zero bit -- beginning of a window */
 176                 *beg = reiserfs_find_next_zero_le_bit
 177                     ((unsigned long *)(bh->b_data), boundary, *beg);
 178
 179                 /*
 180                  * search for a zero bit fails or the rest of bitmap block
 181                  * cannot contain a zero window of minimum size
 182                  */
 183                 if (*beg + min > boundary) {
 184                         brelse(bh);
 185                         return 0;
 186                 }
 187
 188                 if (unfm && is_block_in_journal(s, bmap_n, *beg, beg))
 189                         continue;
 190                 /* first zero bit found; we check next bits */
 191                 for (end = *beg + 1;; end++) {
 192                         if (end >= *beg + max || end >= boundary
 193                             || reiserfs_test_le_bit(end, bh->b_data)) {
 194                                 next = end;
 195                                 break;
 196                         }
 197
 198                         /*
 199                          * finding the other end of zero bit window requires
 200                          * looking into journal structures (in case of
 201                          * searching for free blocks for unformatted nodes)
 202                          */
 203                         if (unfm && is_block_in_journal(s, bmap_n, end, &next))
 204                                 break;
 205                 }
 206
 207                 /*
 208                  * now (*beg) points to beginning of zero bits window,
 209                  * (end) points to one bit after the window end
 210                  */
 211
 212                 /* found window of proper size */
 213                 if (end - *beg >= min) {
 214                         int i;
 215                         reiserfs_prepare_for_journal(s, bh, 1);
 216                         /*
 217                          * try to set all blocks used checking are
 218                          * they still free
 219                          */
 220                         for (i = *beg; i < end; i++) {
 221                                 /* Don't check in journal again. */
 222                                 if (reiserfs_test_and_set_le_bit
 223                                     (i, bh->b_data)) {
 224                                         /*
 225                                          * bit was set by another process while
 226                                          * we slept in prepare_for_journal()
 227                                          */
 228                                         PROC_INFO_INC(s, scan_bitmap.stolen);
 229
 230                                         /*
 231                                          * we can continue with smaller set
 232                                          * of allocated blocks, if length of
 233                                          * this set is more or equal to `min'
 234                                          */
 235                                         if (i >= *beg + min) {
 236                                                 end = i;
 237                                                 break;
 238                                         }
 239
 240                                         /*
 241                                          * otherwise we clear all bit
 242                                          * were set ...
 243                                          */
 244                                         while (--i >= *beg)
 245                                                 reiserfs_clear_le_bit
 246                                                     (i, bh->b_data);
 247                                         reiserfs_restore_prepared_buffer(s, bh);
 248                                         *beg = org;
 249
 250                                         /*
 251                                          * Search again in current block
 252                                          * from beginning
 253                                          */
 254                                         goto cont;
 255                                 }
 256                         }
 257                         bi->free_count -= (end - *beg);
 258                         journal_mark_dirty(th, bh);
 259                         brelse(bh);
 260
 261                         /* free block count calculation */
 262                         reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s),
 263                                                      1);
 264                         PUT_SB_FREE_BLOCKS(s, SB_FREE_BLOCKS(s) - (end - *beg));
 265                         journal_mark_dirty(th, SB_BUFFER_WITH_SB(s));
 266
 267                         return end - (*beg);
 268                 } else {
 269                         *beg = next;
 270                 }
 271         }
 272 }
 273
 274 static int bmap_hash_id(struct super_block *s, u32 id)
 275 {
 276         char *hash_in = NULL;
 277         unsigned long hash;
 278         unsigned bm;
 279
 280         if (id <= 2) {
 281                 bm = 1;
 282         } else {
 283                 hash_in = (char *)(&id);
 284                 hash = keyed_hash(hash_in, 4);
 285                 bm = hash % reiserfs_bmap_count(s);
 286                 if (!bm)
 287                         bm = 1;
 288         }
 289         /* this can only be true when SB_BMAP_NR = 1 */
 290         if (bm >= reiserfs_bmap_count(s))
 291                 bm = 0;
 292         return bm;
 293 }
 294
 295 /*
 296  * hashes the id and then returns > 0 if the block group for the
 297  * corresponding hash is full
 298  */
 299 static inline int block_group_used(struct super_block *s, u32 id)
 300 {
 301         int bm = bmap_hash_id(s, id);
 302         struct reiserfs_bitmap_info *info = &SB_AP_BITMAP(s)[bm];
 303
 304         /*
 305          * If we don't have cached information on this bitmap block, we're
 306          * going to have to load it later anyway. Loading it here allows us
 307          * to make a better decision. This favors long-term performance gain
 308          * with a better on-disk layout vs. a short term gain of skipping the
 309          * read and potentially having a bad placement.
 310          */
 311         if (info->free_count == UINT_MAX) {
 312                 struct buffer_head *bh = reiserfs_read_bitmap_block(s, bm);
 313                 brelse(bh);
 314         }
 315
 316         if (info->free_count > ((s->s_blocksize << 3) * 60 / 100)) {
 317                 return 0;
 318         }
 319         return 1;
 320 }
 321
 322 /*
 323  * the packing is returned in disk byte order
 324  */
 325 __le32 reiserfs_choose_packing(struct inode * dir)
 326 {
 327         __le32 packing;
 328         if (TEST_OPTION(packing_groups, dir->i_sb)) {
 329                 u32 parent_dir = le32_to_cpu(INODE_PKEY(dir)->k_dir_id);
 330                 /*
 331                  * some versions of reiserfsck expect packing locality 1 to be
 332                  * special
 333                  */
 334                 if (parent_dir == 1 || block_group_used(dir->i_sb, parent_dir))
 335                         packing = INODE_PKEY(dir)->k_objectid;
 336                 else
 337                         packing = INODE_PKEY(dir)->k_dir_id;
 338         } else
 339                 packing = INODE_PKEY(dir)->k_objectid;
 340         return packing;
 341 }
 342
 343 /*
 344  * Tries to find contiguous zero bit window (given size) in given region of
 345  * bitmap and place new blocks there. Returns number of allocated blocks.
 346  */
 347 static int scan_bitmap(struct reiserfs_transaction_handle *th,
 348                        b_blocknr_t * start, b_blocknr_t finish,
 349                        int min, int max, int unfm, sector_t file_block)
 350 {
 351         int nr_allocated = 0;
 352         struct super_block *s = th->t_super;
 353         unsigned int bm, off;
 354         unsigned int end_bm, end_off;
 355         unsigned int off_max = s->s_blocksize << 3;
 356
 357         BUG_ON(!th->t_trans_id);
 358         PROC_INFO_INC(s, scan_bitmap.call);
 359
 360         /* No point in looking for more free blocks */
 361         if (SB_FREE_BLOCKS(s) <= 0)
 362                 return 0;
 363
 364         get_bit_address(s, *start, &bm, &off);
 365         get_bit_address(s, finish, &end_bm, &end_off);
 366         if (bm > reiserfs_bmap_count(s))
 367                 return 0;
 368         if (end_bm > reiserfs_bmap_count(s))
 369                 end_bm = reiserfs_bmap_count(s);
 370
 371         /*
 372          * When the bitmap is more than 10% free, anyone can allocate.
 373          * When it's less than 10% free, only files that already use the
 374          * bitmap are allowed. Once we pass 80% full, this restriction
 375          * is lifted.
 376          *
 377          * We do this so that files that grow later still have space close to
 378          * their original allocation. This improves locality, and presumably
 379          * performance as a result.
 380          *
 381          * This is only an allocation policy and does not make up for getting a
 382          * bad hint. Decent hinting must be implemented for this to work well.
 383          */
 384         if (TEST_OPTION(skip_busy, s)
 385             && SB_FREE_BLOCKS(s) > SB_BLOCK_COUNT(s) / 20) {
 386                 for (; bm < end_bm; bm++, off = 0) {
 387                         if ((off && (!unfm || (file_block != 0)))
 388                             || SB_AP_BITMAP(s)[bm].free_count >
 389                             (s->s_blocksize << 3) / 10)
 390                                 nr_allocated =
 391                                     scan_bitmap_block(th, bm, &off, off_max,
 392                                                       min, max, unfm);
 393                         if (nr_allocated)
 394                                 goto ret;
 395                 }
 396                 /* we know from above that start is a reasonable number */
 397                 get_bit_address(s, *start, &bm, &off);
 398         }
 399
 400         for (; bm < end_bm; bm++, off = 0) {
 401                 nr_allocated =
 402                     scan_bitmap_block(th, bm, &off, off_max, min, max, unfm);
 403                 if (nr_allocated)
 404                         goto ret;
 405         }
 406
 407         nr_allocated =
 408             scan_bitmap_block(th, bm, &off, end_off + 1, min, max, unfm);
 409
 410 ret:
 411         *start = bm * off_max + off;
 412         return nr_allocated;
 413
 414 }
 415
 416 static void _reiserfs_free_block(struct reiserfs_transaction_handle *th,
 417                                  struct inode *inode, b_blocknr_t block,
 418                                  int for_unformatted)
 419 {
 420         struct super_block *s = th->t_super;
 421         struct reiserfs_super_block *rs;
 422         struct buffer_head *sbh, *bmbh;
 423         struct reiserfs_bitmap_info *apbi;
 424         unsigned int nr, offset;
 425
 426         BUG_ON(!th->t_trans_id);
 427         PROC_INFO_INC(s, free_block);
 428         rs = SB_DISK_SUPER_BLOCK(s);
 429         sbh = SB_BUFFER_WITH_SB(s);
 430         apbi = SB_AP_BITMAP(s);
 431
 432         get_bit_address(s, block, &nr, &offset);
 433
 434         if (nr >= reiserfs_bmap_count(s)) {
 435                 reiserfs_error(s, "vs-4075", "block %lu is out of range",
 436                                block);
 437                 return;
 438         }
 439
 440         bmbh = reiserfs_read_bitmap_block(s, nr);
 441         if (!bmbh)
 442                 return;
 443
 444         reiserfs_prepare_for_journal(s, bmbh, 1);
 445
 446         /* clear bit for the given block in bit map */
 447         if (!reiserfs_test_and_clear_le_bit(offset, bmbh->b_data)) {
 448                 reiserfs_error(s, "vs-4080",
 449                                "block %lu: bit already cleared", block);
 450         }
 451         apbi[nr].free_count++;
 452         journal_mark_dirty(th, bmbh);
 453         brelse(bmbh);
 454
 455         reiserfs_prepare_for_journal(s, sbh, 1);
 456         /* update super block */
 457         set_sb_free_blocks(rs, sb_free_blocks(rs) + 1);
 458
 459         journal_mark_dirty(th, sbh);
 460         if (for_unformatted) {
 461                 int depth = reiserfs_write_unlock_nested(s);
 462                 dquot_free_block_nodirty(inode, 1);
 463                 reiserfs_write_lock_nested(s, depth);
 464         }
 465 }
 466
 467 void reiserfs_free_block(struct reiserfs_transaction_handle *th,
 468                          struct inode *inode, b_blocknr_t block,
 469                          int for_unformatted)
 470 {
 471         struct super_block *s = th->t_super;
 472
 473         BUG_ON(!th->t_trans_id);
 474         RFALSE(!s, "vs-4061: trying to free block on nonexistent device");
 475         if (!is_reusable(s, block, 1))
 476                 return;
 477
 478         if (block > sb_block_count(REISERFS_SB(s)->s_rs)) {
 479                 reiserfs_error(th->t_super, "bitmap-4072",
 480                                "Trying to free block outside file system "
 481                                "boundaries (%lu > %lu)",
 482                                block, sb_block_count(REISERFS_SB(s)->s_rs));
 483                 return;
 484         }
 485         /* mark it before we clear it, just in case */
 486         journal_mark_freed(th, s, block);
 487         _reiserfs_free_block(th, inode, block, for_unformatted);
 488 }
 489
 490 /* preallocated blocks don't need to be run through journal_mark_freed */
 491 static void reiserfs_free_prealloc_block(struct reiserfs_transaction_handle *th,
 492                                          struct inode *inode, b_blocknr_t block)
 493 {
 494         BUG_ON(!th->t_trans_id);
 495         RFALSE(!th->t_super,
 496                "vs-4060: trying to free block on nonexistent device");
 497         if (!is_reusable(th->t_super, block, 1))
 498                 return;
 499         _reiserfs_free_block(th, inode, block, 1);
 500 }
 501
 502 static void __discard_prealloc(struct reiserfs_transaction_handle *th,
 503                                struct reiserfs_inode_info *ei)
 504 {
 505         unsigned long save = ei->i_prealloc_block;
 506         int dirty = 0;
 507         struct inode *inode = &ei->vfs_inode;
 508
 509         BUG_ON(!th->t_trans_id);
 510 #ifdef CONFIG_REISERFS_CHECK
 511         if (ei->i_prealloc_count < 0)
 512                 reiserfs_error(th->t_super, "zam-4001",
 513                                "inode has negative prealloc blocks count.");
 514 #endif
 515         while (ei->i_prealloc_count > 0) {
 516                 reiserfs_free_prealloc_block(th, inode, ei->i_prealloc_block);
 517                 ei->i_prealloc_block++;
 518                 ei->i_prealloc_count--;
 519                 dirty = 1;
 520         }
 521         if (dirty)
 522                 reiserfs_update_sd(th, inode);
 523         ei->i_prealloc_block = save;
 524         list_del_init(&ei->i_prealloc_list);
 525 }
 526
 527 /* FIXME: It should be inline function */
 528 void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,
 529                                struct inode *inode)
 530 {
 531         struct reiserfs_inode_info *ei = REISERFS_I(inode);
 532
 533         BUG_ON(!th->t_trans_id);
 534         if (ei->i_prealloc_count)
 535                 __discard_prealloc(th, ei);
 536 }
 537
 538 void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th)
 539 {
 540         struct list_head *plist = &SB_JOURNAL(th->t_super)->j_prealloc_list;
 541
 542         BUG_ON(!th->t_trans_id);
 543         while (!list_empty(plist)) {
 544                 struct reiserfs_inode_info *ei;
 545                 ei = list_entry(plist->next, struct reiserfs_inode_info,
 546                                 i_prealloc_list);
 547 #ifdef CONFIG_REISERFS_CHECK
 548                 if (!ei->i_prealloc_count) {
 549                         reiserfs_error(th->t_super, "zam-4001",
 550                                        "inode is in prealloc list but has "
 551                                        "no preallocated blocks.");
 552                 }
 553 #endif
 554                 __discard_prealloc(th, ei);
 555         }
 556 }
 557
 558 void reiserfs_init_alloc_options(struct super_block *s)
 559 {
 560         set_bit(_ALLOC_skip_busy, &SB_ALLOC_OPTS(s));
 561         set_bit(_ALLOC_dirid_groups, &SB_ALLOC_OPTS(s));
 562         set_bit(_ALLOC_packing_groups, &SB_ALLOC_OPTS(s));
 563 }
 564
 565 /* block allocator related options are parsed here */
 566 int reiserfs_parse_alloc_options(struct super_block *s, char *options)
 567 {
 568         char *this_char, *value;
 569
 570         /* clear default settings */
 571         REISERFS_SB(s)->s_alloc_options.bits = 0;
 572
 573         while ((this_char = strsep(&options, ":")) != NULL) {
 574                 if ((value = strchr(this_char, '=')) != NULL)
 575                         *value++ = 0;
 576
 577                 if (!strcmp(this_char, "concentrating_formatted_nodes")) {
 578                         int temp;
 579                         SET_OPTION(concentrating_formatted_nodes);
 580                         temp = (value
 581                                 && *value) ? simple_strtoul(value, &value,
 582                                                             0) : 10;
 583                         if (temp <= 0 || temp > 100) {
 584                                 REISERFS_SB(s)->s_alloc_options.border = 10;
 585                         } else {
 586                                 REISERFS_SB(s)->s_alloc_options.border =
 587                                     100 / temp;
 588                         }
 589                         continue;
 590                 }
 591                 if (!strcmp(this_char, "displacing_large_files")) {
 592                         SET_OPTION(displacing_large_files);
 593                         REISERFS_SB(s)->s_alloc_options.large_file_size =
 594                             (value
 595                              && *value) ? simple_strtoul(value, &value, 0) : 16;
 596                         continue;
 597                 }
 598                 if (!strcmp(this_char, "displacing_new_packing_localities")) {
 599                         SET_OPTION(displacing_new_packing_localities);
 600                         continue;
 601                 }
 602
 603                 if (!strcmp(this_char, "old_hashed_relocation")) {
 604                         SET_OPTION(old_hashed_relocation);
 605                         continue;
 606                 }
 607
 608                 if (!strcmp(this_char, "new_hashed_relocation")) {
 609                         SET_OPTION(new_hashed_relocation);
 610                         continue;
 611                 }
 612
 613                 if (!strcmp(this_char, "dirid_groups")) {
 614                         SET_OPTION(dirid_groups);
 615                         continue;
 616                 }
 617                 if (!strcmp(this_char, "oid_groups")) {
 618                         SET_OPTION(oid_groups);
 619                         continue;
 620                 }
 621                 if (!strcmp(this_char, "packing_groups")) {
 622                         SET_OPTION(packing_groups);
 623                         continue;
 624                 }
 625                 if (!strcmp(this_char, "hashed_formatted_nodes")) {
 626                         SET_OPTION(hashed_formatted_nodes);
 627                         continue;
 628                 }
 629
 630                 if (!strcmp(this_char, "skip_busy")) {
 631                         SET_OPTION(skip_busy);
 632                         continue;
 633                 }
 634
 635                 if (!strcmp(this_char, "hundredth_slices")) {
 636                         SET_OPTION(hundredth_slices);
 637                         continue;
 638                 }
 639
 640                 if (!strcmp(this_char, "old_way")) {
 641                         SET_OPTION(old_way);
 642                         continue;
 643                 }
 644
 645                 if (!strcmp(this_char, "displace_based_on_dirid")) {
 646                         SET_OPTION(displace_based_on_dirid);
 647                         continue;
 648                 }
 649
 650                 if (!strcmp(this_char, "preallocmin")) {
 651                         REISERFS_SB(s)->s_alloc_options.preallocmin =
 652                             (value
 653                              && *value) ? simple_strtoul(value, &value, 0) : 4;
 654                         continue;
 655                 }
 656
 657                 if (!strcmp(this_char, "preallocsize")) {
 658                         REISERFS_SB(s)->s_alloc_options.preallocsize =
 659                             (value
 660                              && *value) ? simple_strtoul(value, &value,
 661                                                          0) :
 662                             PREALLOCATION_SIZE;
 663                         continue;
 664                 }
 665
 666                 reiserfs_warning(s, "zam-4001", "unknown option - %s",
 667                                  this_char);
 668                 return 1;
 669         }
 670
 671         reiserfs_info(s, "allocator options = [%08x]\n", SB_ALLOC_OPTS(s));
 672         return 0;
 673 }
 674
 675 static void print_sep(struct seq_file *seq, int *first)
 676 {
 677         if (!*first)
 678                 seq_puts(seq, ":");
 679         else
 680                 *first = 0;
 681 }
 682
 683 void show_alloc_options(struct seq_file *seq, struct super_block *s)
 684 {
 685         int first = 1;
 686
 687         if (SB_ALLOC_OPTS(s) == ((1 << _ALLOC_skip_busy) |
 688                 (1 << _ALLOC_dirid_groups) | (1 << _ALLOC_packing_groups)))
 689                 return;
 690
 691         seq_puts(seq, ",alloc=");
 692
 693         if (TEST_OPTION(concentrating_formatted_nodes, s)) {
 694                 print_sep(seq, &first);
 695                 if (REISERFS_SB(s)->s_alloc_options.border != 10) {
 696                         seq_printf(seq, "concentrating_formatted_nodes=%d",
 697                                 100 / REISERFS_SB(s)->s_alloc_options.border);
 698                 } else
 699                         seq_puts(seq, "concentrating_formatted_nodes");
 700         }
 701         if (TEST_OPTION(displacing_large_files, s)) {
 702                 print_sep(seq, &first);
 703                 if (REISERFS_SB(s)->s_alloc_options.large_file_size != 16) {
 704                         seq_printf(seq, "displacing_large_files=%lu",
 705                             REISERFS_SB(s)->s_alloc_options.large_file_size);
 706                 } else
 707                         seq_puts(seq, "displacing_large_files");
 708         }
 709         if (TEST_OPTION(displacing_new_packing_localities, s)) {
 710                 print_sep(seq, &first);
 711                 seq_puts(seq, "displacing_new_packing_localities");
 712         }
 713         if (TEST_OPTION(old_hashed_relocation, s)) {
 714                 print_sep(seq, &first);
 715                 seq_puts(seq, "old_hashed_relocation");
 716         }
 717         if (TEST_OPTION(new_hashed_relocation, s)) {
 718                 print_sep(seq, &first);
 719                 seq_puts(seq, "new_hashed_relocation");
 720         }
 721         if (TEST_OPTION(dirid_groups, s)) {
 722                 print_sep(seq, &first);
 723                 seq_puts(seq, "dirid_groups");
 724         }
 725         if (TEST_OPTION(oid_groups, s)) {
 726                 print_sep(seq, &first);
 727                 seq_puts(seq, "oid_groups");
 728         }
 729         if (TEST_OPTION(packing_groups, s)) {
 730                 print_sep(seq, &first);
 731                 seq_puts(seq, "packing_groups");
 732         }
 733         if (TEST_OPTION(hashed_formatted_nodes, s)) {
 734                 print_sep(seq, &first);
 735                 seq_puts(seq, "hashed_formatted_nodes");
 736         }
 737         if (TEST_OPTION(skip_busy, s)) {
 738                 print_sep(seq, &first);
 739                 seq_puts(seq, "skip_busy");
 740         }
 741         if (TEST_OPTION(hundredth_slices, s)) {
 742                 print_sep(seq, &first);
 743                 seq_puts(seq, "hundredth_slices");
 744         }
 745         if (TEST_OPTION(old_way, s)) {
 746                 print_sep(seq, &first);
 747                 seq_puts(seq, "old_way");
 748         }
 749         if (TEST_OPTION(displace_based_on_dirid, s)) {
 750                 print_sep(seq, &first);
 751                 seq_puts(seq, "displace_based_on_dirid");
 752         }
 753         if (REISERFS_SB(s)->s_alloc_options.preallocmin != 0) {
 754                 print_sep(seq, &first);
 755                 seq_printf(seq, "preallocmin=%d",
 756                                 REISERFS_SB(s)->s_alloc_options.preallocmin);
 757         }
 758         if (REISERFS_SB(s)->s_alloc_options.preallocsize != 17) {
 759                 print_sep(seq, &first);
 760                 seq_printf(seq, "preallocsize=%d",
 761                                 REISERFS_SB(s)->s_alloc_options.preallocsize);
 762         }
 763 }
 764
 765 static inline void new_hashed_relocation(reiserfs_blocknr_hint_t * hint)
 766 {
 767         char *hash_in;
 768
 769         if (hint->formatted_node) {
 770                 hash_in = (char *)&hint->key.k_dir_id;
 771         } else {
 772                 if (!hint->inode) {
 773                         /*hint->search_start = hint->beg;*/
 774                         hash_in = (char *)&hint->key.k_dir_id;
 775                 } else
 776                     if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super))
 777                         hash_in = (char *)(&INODE_PKEY(hint->inode)->k_dir_id);
 778                 else
 779                         hash_in =
 780                             (char *)(&INODE_PKEY(hint->inode)->k_objectid);
 781         }
 782
 783         hint->search_start =
 784             hint->beg + keyed_hash(hash_in, 4) % (hint->end - hint->beg);
 785 }
 786
 787 /*
 788  * Relocation based on dirid, hashing them into a given bitmap block
 789  * files. Formatted nodes are unaffected, a separate policy covers them
 790  */
 791 static void dirid_groups(reiserfs_blocknr_hint_t * hint)
 792 {
 793         unsigned long hash;
 794         __u32 dirid = 0;
 795         int bm = 0;
 796         struct super_block *sb = hint->th->t_super;
 797
 798         if (hint->inode)
 799                 dirid = le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id);
 800         else if (hint->formatted_node)
 801                 dirid = hint->key.k_dir_id;
 802
 803         if (dirid) {
 804                 bm = bmap_hash_id(sb, dirid);
 805                 hash = bm * (sb->s_blocksize << 3);
 806                 /* give a portion of the block group to metadata */
 807                 if (hint->inode)
 808                         hash += sb->s_blocksize / 2;
 809                 hint->search_start = hash;
 810         }
 811 }
 812
 813 /*
 814  * Relocation based on oid, hashing them into a given bitmap block
 815  * files. Formatted nodes are unaffected, a separate policy covers them
 816  */
 817 static void oid_groups(reiserfs_blocknr_hint_t * hint)
 818 {
 819         if (hint->inode) {
 820                 unsigned long hash;
 821                 __u32 oid;
 822                 __u32 dirid;
 823                 int bm;
 824
 825                 dirid = le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id);
 826
 827                 /*
 828                  * keep the root dir and it's first set of subdirs close to
 829                  * the start of the disk
 830                  */
 831                 if (dirid <= 2)
 832                         hash = (hint->inode->i_sb->s_blocksize << 3);
 833                 else {
 834                         oid = le32_to_cpu(INODE_PKEY(hint->inode)->k_objectid);
 835                         bm = bmap_hash_id(hint->inode->i_sb, oid);
 836                         hash = bm * (hint->inode->i_sb->s_blocksize << 3);
 837                 }
 838                 hint->search_start = hash;
 839         }
 840 }
 841
 842 /*
 843  * returns 1 if it finds an indirect item and gets valid hint info
 844  * from it, otherwise 0
 845  */
 846 static int get_left_neighbor(reiserfs_blocknr_hint_t * hint)
 847 {
 848         struct treepath *path;
 849         struct buffer_head *bh;
 850         struct item_head *ih;
 851         int pos_in_item;
 852         __le32 *item;
 853         int ret = 0;
 854
 855         /*
 856          * reiserfs code can call this function w/o pointer to path
 857          * structure supplied; then we rely on supplied search_start
 858          */
 859         if (!hint->path)
 860                 return 0;
 861
 862         path = hint->path;
 863         bh = get_last_bh(path);
 864         RFALSE(!bh, "green-4002: Illegal path specified to get_left_neighbor");
 865         ih = tp_item_head(path);
 866         pos_in_item = path->pos_in_item;
 867         item = tp_item_body(path);
 868
 869         hint->search_start = bh->b_blocknr;
 870
 871         /*
 872          * for indirect item: go to left and look for the first non-hole entry
 873          * in the indirect item
 874          */
 875         if (!hint->formatted_node && is_indirect_le_ih(ih)) {
 876                 if (pos_in_item == I_UNFM_NUM(ih))
 877                         pos_in_item--;
 878                 while (pos_in_item >= 0) {
 879                         int t = get_block_num(item, pos_in_item);
 880                         if (t) {
 881                                 hint->search_start = t;
 882                                 ret = 1;
 883                                 break;
 884                         }
 885                         pos_in_item--;
 886                 }
 887         }
 888
 889         /* does result value fit into specified region? */
 890         return ret;
 891 }
 892
 893 /*
 894  * should be, if formatted node, then try to put on first part of the device
 895  * specified as number of percent with mount option device, else try to put
 896  * on last of device.  This is not to say it is good code to do so,
 897  * but the effect should be measured.
 898  */
 899 static inline void set_border_in_hint(struct super_block *s,
 900                                       reiserfs_blocknr_hint_t * hint)
 901 {
 902         b_blocknr_t border =
 903             SB_BLOCK_COUNT(s) / REISERFS_SB(s)->s_alloc_options.border;
 904
 905         if (hint->formatted_node)
 906                 hint->end = border - 1;
 907         else
 908                 hint->beg = border;
 909 }
 910
 911 static inline void displace_large_file(reiserfs_blocknr_hint_t * hint)
 912 {
 913         if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super))
 914                 hint->search_start =
 915                     hint->beg +
 916                     keyed_hash((char *)(&INODE_PKEY(hint->inode)->k_dir_id),
 917                                4) % (hint->end - hint->beg);
 918         else
 919                 hint->search_start =
 920                     hint->beg +
 921                     keyed_hash((char *)(&INODE_PKEY(hint->inode)->k_objectid),
 922                                4) % (hint->end - hint->beg);
 923 }
 924
 925 static inline void hash_formatted_node(reiserfs_blocknr_hint_t * hint)
 926 {
 927         char *hash_in;
 928
 929         if (!hint->inode)
 930                 hash_in = (char *)&hint->key.k_dir_id;
 931         else if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super))
 932                 hash_in = (char *)(&INODE_PKEY(hint->inode)->k_dir_id);
 933         else
 934                 hash_in = (char *)(&INODE_PKEY(hint->inode)->k_objectid);
 935
 936         hint->search_start =
 937             hint->beg + keyed_hash(hash_in, 4) % (hint->end - hint->beg);
 938 }
 939
 940 static inline int
 941 this_blocknr_allocation_would_make_it_a_large_file(reiserfs_blocknr_hint_t *
 942                                                    hint)
 943 {
 944         return hint->block ==
 945             REISERFS_SB(hint->th->t_super)->s_alloc_options.large_file_size;
 946 }
 947
 948 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
 949 static inline void displace_new_packing_locality(reiserfs_blocknr_hint_t * hint)
 950 {
 951         struct in_core_key *key = &hint->key;
 952
 953         hint->th->displace_new_blocks = 0;
 954         hint->search_start =
 955             hint->beg + keyed_hash((char *)(&key->k_objectid),
 956                                    4) % (hint->end - hint->beg);
 957 }
 958 #endif
 959
 960 static inline int old_hashed_relocation(reiserfs_blocknr_hint_t * hint)
 961 {
 962         b_blocknr_t border;
 963         u32 hash_in;
 964
 965         if (hint->formatted_node || hint->inode == NULL) {
 966                 return 0;
 967         }
 968
 969         hash_in = le32_to_cpu((INODE_PKEY(hint->inode))->k_dir_id);
 970         border =
 971             hint->beg + (u32) keyed_hash(((char *)(&hash_in)),
 972                                          4) % (hint->end - hint->beg - 1);
 973         if (border > hint->search_start)
 974                 hint->search_start = border;
 975
 976         return 1;
 977 }
 978
 979 static inline int old_way(reiserfs_blocknr_hint_t * hint)
 980 {
 981         b_blocknr_t border;
 982
 983         if (hint->formatted_node || hint->inode == NULL) {
 984                 return 0;
 985         }
 986
 987         border =
 988             hint->beg +
 989             le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id) % (hint->end -
 990                                                               hint->beg);
 991         if (border > hint->search_start)
 992                 hint->search_start = border;
 993
 994         return 1;
 995 }
 996
 997 static inline void hundredth_slices(reiserfs_blocknr_hint_t * hint)
 998 {
 999         struct in_core_key *key = &hint->key;
1000         b_blocknr_t slice_start;
1001
1002         slice_start =
1003             (keyed_hash((char *)(&key->k_dir_id), 4) % 100) * (hint->end / 100);
1004         if (slice_start > hint->search_start
1005             || slice_start + (hint->end / 100) <= hint->search_start) {
1006                 hint->search_start = slice_start;
1007         }
1008 }
1009
1010 static void determine_search_start(reiserfs_blocknr_hint_t * hint,
1011                                    int amount_needed)
1012 {
1013         struct super_block *s = hint->th->t_super;
1014         int unfm_hint;
1015
1016         hint->beg = 0;
1017         hint->end = SB_BLOCK_COUNT(s) - 1;
1018
1019         /* This is former border algorithm. Now with tunable border offset */
1020         if (concentrating_formatted_nodes(s))
1021                 set_border_in_hint(s, hint);
1022
1023 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1024         /*
1025          * whenever we create a new directory, we displace it.  At first
1026          * we will hash for location, later we might look for a moderately
1027          * empty place for it
1028          */
1029         if (displacing_new_packing_localities(s)
1030             && hint->th->displace_new_blocks) {
1031                 displace_new_packing_locality(hint);
1032
1033                 /*
1034                  * we do not continue determine_search_start,
1035                  * if new packing locality is being displaced
1036                  */
1037                 return;
1038         }
1039 #endif
1040
1041         /*
1042          * all persons should feel encouraged to add more special cases
1043          * here and test them
1044          */
1045
1046         if (displacing_large_files(s) && !hint->formatted_node
1047             && this_blocknr_allocation_would_make_it_a_large_file(hint)) {
1048                 displace_large_file(hint);
1049                 return;
1050         }
1051
1052         /*
1053          * if none of our special cases is relevant, use the left
1054          * neighbor in the tree order of the new node we are allocating for
1055          */
1056         if (hint->formatted_node && TEST_OPTION(hashed_formatted_nodes, s)) {
1057                 hash_formatted_node(hint);
1058                 return;
1059         }
1060
1061         unfm_hint = get_left_neighbor(hint);
1062
1063         /*
1064          * Mimic old block allocator behaviour, that is if VFS allowed for
1065          * preallocation, new blocks are displaced based on directory ID.
1066          * Also, if suggested search_start is less than last preallocated
1067          * block, we start searching from it, assuming that HDD dataflow
1068          * is faster in forward direction
1069          */
1070         if (TEST_OPTION(old_way, s)) {
1071                 if (!hint->formatted_node) {
1072                         if (!reiserfs_hashed_relocation(s))
1073                                 old_way(hint);
1074                         else if (!reiserfs_no_unhashed_relocation(s))
1075                                 old_hashed_relocation(hint);
1076
1077                         if (hint->inode
1078                             && hint->search_start <
1079                             REISERFS_I(hint->inode)->i_prealloc_block)
1080                                 hint->search_start =
1081                                     REISERFS_I(hint->inode)->i_prealloc_block;
1082                 }
1083                 return;
1084         }
1085
1086         /* This is an approach proposed by Hans */
1087         if (TEST_OPTION(hundredth_slices, s)
1088             && !(displacing_large_files(s) && !hint->formatted_node)) {
1089                 hundredth_slices(hint);
1090                 return;
1091         }
1092
1093         /* old_hashed_relocation only works on unformatted */
1094         if (!unfm_hint && !hint->formatted_node &&
1095             TEST_OPTION(old_hashed_relocation, s)) {
1096                 old_hashed_relocation(hint);
1097         }
1098
1099         /* new_hashed_relocation works with both formatted/unformatted nodes */
1100         if ((!unfm_hint || hint->formatted_node) &&
1101             TEST_OPTION(new_hashed_relocation, s)) {
1102                 new_hashed_relocation(hint);
1103         }
1104
1105         /* dirid grouping works only on unformatted nodes */
1106         if (!unfm_hint && !hint->formatted_node && TEST_OPTION(dirid_groups, s)) {
1107                 dirid_groups(hint);
1108         }
1109 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1110         if (hint->formatted_node && TEST_OPTION(dirid_groups, s)) {
1111                 dirid_groups(hint);
1112         }
1113 #endif
1114
1115         /* oid grouping works only on unformatted nodes */
1116         if (!unfm_hint && !hint->formatted_node && TEST_OPTION(oid_groups, s)) {
1117                 oid_groups(hint);
1118         }
1119         return;
1120 }
1121
1122 static int determine_prealloc_size(reiserfs_blocknr_hint_t * hint)
1123 {
1124         /* make minimum size a mount option and benchmark both ways */
1125         /* we preallocate blocks only for regular files, specific size */
1126         /* benchmark preallocating always and see what happens */
1127
1128         hint->prealloc_size = 0;
1129
1130         if (!hint->formatted_node && hint->preallocate) {
1131                 if (S_ISREG(hint->inode->i_mode)
1132                     && hint->inode->i_size >=
1133                     REISERFS_SB(hint->th->t_super)->s_alloc_options.
1134                     preallocmin * hint->inode->i_sb->s_blocksize)
1135                         hint->prealloc_size =
1136                             REISERFS_SB(hint->th->t_super)->s_alloc_options.
1137                             preallocsize - 1;
1138         }
1139         return CARRY_ON;
1140 }
1141
1142 static inline int allocate_without_wrapping_disk(reiserfs_blocknr_hint_t * hint,
1143                                                  b_blocknr_t * new_blocknrs,
1144                                                  b_blocknr_t start,
1145                                                  b_blocknr_t finish, int min,
1146                                                  int amount_needed,
1147                                                  int prealloc_size)
1148 {
1149         int rest = amount_needed;
1150         int nr_allocated;
1151
1152         while (rest > 0 && start <= finish) {
1153                 nr_allocated = scan_bitmap(hint->th, &start, finish, min,
1154                                            rest + prealloc_size,
1155                                            !hint->formatted_node, hint->block);
1156
1157                 if (nr_allocated == 0)  /* no new blocks allocated, return */
1158                         break;
1159
1160                 /* fill free_blocknrs array first */
1161                 while (rest > 0 && nr_allocated > 0) {
1162                         *new_blocknrs++ = start++;
1163                         rest--;
1164                         nr_allocated--;
1165                 }
1166
1167                 /* do we have something to fill prealloc. array also ? */
1168                 if (nr_allocated > 0) {
1169                         /*
1170                          * it means prealloc_size was greater that 0 and
1171                          * we do preallocation
1172                          */
1173                         list_add(&REISERFS_I(hint->inode)->i_prealloc_list,
1174                                  &SB_JOURNAL(hint->th->t_super)->
1175                                  j_prealloc_list);
1176                         REISERFS_I(hint->inode)->i_prealloc_block = start;
1177                         REISERFS_I(hint->inode)->i_prealloc_count =
1178                             nr_allocated;
1179                         break;
1180                 }
1181         }
1182
1183         return (amount_needed - rest);
1184 }
1185
1186 static inline int blocknrs_and_prealloc_arrays_from_search_start
1187     (reiserfs_blocknr_hint_t * hint, b_blocknr_t * new_blocknrs,
1188      int amount_needed) {
1189         struct super_block *s = hint->th->t_super;
1190         b_blocknr_t start = hint->search_start;
1191         b_blocknr_t finish = SB_BLOCK_COUNT(s) - 1;
1192         int passno = 0;
1193         int nr_allocated = 0;
1194         int depth;
1195
1196         determine_prealloc_size(hint);
1197         if (!hint->formatted_node) {
1198                 int quota_ret;
1199 #ifdef REISERQUOTA_DEBUG
1200                 reiserfs_debug(s, REISERFS_DEBUG_CODE,
1201                                "reiserquota: allocating %d blocks id=%u",
1202                                amount_needed, hint->inode->i_uid);
1203 #endif
1204                 depth = reiserfs_write_unlock_nested(s);
1205                 quota_ret =
1206                     dquot_alloc_block_nodirty(hint->inode, amount_needed);
1207                 if (quota_ret) {        /* Quota exceeded? */
1208                         reiserfs_write_lock_nested(s, depth);
1209                         return QUOTA_EXCEEDED;
1210                 }
1211                 if (hint->preallocate && hint->prealloc_size) {
1212 #ifdef REISERQUOTA_DEBUG
1213                         reiserfs_debug(s, REISERFS_DEBUG_CODE,
1214                                        "reiserquota: allocating (prealloc) %d blocks id=%u",
1215                                        hint->prealloc_size, hint->inode->i_uid);
1216 #endif
1217                         quota_ret = dquot_prealloc_block_nodirty(hint->inode,
1218                                                          hint->prealloc_size);
1219                         if (quota_ret)
1220                                 hint->preallocate = hint->prealloc_size = 0;
1221                 }
1222                 /* for unformatted nodes, force large allocations */
1223                 reiserfs_write_lock_nested(s, depth);
1224         }
1225
1226         do {
1227                 switch (passno++) {
1228                 case 0: /* Search from hint->search_start to end of disk */
1229                         start = hint->search_start;
1230                         finish = SB_BLOCK_COUNT(s) - 1;
1231                         break;
1232                 case 1: /* Search from hint->beg to hint->search_start */
1233                         start = hint->beg;
1234                         finish = hint->search_start;
1235                         break;
1236                 case 2: /* Last chance: Search from 0 to hint->beg */
1237                         start = 0;
1238                         finish = hint->beg;
1239                         break;
1240                 default:
1241                         /* We've tried searching everywhere, not enough space */
1242                         /* Free the blocks */
1243                         if (!hint->formatted_node) {
1244 #ifdef REISERQUOTA_DEBUG
1245                                 reiserfs_debug(s, REISERFS_DEBUG_CODE,
1246                                                "reiserquota: freeing (nospace) %d blocks id=%u",
1247                                                amount_needed +
1248                                                hint->prealloc_size -
1249                                                nr_allocated,
1250                                                hint->inode->i_uid);
1251 #endif
1252                                 /* Free not allocated blocks */
1253                                 depth = reiserfs_write_unlock_nested(s);
1254                                 dquot_free_block_nodirty(hint->inode,
1255                                         amount_needed + hint->prealloc_size -
1256                                         nr_allocated);
1257                                 reiserfs_write_lock_nested(s, depth);
1258                         }
1259                         while (nr_allocated--)
1260                                 reiserfs_free_block(hint->th, hint->inode,
1261                                                     new_blocknrs[nr_allocated],
1262                                                     !hint->formatted_node);
1263
1264                         return NO_DISK_SPACE;
1265                 }
1266         } while ((nr_allocated += allocate_without_wrapping_disk(hint,
1267                                                                  new_blocknrs +
1268                                                                  nr_allocated,
1269                                                                  start, finish,
1270                                                                  1,
1271                                                                  amount_needed -
1272                                                                  nr_allocated,
1273                                                                  hint->
1274                                                                  prealloc_size))
1275                  < amount_needed);
1276         if (!hint->formatted_node &&
1277             amount_needed + hint->prealloc_size >
1278             nr_allocated + REISERFS_I(hint->inode)->i_prealloc_count) {
1279                 /* Some of preallocation blocks were not allocated */
1280 #ifdef REISERQUOTA_DEBUG
1281                 reiserfs_debug(s, REISERFS_DEBUG_CODE,
1282                                "reiserquota: freeing (failed prealloc) %d blocks id=%u",
1283                                amount_needed + hint->prealloc_size -
1284                                nr_allocated -
1285                                REISERFS_I(hint->inode)->i_prealloc_count,
1286                                hint->inode->i_uid);
1287 #endif
1288
1289                 depth = reiserfs_write_unlock_nested(s);
1290                 dquot_free_block_nodirty(hint->inode, amount_needed +
1291                                          hint->prealloc_size - nr_allocated -
1292                                          REISERFS_I(hint->inode)->
1293                                          i_prealloc_count);
1294                 reiserfs_write_lock_nested(s, depth);
1295         }
1296
1297         return CARRY_ON;
1298 }
1299
1300 /* grab new blocknrs from preallocated list */
1301 /* return amount still needed after using them */
1302 static int use_preallocated_list_if_available(reiserfs_blocknr_hint_t * hint,
1303                                               b_blocknr_t * new_blocknrs,
1304                                               int amount_needed)
1305 {
1306         struct inode *inode = hint->inode;
1307
1308         if (REISERFS_I(inode)->i_prealloc_count > 0) {
1309                 while (amount_needed) {
1310
1311                         *new_blocknrs++ = REISERFS_I(inode)->i_prealloc_block++;
1312                         REISERFS_I(inode)->i_prealloc_count--;
1313
1314                         amount_needed--;
1315
1316                         if (REISERFS_I(inode)->i_prealloc_count <= 0) {
1317                                 list_del(&REISERFS_I(inode)->i_prealloc_list);
1318                                 break;
1319                         }
1320                 }
1321         }
1322         /* return amount still needed after using preallocated blocks */
1323         return amount_needed;
1324 }
1325
1326 int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *hint,
1327                                b_blocknr_t *new_blocknrs,
1328                                int amount_needed,
1329                                /* Amount of blocks we have already reserved */
1330                                int reserved_by_us)
1331 {
1332         int initial_amount_needed = amount_needed;
1333         int ret;
1334         struct super_block *s = hint->th->t_super;
1335
1336         /* Check if there is enough space, taking into account reserved space */
1337         if (SB_FREE_BLOCKS(s) - REISERFS_SB(s)->reserved_blocks <
1338             amount_needed - reserved_by_us)
1339                 return NO_DISK_SPACE;
1340         /* should this be if !hint->inode &&  hint->preallocate? */
1341         /* do you mean hint->formatted_node can be removed ? - Zam */
1342         /*
1343          * hint->formatted_node cannot be removed because we try to access
1344          * inode information here, and there is often no inode associated with
1345          * metadata allocations - green
1346          */
1347
1348         if (!hint->formatted_node && hint->preallocate) {
1349                 amount_needed = use_preallocated_list_if_available
1350                     (hint, new_blocknrs, amount_needed);
1351
1352                 /*
1353                  * We have all the block numbers we need from the
1354                  * prealloc list
1355                  */
1356                 if (amount_needed == 0)
1357                         return CARRY_ON;
1358                 new_blocknrs += (initial_amount_needed - amount_needed);
1359         }
1360
1361         /* find search start and save it in hint structure */
1362         determine_search_start(hint, amount_needed);
1363         if (hint->search_start >= SB_BLOCK_COUNT(s))
1364                 hint->search_start = SB_BLOCK_COUNT(s) - 1;
1365
1366         /* allocation itself; fill new_blocknrs and preallocation arrays */
1367         ret = blocknrs_and_prealloc_arrays_from_search_start
1368             (hint, new_blocknrs, amount_needed);
1369
1370         /*
1371          * We used prealloc. list to fill (partially) new_blocknrs array.
1372          * If final allocation fails we need to return blocks back to
1373          * prealloc. list or just free them. -- Zam (I chose second
1374          * variant)
1375          */
1376         if (ret != CARRY_ON) {
1377                 while (amount_needed++ < initial_amount_needed) {
1378                         reiserfs_free_block(hint->th, hint->inode,
1379                                             *(--new_blocknrs), 1);
1380                 }
1381         }
1382         return ret;
1383 }
1384
1385 void reiserfs_cache_bitmap_metadata(struct super_block *sb,
1386                                     struct buffer_head *bh,
1387                                     struct reiserfs_bitmap_info *info)
1388 {
1389         unsigned long *cur = (unsigned long *)(bh->b_data + bh->b_size);
1390
1391         /* The first bit must ALWAYS be 1 */
1392         if (!reiserfs_test_le_bit(0, (unsigned long *)bh->b_data))
1393                 reiserfs_error(sb, "reiserfs-2025", "bitmap block %lu is "
1394                                "corrupted: first bit must be 1", bh->b_blocknr);
1395
1396         info->free_count = 0;
1397
1398         while (--cur >= (unsigned long *)bh->b_data) {
1399                 /* 0 and ~0 are special, we can optimize for them */
1400                 if (*cur == 0)
1401                         info->free_count += BITS_PER_LONG;
1402                 else if (*cur != ~0L)   /* A mix, investigate */
1403                         info->free_count += BITS_PER_LONG - hweight_long(*cur);
1404         }
1405 }
1406
1407 struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb,
1408                                                unsigned int bitmap)
1409 {
1410         b_blocknr_t block = (sb->s_blocksize << 3) * bitmap;
1411         struct reiserfs_bitmap_info *info = SB_AP_BITMAP(sb) + bitmap;
1412         struct buffer_head *bh;
1413
1414         /*
1415          * Way old format filesystems had the bitmaps packed up front.
1416          * I doubt there are any of these left, but just in case...
1417          */
1418         if (unlikely(test_bit(REISERFS_OLD_FORMAT,
1419                               &REISERFS_SB(sb)->s_properties)))
1420                 block = REISERFS_SB(sb)->s_sbh->b_blocknr + 1 + bitmap;
1421         else if (bitmap == 0)
1422                 block = (REISERFS_DISK_OFFSET_IN_BYTES >> sb->s_blocksize_bits) + 1;
1423
1424         bh = sb_bread(sb, block);
1425         if (bh == NULL)
1426                 reiserfs_warning(sb, "sh-2029: %s: bitmap block (#%u) "
1427                                  "reading failed", __func__, block);
1428         else {
1429                 if (buffer_locked(bh)) {
1430                         int depth;
1431                         PROC_INFO_INC(sb, scan_bitmap.wait);
1432                         depth = reiserfs_write_unlock_nested(sb);
1433                         __wait_on_buffer(bh);
1434                         reiserfs_write_lock_nested(sb, depth);
1435                 }
1436                 BUG_ON(!buffer_uptodate(bh));
1437                 BUG_ON(atomic_read(&bh->b_count) == 0);
1438
1439                 if (info->free_count == UINT_MAX)
1440                         reiserfs_cache_bitmap_metadata(sb, bh, info);
1441         }
1442
1443         return bh;
1444 }
1445
1446 int reiserfs_init_bitmap_cache(struct super_block *sb)
1447 {
1448         struct reiserfs_bitmap_info *bitmap;
1449         unsigned int bmap_nr = reiserfs_bmap_count(sb);
1450
1451         bitmap = vmalloc(sizeof(*bitmap) * bmap_nr);
1452         if (bitmap == NULL)
1453                 return -ENOMEM;
1454
1455         memset(bitmap, 0xff, sizeof(*bitmap) * bmap_nr);
1456
1457         SB_AP_BITMAP(sb) = bitmap;
1458
1459         return 0;
1460 }
1461
1462 void reiserfs_free_bitmap_cache(struct super_block *sb)
1463 {
1464         if (SB_AP_BITMAP(sb)) {
1465                 vfree(SB_AP_BITMAP(sb));
1466                 SB_AP_BITMAP(sb) = NULL;
1467         }
1468 }