at v2.6.13 1266 lines 37 kB view raw
1/* 2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README 3 */ 4/* Reiserfs block (de)allocator, bitmap-based. */ 5 6#include <linux/config.h> 7#include <linux/time.h> 8#include <linux/reiserfs_fs.h> 9#include <linux/errno.h> 10#include <linux/buffer_head.h> 11#include <linux/kernel.h> 12#include <linux/pagemap.h> 13#include <linux/reiserfs_fs_sb.h> 14#include <linux/reiserfs_fs_i.h> 15#include <linux/quotaops.h> 16 17#define PREALLOCATION_SIZE 9 18 19/* different reiserfs block allocator options */ 20 21#define SB_ALLOC_OPTS(s) (REISERFS_SB(s)->s_alloc_options.bits) 22 23#define _ALLOC_concentrating_formatted_nodes 0 24#define _ALLOC_displacing_large_files 1 25#define _ALLOC_displacing_new_packing_localities 2 26#define _ALLOC_old_hashed_relocation 3 27#define _ALLOC_new_hashed_relocation 4 28#define _ALLOC_skip_busy 5 29#define _ALLOC_displace_based_on_dirid 6 30#define _ALLOC_hashed_formatted_nodes 7 31#define _ALLOC_old_way 8 32#define _ALLOC_hundredth_slices 9 33#define _ALLOC_dirid_groups 10 34#define _ALLOC_oid_groups 11 35#define _ALLOC_packing_groups 12 36 37#define concentrating_formatted_nodes(s) test_bit(_ALLOC_concentrating_formatted_nodes, &SB_ALLOC_OPTS(s)) 38#define displacing_large_files(s) test_bit(_ALLOC_displacing_large_files, &SB_ALLOC_OPTS(s)) 39#define displacing_new_packing_localities(s) test_bit(_ALLOC_displacing_new_packing_localities, &SB_ALLOC_OPTS(s)) 40 41#define SET_OPTION(optname) \ 42 do { \ 43 reiserfs_warning(s, "reiserfs: option \"%s\" is set", #optname); \ 44 set_bit(_ALLOC_ ## optname , &SB_ALLOC_OPTS(s)); \ 45 } while(0) 46#define TEST_OPTION(optname, s) \ 47 test_bit(_ALLOC_ ## optname , &SB_ALLOC_OPTS(s)) 48 49static inline void get_bit_address(struct super_block *s, 50 b_blocknr_t block, int *bmap_nr, int *offset) 51{ 52 /* It is in the bitmap block number equal to the block 53 * number divided by the number of bits in a block. */ 54 *bmap_nr = block / (s->s_blocksize << 3); 55 /* Within that bitmap block it is located at bit offset *offset. */ 56 *offset = block & ((s->s_blocksize << 3) - 1); 57 return; 58} 59 60#ifdef CONFIG_REISERFS_CHECK 61int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value) 62{ 63 int i, j; 64 65 if (block == 0 || block >= SB_BLOCK_COUNT(s)) { 66 reiserfs_warning(s, 67 "vs-4010: is_reusable: block number is out of range %lu (%u)", 68 block, SB_BLOCK_COUNT(s)); 69 return 0; 70 } 71 72 /* it can't be one of the bitmap blocks */ 73 for (i = 0; i < SB_BMAP_NR(s); i++) 74 if (block == SB_AP_BITMAP(s)[i].bh->b_blocknr) { 75 reiserfs_warning(s, "vs: 4020: is_reusable: " 76 "bitmap block %lu(%u) can't be freed or reused", 77 block, SB_BMAP_NR(s)); 78 return 0; 79 } 80 81 get_bit_address(s, block, &i, &j); 82 83 if (i >= SB_BMAP_NR(s)) { 84 reiserfs_warning(s, 85 "vs-4030: is_reusable: there is no so many bitmap blocks: " 86 "block=%lu, bitmap_nr=%d", block, i); 87 return 0; 88 } 89 90 if ((bit_value == 0 && 91 reiserfs_test_le_bit(j, SB_AP_BITMAP(s)[i].bh->b_data)) || 92 (bit_value == 1 && 93 reiserfs_test_le_bit(j, SB_AP_BITMAP(s)[i].bh->b_data) == 0)) { 94 reiserfs_warning(s, 95 "vs-4040: is_reusable: corresponding bit of block %lu does not " 96 "match required value (i==%d, j==%d) test_bit==%d", 97 block, i, j, reiserfs_test_le_bit(j, 98 SB_AP_BITMAP 99 (s)[i].bh-> 100 b_data)); 101 102 return 0; 103 } 104 105 if (bit_value == 0 && block == SB_ROOT_BLOCK(s)) { 106 reiserfs_warning(s, 107 "vs-4050: is_reusable: this is root block (%u), " 108 "it must be busy", SB_ROOT_BLOCK(s)); 109 return 0; 110 } 111 112 return 1; 113} 114#endif /* CONFIG_REISERFS_CHECK */ 115 116/* searches in journal structures for a given block number (bmap, off). If block 117 is found in reiserfs journal it suggests next free block candidate to test. */ 118static inline int is_block_in_journal(struct super_block *s, int bmap, int 119 off, int *next) 120{ 121 b_blocknr_t tmp; 122 123 if (reiserfs_in_journal(s, bmap, off, 1, &tmp)) { 124 if (tmp) { /* hint supplied */ 125 *next = tmp; 126 PROC_INFO_INC(s, scan_bitmap.in_journal_hint); 127 } else { 128 (*next) = off + 1; /* inc offset to avoid looping. */ 129 PROC_INFO_INC(s, scan_bitmap.in_journal_nohint); 130 } 131 PROC_INFO_INC(s, scan_bitmap.retry); 132 return 1; 133 } 134 return 0; 135} 136 137/* it searches for a window of zero bits with given minimum and maximum lengths in one bitmap 138 * block; */ 139static int scan_bitmap_block(struct reiserfs_transaction_handle *th, 140 int bmap_n, int *beg, int boundary, int min, 141 int max, int unfm) 142{ 143 struct super_block *s = th->t_super; 144 struct reiserfs_bitmap_info *bi = &SB_AP_BITMAP(s)[bmap_n]; 145 int end, next; 146 int org = *beg; 147 148 BUG_ON(!th->t_trans_id); 149 150 RFALSE(bmap_n >= SB_BMAP_NR(s), "Bitmap %d is out of range (0..%d)", 151 bmap_n, SB_BMAP_NR(s) - 1); 152 PROC_INFO_INC(s, scan_bitmap.bmap); 153/* this is unclear and lacks comments, explain how journal bitmaps 154 work here for the reader. Convey a sense of the design here. What 155 is a window? */ 156/* - I mean `a window of zero bits' as in description of this function - Zam. */ 157 158 if (!bi) { 159 reiserfs_warning(s, "NULL bitmap info pointer for bitmap %d", 160 bmap_n); 161 return 0; 162 } 163 if (buffer_locked(bi->bh)) { 164 PROC_INFO_INC(s, scan_bitmap.wait); 165 __wait_on_buffer(bi->bh); 166 } 167 168 while (1) { 169 cont: 170 if (bi->free_count < min) 171 return 0; // No free blocks in this bitmap 172 173 /* search for a first zero bit -- beggining of a window */ 174 *beg = reiserfs_find_next_zero_le_bit 175 ((unsigned long *)(bi->bh->b_data), boundary, *beg); 176 177 if (*beg + min > boundary) { /* search for a zero bit fails or the rest of bitmap block 178 * cannot contain a zero window of minimum size */ 179 return 0; 180 } 181 182 if (unfm && is_block_in_journal(s, bmap_n, *beg, beg)) 183 continue; 184 /* first zero bit found; we check next bits */ 185 for (end = *beg + 1;; end++) { 186 if (end >= *beg + max || end >= boundary 187 || reiserfs_test_le_bit(end, bi->bh->b_data)) { 188 next = end; 189 break; 190 } 191 /* finding the other end of zero bit window requires looking into journal structures (in 192 * case of searching for free blocks for unformatted nodes) */ 193 if (unfm && is_block_in_journal(s, bmap_n, end, &next)) 194 break; 195 } 196 197 /* now (*beg) points to beginning of zero bits window, 198 * (end) points to one bit after the window end */ 199 if (end - *beg >= min) { /* it seems we have found window of proper size */ 200 int i; 201 reiserfs_prepare_for_journal(s, bi->bh, 1); 202 /* try to set all blocks used checking are they still free */ 203 for (i = *beg; i < end; i++) { 204 /* It seems that we should not check in journal again. */ 205 if (reiserfs_test_and_set_le_bit 206 (i, bi->bh->b_data)) { 207 /* bit was set by another process 208 * while we slept in prepare_for_journal() */ 209 PROC_INFO_INC(s, scan_bitmap.stolen); 210 if (i >= *beg + min) { /* we can continue with smaller set of allocated blocks, 211 * if length of this set is more or equal to `min' */ 212 end = i; 213 break; 214 } 215 /* otherwise we clear all bit were set ... */ 216 while (--i >= *beg) 217 reiserfs_test_and_clear_le_bit 218 (i, bi->bh->b_data); 219 reiserfs_restore_prepared_buffer(s, 220 bi-> 221 bh); 222 *beg = org; 223 /* ... and search again in current block from beginning */ 224 goto cont; 225 } 226 } 227 bi->free_count -= (end - *beg); 228 journal_mark_dirty(th, s, bi->bh); 229 230 /* free block count calculation */ 231 reiserfs_prepare_for_journal(s, SB_BUFFER_WITH_SB(s), 232 1); 233 PUT_SB_FREE_BLOCKS(s, SB_FREE_BLOCKS(s) - (end - *beg)); 234 journal_mark_dirty(th, s, SB_BUFFER_WITH_SB(s)); 235 236 return end - (*beg); 237 } else { 238 *beg = next; 239 } 240 } 241} 242 243static int bmap_hash_id(struct super_block *s, u32 id) 244{ 245 char *hash_in = NULL; 246 unsigned long hash; 247 unsigned bm; 248 249 if (id <= 2) { 250 bm = 1; 251 } else { 252 hash_in = (char *)(&id); 253 hash = keyed_hash(hash_in, 4); 254 bm = hash % SB_BMAP_NR(s); 255 if (!bm) 256 bm = 1; 257 } 258 /* this can only be true when SB_BMAP_NR = 1 */ 259 if (bm >= SB_BMAP_NR(s)) 260 bm = 0; 261 return bm; 262} 263 264/* 265 * hashes the id and then returns > 0 if the block group for the 266 * corresponding hash is full 267 */ 268static inline int block_group_used(struct super_block *s, u32 id) 269{ 270 int bm; 271 bm = bmap_hash_id(s, id); 272 if (SB_AP_BITMAP(s)[bm].free_count > ((s->s_blocksize << 3) * 60 / 100)) { 273 return 0; 274 } 275 return 1; 276} 277 278/* 279 * the packing is returned in disk byte order 280 */ 281__le32 reiserfs_choose_packing(struct inode * dir) 282{ 283 __le32 packing; 284 if (TEST_OPTION(packing_groups, dir->i_sb)) { 285 u32 parent_dir = le32_to_cpu(INODE_PKEY(dir)->k_dir_id); 286 /* 287 * some versions of reiserfsck expect packing locality 1 to be 288 * special 289 */ 290 if (parent_dir == 1 || block_group_used(dir->i_sb, parent_dir)) 291 packing = INODE_PKEY(dir)->k_objectid; 292 else 293 packing = INODE_PKEY(dir)->k_dir_id; 294 } else 295 packing = INODE_PKEY(dir)->k_objectid; 296 return packing; 297} 298 299/* Tries to find contiguous zero bit window (given size) in given region of 300 * bitmap and place new blocks there. Returns number of allocated blocks. */ 301static int scan_bitmap(struct reiserfs_transaction_handle *th, 302 b_blocknr_t * start, b_blocknr_t finish, 303 int min, int max, int unfm, unsigned long file_block) 304{ 305 int nr_allocated = 0; 306 struct super_block *s = th->t_super; 307 /* find every bm and bmap and bmap_nr in this file, and change them all to bitmap_blocknr 308 * - Hans, it is not a block number - Zam. */ 309 310 int bm, off; 311 int end_bm, end_off; 312 int off_max = s->s_blocksize << 3; 313 314 BUG_ON(!th->t_trans_id); 315 316 PROC_INFO_INC(s, scan_bitmap.call); 317 if (SB_FREE_BLOCKS(s) <= 0) 318 return 0; // No point in looking for more free blocks 319 320 get_bit_address(s, *start, &bm, &off); 321 get_bit_address(s, finish, &end_bm, &end_off); 322 if (bm > SB_BMAP_NR(s)) 323 return 0; 324 if (end_bm > SB_BMAP_NR(s)) 325 end_bm = SB_BMAP_NR(s); 326 327 /* When the bitmap is more than 10% free, anyone can allocate. 328 * When it's less than 10% free, only files that already use the 329 * bitmap are allowed. Once we pass 80% full, this restriction 330 * is lifted. 331 * 332 * We do this so that files that grow later still have space close to 333 * their original allocation. This improves locality, and presumably 334 * performance as a result. 335 * 336 * This is only an allocation policy and does not make up for getting a 337 * bad hint. Decent hinting must be implemented for this to work well. 338 */ 339 if (TEST_OPTION(skip_busy, s) 340 && SB_FREE_BLOCKS(s) > SB_BLOCK_COUNT(s) / 20) { 341 for (; bm < end_bm; bm++, off = 0) { 342 if ((off && (!unfm || (file_block != 0))) 343 || SB_AP_BITMAP(s)[bm].free_count > 344 (s->s_blocksize << 3) / 10) 345 nr_allocated = 346 scan_bitmap_block(th, bm, &off, off_max, 347 min, max, unfm); 348 if (nr_allocated) 349 goto ret; 350 } 351 /* we know from above that start is a reasonable number */ 352 get_bit_address(s, *start, &bm, &off); 353 } 354 355 for (; bm < end_bm; bm++, off = 0) { 356 nr_allocated = 357 scan_bitmap_block(th, bm, &off, off_max, min, max, unfm); 358 if (nr_allocated) 359 goto ret; 360 } 361 362 nr_allocated = 363 scan_bitmap_block(th, bm, &off, end_off + 1, min, max, unfm); 364 365 ret: 366 *start = bm * off_max + off; 367 return nr_allocated; 368 369} 370 371static void _reiserfs_free_block(struct reiserfs_transaction_handle *th, 372 struct inode *inode, b_blocknr_t block, 373 int for_unformatted) 374{ 375 struct super_block *s = th->t_super; 376 struct reiserfs_super_block *rs; 377 struct buffer_head *sbh; 378 struct reiserfs_bitmap_info *apbi; 379 int nr, offset; 380 381 BUG_ON(!th->t_trans_id); 382 383 PROC_INFO_INC(s, free_block); 384 385 rs = SB_DISK_SUPER_BLOCK(s); 386 sbh = SB_BUFFER_WITH_SB(s); 387 apbi = SB_AP_BITMAP(s); 388 389 get_bit_address(s, block, &nr, &offset); 390 391 if (nr >= sb_bmap_nr(rs)) { 392 reiserfs_warning(s, "vs-4075: reiserfs_free_block: " 393 "block %lu is out of range on %s", 394 block, reiserfs_bdevname(s)); 395 return; 396 } 397 398 reiserfs_prepare_for_journal(s, apbi[nr].bh, 1); 399 400 /* clear bit for the given block in bit map */ 401 if (!reiserfs_test_and_clear_le_bit(offset, apbi[nr].bh->b_data)) { 402 reiserfs_warning(s, "vs-4080: reiserfs_free_block: " 403 "free_block (%s:%lu)[dev:blocknr]: bit already cleared", 404 reiserfs_bdevname(s), block); 405 } 406 apbi[nr].free_count++; 407 journal_mark_dirty(th, s, apbi[nr].bh); 408 409 reiserfs_prepare_for_journal(s, sbh, 1); 410 /* update super block */ 411 set_sb_free_blocks(rs, sb_free_blocks(rs) + 1); 412 413 journal_mark_dirty(th, s, sbh); 414 if (for_unformatted) 415 DQUOT_FREE_BLOCK_NODIRTY(inode, 1); 416} 417 418void reiserfs_free_block(struct reiserfs_transaction_handle *th, 419 struct inode *inode, b_blocknr_t block, 420 int for_unformatted) 421{ 422 struct super_block *s = th->t_super; 423 424 BUG_ON(!th->t_trans_id); 425 426 RFALSE(!s, "vs-4061: trying to free block on nonexistent device"); 427 RFALSE(is_reusable(s, block, 1) == 0, 428 "vs-4071: can not free such block"); 429 /* mark it before we clear it, just in case */ 430 journal_mark_freed(th, s, block); 431 _reiserfs_free_block(th, inode, block, for_unformatted); 432} 433 434/* preallocated blocks don't need to be run through journal_mark_freed */ 435static void reiserfs_free_prealloc_block(struct reiserfs_transaction_handle *th, 436 struct inode *inode, b_blocknr_t block) 437{ 438 RFALSE(!th->t_super, 439 "vs-4060: trying to free block on nonexistent device"); 440 RFALSE(is_reusable(th->t_super, block, 1) == 0, 441 "vs-4070: can not free such block"); 442 BUG_ON(!th->t_trans_id); 443 _reiserfs_free_block(th, inode, block, 1); 444} 445 446static void __discard_prealloc(struct reiserfs_transaction_handle *th, 447 struct reiserfs_inode_info *ei) 448{ 449 unsigned long save = ei->i_prealloc_block; 450 int dirty = 0; 451 struct inode *inode = &ei->vfs_inode; 452 BUG_ON(!th->t_trans_id); 453#ifdef CONFIG_REISERFS_CHECK 454 if (ei->i_prealloc_count < 0) 455 reiserfs_warning(th->t_super, 456 "zam-4001:%s: inode has negative prealloc blocks count.", 457 __FUNCTION__); 458#endif 459 while (ei->i_prealloc_count > 0) { 460 reiserfs_free_prealloc_block(th, inode, ei->i_prealloc_block); 461 ei->i_prealloc_block++; 462 ei->i_prealloc_count--; 463 dirty = 1; 464 } 465 if (dirty) 466 reiserfs_update_sd(th, inode); 467 ei->i_prealloc_block = save; 468 list_del_init(&(ei->i_prealloc_list)); 469} 470 471/* FIXME: It should be inline function */ 472void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th, 473 struct inode *inode) 474{ 475 struct reiserfs_inode_info *ei = REISERFS_I(inode); 476 BUG_ON(!th->t_trans_id); 477 if (ei->i_prealloc_count) 478 __discard_prealloc(th, ei); 479} 480 481void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th) 482{ 483 struct list_head *plist = &SB_JOURNAL(th->t_super)->j_prealloc_list; 484 485 BUG_ON(!th->t_trans_id); 486 487 while (!list_empty(plist)) { 488 struct reiserfs_inode_info *ei; 489 ei = list_entry(plist->next, struct reiserfs_inode_info, 490 i_prealloc_list); 491#ifdef CONFIG_REISERFS_CHECK 492 if (!ei->i_prealloc_count) { 493 reiserfs_warning(th->t_super, 494 "zam-4001:%s: inode is in prealloc list but has no preallocated blocks.", 495 __FUNCTION__); 496 } 497#endif 498 __discard_prealloc(th, ei); 499 } 500} 501 502void reiserfs_init_alloc_options(struct super_block *s) 503{ 504 set_bit(_ALLOC_skip_busy, &SB_ALLOC_OPTS(s)); 505 set_bit(_ALLOC_dirid_groups, &SB_ALLOC_OPTS(s)); 506 set_bit(_ALLOC_packing_groups, &SB_ALLOC_OPTS(s)); 507} 508 509/* block allocator related options are parsed here */ 510int reiserfs_parse_alloc_options(struct super_block *s, char *options) 511{ 512 char *this_char, *value; 513 514 REISERFS_SB(s)->s_alloc_options.bits = 0; /* clear default settings */ 515 516 while ((this_char = strsep(&options, ":")) != NULL) { 517 if ((value = strchr(this_char, '=')) != NULL) 518 *value++ = 0; 519 520 if (!strcmp(this_char, "concentrating_formatted_nodes")) { 521 int temp; 522 SET_OPTION(concentrating_formatted_nodes); 523 temp = (value 524 && *value) ? simple_strtoul(value, &value, 525 0) : 10; 526 if (temp <= 0 || temp > 100) { 527 REISERFS_SB(s)->s_alloc_options.border = 10; 528 } else { 529 REISERFS_SB(s)->s_alloc_options.border = 530 100 / temp; 531 } 532 continue; 533 } 534 if (!strcmp(this_char, "displacing_large_files")) { 535 SET_OPTION(displacing_large_files); 536 REISERFS_SB(s)->s_alloc_options.large_file_size = 537 (value 538 && *value) ? simple_strtoul(value, &value, 0) : 16; 539 continue; 540 } 541 if (!strcmp(this_char, "displacing_new_packing_localities")) { 542 SET_OPTION(displacing_new_packing_localities); 543 continue; 544 }; 545 546 if (!strcmp(this_char, "old_hashed_relocation")) { 547 SET_OPTION(old_hashed_relocation); 548 continue; 549 } 550 551 if (!strcmp(this_char, "new_hashed_relocation")) { 552 SET_OPTION(new_hashed_relocation); 553 continue; 554 } 555 556 if (!strcmp(this_char, "dirid_groups")) { 557 SET_OPTION(dirid_groups); 558 continue; 559 } 560 if (!strcmp(this_char, "oid_groups")) { 561 SET_OPTION(oid_groups); 562 continue; 563 } 564 if (!strcmp(this_char, "packing_groups")) { 565 SET_OPTION(packing_groups); 566 continue; 567 } 568 if (!strcmp(this_char, "hashed_formatted_nodes")) { 569 SET_OPTION(hashed_formatted_nodes); 570 continue; 571 } 572 573 if (!strcmp(this_char, "skip_busy")) { 574 SET_OPTION(skip_busy); 575 continue; 576 } 577 578 if (!strcmp(this_char, "hundredth_slices")) { 579 SET_OPTION(hundredth_slices); 580 continue; 581 } 582 583 if (!strcmp(this_char, "old_way")) { 584 SET_OPTION(old_way); 585 continue; 586 } 587 588 if (!strcmp(this_char, "displace_based_on_dirid")) { 589 SET_OPTION(displace_based_on_dirid); 590 continue; 591 } 592 593 if (!strcmp(this_char, "preallocmin")) { 594 REISERFS_SB(s)->s_alloc_options.preallocmin = 595 (value 596 && *value) ? simple_strtoul(value, &value, 0) : 4; 597 continue; 598 } 599 600 if (!strcmp(this_char, "preallocsize")) { 601 REISERFS_SB(s)->s_alloc_options.preallocsize = 602 (value 603 && *value) ? simple_strtoul(value, &value, 604 0) : 605 PREALLOCATION_SIZE; 606 continue; 607 } 608 609 reiserfs_warning(s, "zam-4001: %s : unknown option - %s", 610 __FUNCTION__, this_char); 611 return 1; 612 } 613 614 reiserfs_warning(s, "allocator options = [%08x]\n", SB_ALLOC_OPTS(s)); 615 return 0; 616} 617 618static inline void new_hashed_relocation(reiserfs_blocknr_hint_t * hint) 619{ 620 char *hash_in; 621 if (hint->formatted_node) { 622 hash_in = (char *)&hint->key.k_dir_id; 623 } else { 624 if (!hint->inode) { 625 //hint->search_start = hint->beg; 626 hash_in = (char *)&hint->key.k_dir_id; 627 } else 628 if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super)) 629 hash_in = (char *)(&INODE_PKEY(hint->inode)->k_dir_id); 630 else 631 hash_in = 632 (char *)(&INODE_PKEY(hint->inode)->k_objectid); 633 } 634 635 hint->search_start = 636 hint->beg + keyed_hash(hash_in, 4) % (hint->end - hint->beg); 637} 638 639/* 640 * Relocation based on dirid, hashing them into a given bitmap block 641 * files. Formatted nodes are unaffected, a seperate policy covers them 642 */ 643static void dirid_groups(reiserfs_blocknr_hint_t * hint) 644{ 645 unsigned long hash; 646 __u32 dirid = 0; 647 int bm = 0; 648 struct super_block *sb = hint->th->t_super; 649 if (hint->inode) 650 dirid = le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id); 651 else if (hint->formatted_node) 652 dirid = hint->key.k_dir_id; 653 654 if (dirid) { 655 bm = bmap_hash_id(sb, dirid); 656 hash = bm * (sb->s_blocksize << 3); 657 /* give a portion of the block group to metadata */ 658 if (hint->inode) 659 hash += sb->s_blocksize / 2; 660 hint->search_start = hash; 661 } 662} 663 664/* 665 * Relocation based on oid, hashing them into a given bitmap block 666 * files. Formatted nodes are unaffected, a seperate policy covers them 667 */ 668static void oid_groups(reiserfs_blocknr_hint_t * hint) 669{ 670 if (hint->inode) { 671 unsigned long hash; 672 __u32 oid; 673 __u32 dirid; 674 int bm; 675 676 dirid = le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id); 677 678 /* keep the root dir and it's first set of subdirs close to 679 * the start of the disk 680 */ 681 if (dirid <= 2) 682 hash = (hint->inode->i_sb->s_blocksize << 3); 683 else { 684 oid = le32_to_cpu(INODE_PKEY(hint->inode)->k_objectid); 685 bm = bmap_hash_id(hint->inode->i_sb, oid); 686 hash = bm * (hint->inode->i_sb->s_blocksize << 3); 687 } 688 hint->search_start = hash; 689 } 690} 691 692/* returns 1 if it finds an indirect item and gets valid hint info 693 * from it, otherwise 0 694 */ 695static int get_left_neighbor(reiserfs_blocknr_hint_t * hint) 696{ 697 struct path *path; 698 struct buffer_head *bh; 699 struct item_head *ih; 700 int pos_in_item; 701 __le32 *item; 702 int ret = 0; 703 704 if (!hint->path) /* reiserfs code can call this function w/o pointer to path 705 * structure supplied; then we rely on supplied search_start */ 706 return 0; 707 708 path = hint->path; 709 bh = get_last_bh(path); 710 RFALSE(!bh, "green-4002: Illegal path specified to get_left_neighbor"); 711 ih = get_ih(path); 712 pos_in_item = path->pos_in_item; 713 item = get_item(path); 714 715 hint->search_start = bh->b_blocknr; 716 717 if (!hint->formatted_node && is_indirect_le_ih(ih)) { 718 /* for indirect item: go to left and look for the first non-hole entry 719 in the indirect item */ 720 if (pos_in_item == I_UNFM_NUM(ih)) 721 pos_in_item--; 722// pos_in_item = I_UNFM_NUM (ih) - 1; 723 while (pos_in_item >= 0) { 724 int t = get_block_num(item, pos_in_item); 725 if (t) { 726 hint->search_start = t; 727 ret = 1; 728 break; 729 } 730 pos_in_item--; 731 } 732 } 733 734 /* does result value fit into specified region? */ 735 return ret; 736} 737 738/* should be, if formatted node, then try to put on first part of the device 739 specified as number of percent with mount option device, else try to put 740 on last of device. This is not to say it is good code to do so, 741 but the effect should be measured. */ 742static inline void set_border_in_hint(struct super_block *s, 743 reiserfs_blocknr_hint_t * hint) 744{ 745 b_blocknr_t border = 746 SB_BLOCK_COUNT(s) / REISERFS_SB(s)->s_alloc_options.border; 747 748 if (hint->formatted_node) 749 hint->end = border - 1; 750 else 751 hint->beg = border; 752} 753 754static inline void displace_large_file(reiserfs_blocknr_hint_t * hint) 755{ 756 if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super)) 757 hint->search_start = 758 hint->beg + 759 keyed_hash((char *)(&INODE_PKEY(hint->inode)->k_dir_id), 760 4) % (hint->end - hint->beg); 761 else 762 hint->search_start = 763 hint->beg + 764 keyed_hash((char *)(&INODE_PKEY(hint->inode)->k_objectid), 765 4) % (hint->end - hint->beg); 766} 767 768static inline void hash_formatted_node(reiserfs_blocknr_hint_t * hint) 769{ 770 char *hash_in; 771 772 if (!hint->inode) 773 hash_in = (char *)&hint->key.k_dir_id; 774 else if (TEST_OPTION(displace_based_on_dirid, hint->th->t_super)) 775 hash_in = (char *)(&INODE_PKEY(hint->inode)->k_dir_id); 776 else 777 hash_in = (char *)(&INODE_PKEY(hint->inode)->k_objectid); 778 779 hint->search_start = 780 hint->beg + keyed_hash(hash_in, 4) % (hint->end - hint->beg); 781} 782 783static inline int 784this_blocknr_allocation_would_make_it_a_large_file(reiserfs_blocknr_hint_t * 785 hint) 786{ 787 return hint->block == 788 REISERFS_SB(hint->th->t_super)->s_alloc_options.large_file_size; 789} 790 791#ifdef DISPLACE_NEW_PACKING_LOCALITIES 792static inline void displace_new_packing_locality(reiserfs_blocknr_hint_t * hint) 793{ 794 struct in_core_key *key = &hint->key; 795 796 hint->th->displace_new_blocks = 0; 797 hint->search_start = 798 hint->beg + keyed_hash((char *)(&key->k_objectid), 799 4) % (hint->end - hint->beg); 800} 801#endif 802 803static inline int old_hashed_relocation(reiserfs_blocknr_hint_t * hint) 804{ 805 b_blocknr_t border; 806 u32 hash_in; 807 808 if (hint->formatted_node || hint->inode == NULL) { 809 return 0; 810 } 811 812 hash_in = le32_to_cpu((INODE_PKEY(hint->inode))->k_dir_id); 813 border = 814 hint->beg + (u32) keyed_hash(((char *)(&hash_in)), 815 4) % (hint->end - hint->beg - 1); 816 if (border > hint->search_start) 817 hint->search_start = border; 818 819 return 1; 820} 821 822static inline int old_way(reiserfs_blocknr_hint_t * hint) 823{ 824 b_blocknr_t border; 825 826 if (hint->formatted_node || hint->inode == NULL) { 827 return 0; 828 } 829 830 border = 831 hint->beg + 832 le32_to_cpu(INODE_PKEY(hint->inode)->k_dir_id) % (hint->end - 833 hint->beg); 834 if (border > hint->search_start) 835 hint->search_start = border; 836 837 return 1; 838} 839 840static inline void hundredth_slices(reiserfs_blocknr_hint_t * hint) 841{ 842 struct in_core_key *key = &hint->key; 843 b_blocknr_t slice_start; 844 845 slice_start = 846 (keyed_hash((char *)(&key->k_dir_id), 4) % 100) * (hint->end / 100); 847 if (slice_start > hint->search_start 848 || slice_start + (hint->end / 100) <= hint->search_start) { 849 hint->search_start = slice_start; 850 } 851} 852 853static void determine_search_start(reiserfs_blocknr_hint_t * hint, 854 int amount_needed) 855{ 856 struct super_block *s = hint->th->t_super; 857 int unfm_hint; 858 859 hint->beg = 0; 860 hint->end = SB_BLOCK_COUNT(s) - 1; 861 862 /* This is former border algorithm. Now with tunable border offset */ 863 if (concentrating_formatted_nodes(s)) 864 set_border_in_hint(s, hint); 865 866#ifdef DISPLACE_NEW_PACKING_LOCALITIES 867 /* whenever we create a new directory, we displace it. At first we will 868 hash for location, later we might look for a moderately empty place for 869 it */ 870 if (displacing_new_packing_localities(s) 871 && hint->th->displace_new_blocks) { 872 displace_new_packing_locality(hint); 873 874 /* we do not continue determine_search_start, 875 * if new packing locality is being displaced */ 876 return; 877 } 878#endif 879 880 /* all persons should feel encouraged to add more special cases here and 881 * test them */ 882 883 if (displacing_large_files(s) && !hint->formatted_node 884 && this_blocknr_allocation_would_make_it_a_large_file(hint)) { 885 displace_large_file(hint); 886 return; 887 } 888 889 /* if none of our special cases is relevant, use the left neighbor in the 890 tree order of the new node we are allocating for */ 891 if (hint->formatted_node && TEST_OPTION(hashed_formatted_nodes, s)) { 892 hash_formatted_node(hint); 893 return; 894 } 895 896 unfm_hint = get_left_neighbor(hint); 897 898 /* Mimic old block allocator behaviour, that is if VFS allowed for preallocation, 899 new blocks are displaced based on directory ID. Also, if suggested search_start 900 is less than last preallocated block, we start searching from it, assuming that 901 HDD dataflow is faster in forward direction */ 902 if (TEST_OPTION(old_way, s)) { 903 if (!hint->formatted_node) { 904 if (!reiserfs_hashed_relocation(s)) 905 old_way(hint); 906 else if (!reiserfs_no_unhashed_relocation(s)) 907 old_hashed_relocation(hint); 908 909 if (hint->inode 910 && hint->search_start < 911 REISERFS_I(hint->inode)->i_prealloc_block) 912 hint->search_start = 913 REISERFS_I(hint->inode)->i_prealloc_block; 914 } 915 return; 916 } 917 918 /* This is an approach proposed by Hans */ 919 if (TEST_OPTION(hundredth_slices, s) 920 && !(displacing_large_files(s) && !hint->formatted_node)) { 921 hundredth_slices(hint); 922 return; 923 } 924 925 /* old_hashed_relocation only works on unformatted */ 926 if (!unfm_hint && !hint->formatted_node && 927 TEST_OPTION(old_hashed_relocation, s)) { 928 old_hashed_relocation(hint); 929 } 930 /* new_hashed_relocation works with both formatted/unformatted nodes */ 931 if ((!unfm_hint || hint->formatted_node) && 932 TEST_OPTION(new_hashed_relocation, s)) { 933 new_hashed_relocation(hint); 934 } 935 /* dirid grouping works only on unformatted nodes */ 936 if (!unfm_hint && !hint->formatted_node && TEST_OPTION(dirid_groups, s)) { 937 dirid_groups(hint); 938 } 939#ifdef DISPLACE_NEW_PACKING_LOCALITIES 940 if (hint->formatted_node && TEST_OPTION(dirid_groups, s)) { 941 dirid_groups(hint); 942 } 943#endif 944 945 /* oid grouping works only on unformatted nodes */ 946 if (!unfm_hint && !hint->formatted_node && TEST_OPTION(oid_groups, s)) { 947 oid_groups(hint); 948 } 949 return; 950} 951 952static int determine_prealloc_size(reiserfs_blocknr_hint_t * hint) 953{ 954 /* make minimum size a mount option and benchmark both ways */ 955 /* we preallocate blocks only for regular files, specific size */ 956 /* benchmark preallocating always and see what happens */ 957 958 hint->prealloc_size = 0; 959 960 if (!hint->formatted_node && hint->preallocate) { 961 if (S_ISREG(hint->inode->i_mode) 962 && hint->inode->i_size >= 963 REISERFS_SB(hint->th->t_super)->s_alloc_options. 964 preallocmin * hint->inode->i_sb->s_blocksize) 965 hint->prealloc_size = 966 REISERFS_SB(hint->th->t_super)->s_alloc_options. 967 preallocsize - 1; 968 } 969 return CARRY_ON; 970} 971 972/* XXX I know it could be merged with upper-level function; 973 but may be result function would be too complex. */ 974static inline int allocate_without_wrapping_disk(reiserfs_blocknr_hint_t * hint, 975 b_blocknr_t * new_blocknrs, 976 b_blocknr_t start, 977 b_blocknr_t finish, int min, 978 int amount_needed, 979 int prealloc_size) 980{ 981 int rest = amount_needed; 982 int nr_allocated; 983 984 while (rest > 0 && start <= finish) { 985 nr_allocated = scan_bitmap(hint->th, &start, finish, min, 986 rest + prealloc_size, 987 !hint->formatted_node, hint->block); 988 989 if (nr_allocated == 0) /* no new blocks allocated, return */ 990 break; 991 992 /* fill free_blocknrs array first */ 993 while (rest > 0 && nr_allocated > 0) { 994 *new_blocknrs++ = start++; 995 rest--; 996 nr_allocated--; 997 } 998 999 /* do we have something to fill prealloc. array also ? */ 1000 if (nr_allocated > 0) { 1001 /* it means prealloc_size was greater that 0 and we do preallocation */ 1002 list_add(&REISERFS_I(hint->inode)->i_prealloc_list, 1003 &SB_JOURNAL(hint->th->t_super)-> 1004 j_prealloc_list); 1005 REISERFS_I(hint->inode)->i_prealloc_block = start; 1006 REISERFS_I(hint->inode)->i_prealloc_count = 1007 nr_allocated; 1008 break; 1009 } 1010 } 1011 1012 return (amount_needed - rest); 1013} 1014 1015static inline int blocknrs_and_prealloc_arrays_from_search_start 1016 (reiserfs_blocknr_hint_t * hint, b_blocknr_t * new_blocknrs, 1017 int amount_needed) { 1018 struct super_block *s = hint->th->t_super; 1019 b_blocknr_t start = hint->search_start; 1020 b_blocknr_t finish = SB_BLOCK_COUNT(s) - 1; 1021 int passno = 0; 1022 int nr_allocated = 0; 1023 int bigalloc = 0; 1024 1025 determine_prealloc_size(hint); 1026 if (!hint->formatted_node) { 1027 int quota_ret; 1028#ifdef REISERQUOTA_DEBUG 1029 reiserfs_debug(s, REISERFS_DEBUG_CODE, 1030 "reiserquota: allocating %d blocks id=%u", 1031 amount_needed, hint->inode->i_uid); 1032#endif 1033 quota_ret = 1034 DQUOT_ALLOC_BLOCK_NODIRTY(hint->inode, amount_needed); 1035 if (quota_ret) /* Quota exceeded? */ 1036 return QUOTA_EXCEEDED; 1037 if (hint->preallocate && hint->prealloc_size) { 1038#ifdef REISERQUOTA_DEBUG 1039 reiserfs_debug(s, REISERFS_DEBUG_CODE, 1040 "reiserquota: allocating (prealloc) %d blocks id=%u", 1041 hint->prealloc_size, hint->inode->i_uid); 1042#endif 1043 quota_ret = 1044 DQUOT_PREALLOC_BLOCK_NODIRTY(hint->inode, 1045 hint->prealloc_size); 1046 if (quota_ret) 1047 hint->preallocate = hint->prealloc_size = 0; 1048 } 1049 /* for unformatted nodes, force large allocations */ 1050 bigalloc = amount_needed; 1051 } 1052 1053 do { 1054 /* in bigalloc mode, nr_allocated should stay zero until 1055 * the entire allocation is filled 1056 */ 1057 if (unlikely(bigalloc && nr_allocated)) { 1058 reiserfs_warning(s, "bigalloc is %d, nr_allocated %d\n", 1059 bigalloc, nr_allocated); 1060 /* reset things to a sane value */ 1061 bigalloc = amount_needed - nr_allocated; 1062 } 1063 /* 1064 * try pass 0 and pass 1 looking for a nice big 1065 * contiguous allocation. Then reset and look 1066 * for anything you can find. 1067 */ 1068 if (passno == 2 && bigalloc) { 1069 passno = 0; 1070 bigalloc = 0; 1071 } 1072 switch (passno++) { 1073 case 0: /* Search from hint->search_start to end of disk */ 1074 start = hint->search_start; 1075 finish = SB_BLOCK_COUNT(s) - 1; 1076 break; 1077 case 1: /* Search from hint->beg to hint->search_start */ 1078 start = hint->beg; 1079 finish = hint->search_start; 1080 break; 1081 case 2: /* Last chance: Search from 0 to hint->beg */ 1082 start = 0; 1083 finish = hint->beg; 1084 break; 1085 default: /* We've tried searching everywhere, not enough space */ 1086 /* Free the blocks */ 1087 if (!hint->formatted_node) { 1088#ifdef REISERQUOTA_DEBUG 1089 reiserfs_debug(s, REISERFS_DEBUG_CODE, 1090 "reiserquota: freeing (nospace) %d blocks id=%u", 1091 amount_needed + 1092 hint->prealloc_size - 1093 nr_allocated, 1094 hint->inode->i_uid); 1095#endif 1096 DQUOT_FREE_BLOCK_NODIRTY(hint->inode, amount_needed + hint->prealloc_size - nr_allocated); /* Free not allocated blocks */ 1097 } 1098 while (nr_allocated--) 1099 reiserfs_free_block(hint->th, hint->inode, 1100 new_blocknrs[nr_allocated], 1101 !hint->formatted_node); 1102 1103 return NO_DISK_SPACE; 1104 } 1105 } while ((nr_allocated += allocate_without_wrapping_disk(hint, 1106 new_blocknrs + 1107 nr_allocated, 1108 start, finish, 1109 bigalloc ? 1110 bigalloc : 1, 1111 amount_needed - 1112 nr_allocated, 1113 hint-> 1114 prealloc_size)) 1115 < amount_needed); 1116 if (!hint->formatted_node && 1117 amount_needed + hint->prealloc_size > 1118 nr_allocated + REISERFS_I(hint->inode)->i_prealloc_count) { 1119 /* Some of preallocation blocks were not allocated */ 1120#ifdef REISERQUOTA_DEBUG 1121 reiserfs_debug(s, REISERFS_DEBUG_CODE, 1122 "reiserquota: freeing (failed prealloc) %d blocks id=%u", 1123 amount_needed + hint->prealloc_size - 1124 nr_allocated - 1125 REISERFS_I(hint->inode)->i_prealloc_count, 1126 hint->inode->i_uid); 1127#endif 1128 DQUOT_FREE_BLOCK_NODIRTY(hint->inode, amount_needed + 1129 hint->prealloc_size - nr_allocated - 1130 REISERFS_I(hint->inode)-> 1131 i_prealloc_count); 1132 } 1133 1134 return CARRY_ON; 1135} 1136 1137/* grab new blocknrs from preallocated list */ 1138/* return amount still needed after using them */ 1139static int use_preallocated_list_if_available(reiserfs_blocknr_hint_t * hint, 1140 b_blocknr_t * new_blocknrs, 1141 int amount_needed) 1142{ 1143 struct inode *inode = hint->inode; 1144 1145 if (REISERFS_I(inode)->i_prealloc_count > 0) { 1146 while (amount_needed) { 1147 1148 *new_blocknrs++ = REISERFS_I(inode)->i_prealloc_block++; 1149 REISERFS_I(inode)->i_prealloc_count--; 1150 1151 amount_needed--; 1152 1153 if (REISERFS_I(inode)->i_prealloc_count <= 0) { 1154 list_del(&REISERFS_I(inode)->i_prealloc_list); 1155 break; 1156 } 1157 } 1158 } 1159 /* return amount still needed after using preallocated blocks */ 1160 return amount_needed; 1161} 1162 1163int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t * hint, b_blocknr_t * new_blocknrs, int amount_needed, int reserved_by_us /* Amount of blocks we have 1164 already reserved */ ) 1165{ 1166 int initial_amount_needed = amount_needed; 1167 int ret; 1168 struct super_block *s = hint->th->t_super; 1169 1170 /* Check if there is enough space, taking into account reserved space */ 1171 if (SB_FREE_BLOCKS(s) - REISERFS_SB(s)->reserved_blocks < 1172 amount_needed - reserved_by_us) 1173 return NO_DISK_SPACE; 1174 /* should this be if !hint->inode && hint->preallocate? */ 1175 /* do you mean hint->formatted_node can be removed ? - Zam */ 1176 /* hint->formatted_node cannot be removed because we try to access 1177 inode information here, and there is often no inode assotiated with 1178 metadata allocations - green */ 1179 1180 if (!hint->formatted_node && hint->preallocate) { 1181 amount_needed = use_preallocated_list_if_available 1182 (hint, new_blocknrs, amount_needed); 1183 if (amount_needed == 0) /* all blocknrs we need we got from 1184 prealloc. list */ 1185 return CARRY_ON; 1186 new_blocknrs += (initial_amount_needed - amount_needed); 1187 } 1188 1189 /* find search start and save it in hint structure */ 1190 determine_search_start(hint, amount_needed); 1191 if (hint->search_start >= SB_BLOCK_COUNT(s)) 1192 hint->search_start = SB_BLOCK_COUNT(s) - 1; 1193 1194 /* allocation itself; fill new_blocknrs and preallocation arrays */ 1195 ret = blocknrs_and_prealloc_arrays_from_search_start 1196 (hint, new_blocknrs, amount_needed); 1197 1198 /* we used prealloc. list to fill (partially) new_blocknrs array. If final allocation fails we 1199 * need to return blocks back to prealloc. list or just free them. -- Zam (I chose second 1200 * variant) */ 1201 1202 if (ret != CARRY_ON) { 1203 while (amount_needed++ < initial_amount_needed) { 1204 reiserfs_free_block(hint->th, hint->inode, 1205 *(--new_blocknrs), 1); 1206 } 1207 } 1208 return ret; 1209} 1210 1211/* These 2 functions are here to provide blocks reservation to the rest of kernel */ 1212/* Reserve @blocks amount of blocks in fs pointed by @sb. Caller must make sure 1213 there are actually this much blocks on the FS available */ 1214void reiserfs_claim_blocks_to_be_allocated(struct super_block *sb, /* super block of 1215 filesystem where 1216 blocks should be 1217 reserved */ 1218 int blocks /* How much to reserve */ 1219 ) 1220{ 1221 1222 /* Fast case, if reservation is zero - exit immediately. */ 1223 if (!blocks) 1224 return; 1225 1226 spin_lock(&REISERFS_SB(sb)->bitmap_lock); 1227 REISERFS_SB(sb)->reserved_blocks += blocks; 1228 spin_unlock(&REISERFS_SB(sb)->bitmap_lock); 1229} 1230 1231/* Unreserve @blocks amount of blocks in fs pointed by @sb */ 1232void reiserfs_release_claimed_blocks(struct super_block *sb, /* super block of 1233 filesystem where 1234 blocks should be 1235 reserved */ 1236 int blocks /* How much to unreserve */ 1237 ) 1238{ 1239 1240 /* Fast case, if unreservation is zero - exit immediately. */ 1241 if (!blocks) 1242 return; 1243 1244 spin_lock(&REISERFS_SB(sb)->bitmap_lock); 1245 REISERFS_SB(sb)->reserved_blocks -= blocks; 1246 spin_unlock(&REISERFS_SB(sb)->bitmap_lock); 1247 RFALSE(REISERFS_SB(sb)->reserved_blocks < 0, 1248 "amount of blocks reserved became zero?"); 1249} 1250 1251/* This function estimates how much pages we will be able to write to FS 1252 used for reiserfs_file_write() purposes for now. */ 1253int reiserfs_can_fit_pages(struct super_block *sb /* superblock of filesystem 1254 to estimate space */ ) 1255{ 1256 int space; 1257 1258 spin_lock(&REISERFS_SB(sb)->bitmap_lock); 1259 space = 1260 (SB_FREE_BLOCKS(sb) - 1261 REISERFS_SB(sb)->reserved_blocks) >> (PAGE_CACHE_SHIFT - 1262 sb->s_blocksize_bits); 1263 spin_unlock(&REISERFS_SB(sb)->bitmap_lock); 1264 1265 return space > 0 ? space : 0; 1266}