1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149
|
/* -*- mode: C++; c-basic-offset: 4; indent-tabs-mode: nil -*- */
// vim: ft=cpp:expandtab:ts=8:sw=4:softtabstop=4:
#ident "$Id$"
/*======
This file is part of PerconaFT.
Copyright (c) 2006, 2015, Percona and/or its affiliates. All rights reserved.
PerconaFT is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License, version 2,
as published by the Free Software Foundation.
PerconaFT is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with PerconaFT. If not, see <http://www.gnu.org/licenses/>.
----------------------------------------
PerconaFT is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License, version 3,
as published by the Free Software Foundation.
PerconaFT is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with PerconaFT. If not, see <http://www.gnu.org/licenses/>.
======= */
#ident "Copyright (c) 2006, 2015, Percona and/or its affiliates. All rights reserved."
#include "portability/memory.h"
#include "portability/toku_assert.h"
#include "portability/toku_portability.h"
#include "portability/toku_pthread.h"
// ugly but pragmatic, need access to dirty bits while holding translation lock
// TODO: Refactor this (possibly with FT-301)
#include "ft/ft-internal.h"
// TODO: reorganize this dependency (FT-303)
#include "ft/ft-ops.h" // for toku_maybe_truncate_file
#include "ft/serialize/block_table.h"
#include "ft/serialize/rbuf.h"
#include "ft/serialize/wbuf.h"
#include "ft/serialize/block_allocator.h"
#include "util/nb_mutex.h"
#include "util/scoped_malloc.h"
// indicates the end of a freelist
static const BLOCKNUM freelist_null = {-1};
// value of block_translation_pair.size if blocknum is unused
static const DISKOFF size_is_free = (DISKOFF)-1;
// value of block_translation_pair.u.diskoff if blocknum is used but does not
// yet have a diskblock
static const DISKOFF diskoff_unused = (DISKOFF)-2;
void block_table::_mutex_lock() { toku_mutex_lock(&_mutex); }
void block_table::_mutex_unlock() { toku_mutex_unlock(&_mutex); }
// TODO: Move lock to FT
void toku_ft_lock(FT ft) {
block_table *bt = &ft->blocktable;
bt->_mutex_lock();
}
// TODO: Move lock to FT
void toku_ft_unlock(FT ft) {
block_table *bt = &ft->blocktable;
toku_mutex_assert_locked(&bt->_mutex);
bt->_mutex_unlock();
}
// There are two headers: the reserve must fit them both and be suitably
// aligned.
static_assert(BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE %
BlockAllocator::BLOCK_ALLOCATOR_ALIGNMENT ==
0,
"Block allocator's header reserve must be suitibly aligned");
static_assert(
BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE * 2 ==
BlockAllocator::BLOCK_ALLOCATOR_TOTAL_HEADER_RESERVE,
"Block allocator's total header reserve must exactly fit two headers");
// does NOT initialize the block allocator: the caller is responsible
void block_table::_create_internal() {
memset(&_current, 0, sizeof(struct translation));
memset(&_inprogress, 0, sizeof(struct translation));
memset(&_checkpointed, 0, sizeof(struct translation));
memset(&_mutex, 0, sizeof(_mutex));
_bt_block_allocator = new BlockAllocator();
toku_mutex_init(&_mutex, nullptr);
nb_mutex_init(&_safe_file_size_lock);
}
// Fill in the checkpointed translation from buffer, and copy checkpointed to
// current.
// The one read from disk is the last known checkpointed one, so we are keeping
// it in
// place and then setting current (which is never stored on disk) for current
// use.
// The translation_buffer has translation only, we create the rest of the
// block_table.
int block_table::create_from_buffer(
int fd,
DISKOFF location_on_disk, // Location of translation_buffer
DISKOFF size_on_disk,
unsigned char *translation_buffer) {
// Does not initialize the block allocator
_create_internal();
// Deserialize the translation and copy it to current
int r = _translation_deserialize_from_buffer(
&_checkpointed, location_on_disk, size_on_disk, translation_buffer);
if (r != 0) {
return r;
}
_copy_translation(&_current, &_checkpointed, TRANSLATION_CURRENT);
// Determine the file size
int64_t file_size;
r = toku_os_get_file_size(fd, &file_size);
lazy_assert_zero(r);
invariant(file_size >= 0);
_safe_file_size = file_size;
// Gather the non-empty translations and use them to create the block
// allocator
toku::scoped_malloc pairs_buf(_checkpointed.smallest_never_used_blocknum.b *
sizeof(struct BlockAllocator::BlockPair));
struct BlockAllocator::BlockPair *CAST_FROM_VOIDP(pairs, pairs_buf.get());
uint64_t n_pairs = 0;
for (int64_t i = 0; i < _checkpointed.smallest_never_used_blocknum.b; i++) {
struct block_translation_pair pair = _checkpointed.block_translation[i];
if (pair.size > 0) {
invariant(pair.u.diskoff != diskoff_unused);
pairs[n_pairs++] =
BlockAllocator::BlockPair(pair.u.diskoff, pair.size);
}
}
_bt_block_allocator->CreateFromBlockPairs(
BlockAllocator::BLOCK_ALLOCATOR_TOTAL_HEADER_RESERVE,
BlockAllocator::BLOCK_ALLOCATOR_ALIGNMENT,
pairs,
n_pairs);
return 0;
}
void block_table::create() {
// Does not initialize the block allocator
_create_internal();
_checkpointed.type = TRANSLATION_CHECKPOINTED;
_checkpointed.smallest_never_used_blocknum =
make_blocknum(RESERVED_BLOCKNUMS);
_checkpointed.length_of_array =
_checkpointed.smallest_never_used_blocknum.b;
_checkpointed.blocknum_freelist_head = freelist_null;
XMALLOC_N(_checkpointed.length_of_array, _checkpointed.block_translation);
for (int64_t i = 0; i < _checkpointed.length_of_array; i++) {
_checkpointed.block_translation[i].size = 0;
_checkpointed.block_translation[i].u.diskoff = diskoff_unused;
}
// we just created a default checkpointed, now copy it to current.
_copy_translation(&_current, &_checkpointed, TRANSLATION_CURRENT);
// Create an empty block allocator.
_bt_block_allocator->Create(
BlockAllocator::BLOCK_ALLOCATOR_TOTAL_HEADER_RESERVE,
BlockAllocator::BLOCK_ALLOCATOR_ALIGNMENT);
}
// TODO: Refactor with FT-303
static void ft_set_dirty(FT ft, bool for_checkpoint) {
invariant(ft->h->type == FT_CURRENT);
if (for_checkpoint) {
invariant(ft->checkpoint_header->type == FT_CHECKPOINT_INPROGRESS);
ft->checkpoint_header->dirty = 1;
} else {
ft->h->dirty = 1;
}
}
void block_table::_maybe_truncate_file(int fd, uint64_t size_needed_before) {
toku_mutex_assert_locked(&_mutex);
uint64_t new_size_needed = _bt_block_allocator->AllocatedLimit();
// Save a call to toku_os_get_file_size (kernel call) if unlikely to be
// useful.
if (new_size_needed < size_needed_before &&
new_size_needed < _safe_file_size) {
nb_mutex_lock(&_safe_file_size_lock, &_mutex);
// Must hold _safe_file_size_lock to change _safe_file_size.
if (new_size_needed < _safe_file_size) {
int64_t safe_file_size_before = _safe_file_size;
// Not safe to use the 'to-be-truncated' portion until truncate is
// done.
_safe_file_size = new_size_needed;
_mutex_unlock();
uint64_t size_after;
toku_maybe_truncate_file(
fd, new_size_needed, safe_file_size_before, &size_after);
_mutex_lock();
_safe_file_size = size_after;
}
nb_mutex_unlock(&_safe_file_size_lock);
}
}
void block_table::maybe_truncate_file_on_open(int fd) {
_mutex_lock();
_maybe_truncate_file(fd, _safe_file_size);
_mutex_unlock();
}
void block_table::_copy_translation(struct translation *dst,
struct translation *src,
enum translation_type newtype) {
// We intend to malloc a fresh block, so the incoming translation should be
// empty
invariant_null(dst->block_translation);
invariant(src->length_of_array >= src->smallest_never_used_blocknum.b);
invariant(newtype == TRANSLATION_DEBUG ||
(src->type == TRANSLATION_CURRENT &&
newtype == TRANSLATION_INPROGRESS) ||
(src->type == TRANSLATION_CHECKPOINTED &&
newtype == TRANSLATION_CURRENT));
dst->type = newtype;
dst->smallest_never_used_blocknum = src->smallest_never_used_blocknum;
dst->blocknum_freelist_head = src->blocknum_freelist_head;
// destination btt is of fixed size. Allocate + memcpy the exact length
// necessary.
dst->length_of_array = dst->smallest_never_used_blocknum.b;
XMALLOC_N(dst->length_of_array, dst->block_translation);
memcpy(dst->block_translation,
src->block_translation,
dst->length_of_array * sizeof(*dst->block_translation));
// New version of btt is not yet stored on disk.
dst->block_translation[RESERVED_BLOCKNUM_TRANSLATION].size = 0;
dst->block_translation[RESERVED_BLOCKNUM_TRANSLATION].u.diskoff =
diskoff_unused;
}
int64_t block_table::get_blocks_in_use_unlocked() {
BLOCKNUM b;
struct translation *t = &_current;
int64_t num_blocks = 0;
{
// Reserved blocknums do not get upgraded; They are part of the header.
for (b.b = RESERVED_BLOCKNUMS; b.b < t->smallest_never_used_blocknum.b;
b.b++) {
if (t->block_translation[b.b].size != size_is_free) {
num_blocks++;
}
}
}
return num_blocks;
}
void block_table::_maybe_optimize_translation(struct translation *t) {
// Reduce 'smallest_never_used_blocknum.b' (completely free blocknums
// instead of just
// on a free list. Doing so requires us to regenerate the free list.
// This is O(n) work, so do it only if you're already doing that.
BLOCKNUM b;
paranoid_invariant(t->smallest_never_used_blocknum.b >= RESERVED_BLOCKNUMS);
// Calculate how large the free suffix is.
int64_t freed;
{
for (b.b = t->smallest_never_used_blocknum.b; b.b > RESERVED_BLOCKNUMS;
b.b--) {
if (t->block_translation[b.b - 1].size != size_is_free) {
break;
}
}
freed = t->smallest_never_used_blocknum.b - b.b;
}
if (freed > 0) {
t->smallest_never_used_blocknum.b = b.b;
if (t->length_of_array / 4 > t->smallest_never_used_blocknum.b) {
// We're using more memory than necessary to represent this now.
// Reduce.
uint64_t new_length = t->smallest_never_used_blocknum.b * 2;
XREALLOC_N(new_length, t->block_translation);
t->length_of_array = new_length;
// No need to zero anything out.
}
// Regenerate free list.
t->blocknum_freelist_head.b = freelist_null.b;
for (b.b = RESERVED_BLOCKNUMS; b.b < t->smallest_never_used_blocknum.b;
b.b++) {
if (t->block_translation[b.b].size == size_is_free) {
t->block_translation[b.b].u.next_free_blocknum =
t->blocknum_freelist_head;
t->blocknum_freelist_head = b;
}
}
}
}
// block table must be locked by caller of this function
void block_table::note_start_checkpoint_unlocked() {
toku_mutex_assert_locked(&_mutex);
// We're going to do O(n) work to copy the translation, so we
// can afford to do O(n) work by optimizing the translation
_maybe_optimize_translation(&_current);
// Copy current translation to inprogress translation.
_copy_translation(&_inprogress, &_current, TRANSLATION_INPROGRESS);
_checkpoint_skipped = false;
}
void block_table::note_skipped_checkpoint() {
// Purpose, alert block translation that the checkpoint was skipped, e.x.
// for a non-dirty header
_mutex_lock();
paranoid_invariant_notnull(_inprogress.block_translation);
_checkpoint_skipped = true;
_mutex_unlock();
}
// Purpose: free any disk space used by previous checkpoint that isn't in use by
// either
// - current state
// - in-progress checkpoint
// capture inprogress as new checkpointed.
// For each entry in checkpointBTT
// if offset does not match offset in inprogress
// assert offset does not match offset in current
// free (offset,len) from checkpoint
// move inprogress to checkpoint (resetting type)
// inprogress = NULL
void block_table::note_end_checkpoint(int fd) {
// Free unused blocks
_mutex_lock();
uint64_t allocated_limit_at_start = _bt_block_allocator->AllocatedLimit();
paranoid_invariant_notnull(_inprogress.block_translation);
if (_checkpoint_skipped) {
toku_free(_inprogress.block_translation);
memset(&_inprogress, 0, sizeof(_inprogress));
goto end;
}
// Make certain inprogress was allocated space on disk
invariant(
_inprogress.block_translation[RESERVED_BLOCKNUM_TRANSLATION].size > 0);
invariant(
_inprogress.block_translation[RESERVED_BLOCKNUM_TRANSLATION].u.diskoff >
0);
{
struct translation *t = &_checkpointed;
for (int64_t i = 0; i < t->length_of_array; i++) {
struct block_translation_pair *pair = &t->block_translation[i];
if (pair->size > 0 &&
!_translation_prevents_freeing(
&_inprogress, make_blocknum(i), pair)) {
invariant(!_translation_prevents_freeing(
&_current, make_blocknum(i), pair));
_bt_block_allocator->FreeBlock(pair->u.diskoff, pair->size);
}
}
toku_free(_checkpointed.block_translation);
_checkpointed = _inprogress;
_checkpointed.type = TRANSLATION_CHECKPOINTED;
memset(&_inprogress, 0, sizeof(_inprogress));
_maybe_truncate_file(fd, allocated_limit_at_start);
}
end:
_mutex_unlock();
}
bool block_table::_is_valid_blocknum(struct translation *t, BLOCKNUM b) {
invariant(t->length_of_array >= t->smallest_never_used_blocknum.b);
return b.b >= 0 && b.b < t->smallest_never_used_blocknum.b;
}
void block_table::_verify_valid_blocknum(struct translation *UU(t),
BLOCKNUM UU(b)) {
invariant(_is_valid_blocknum(t, b));
}
bool block_table::_is_valid_freeable_blocknum(struct translation *t,
BLOCKNUM b) {
invariant(t->length_of_array >= t->smallest_never_used_blocknum.b);
return b.b >= RESERVED_BLOCKNUMS && b.b < t->smallest_never_used_blocknum.b;
}
// should be freeable
void block_table::_verify_valid_freeable_blocknum(struct translation *UU(t),
BLOCKNUM UU(b)) {
invariant(_is_valid_freeable_blocknum(t, b));
}
// Also used only in ft-serialize-test.
void block_table::block_free(uint64_t offset, uint64_t size) {
_mutex_lock();
_bt_block_allocator->FreeBlock(offset, size);
_mutex_unlock();
}
int64_t block_table::_calculate_size_on_disk(struct translation *t) {
return 8 + // smallest_never_used_blocknum
8 + // blocknum_freelist_head
t->smallest_never_used_blocknum.b * 16 + // Array
4; // 4 for checksum
}
// We cannot free the disk space allocated to this blocknum if it is still in
// use by the given translation table.
bool block_table::_translation_prevents_freeing(
struct translation *t,
BLOCKNUM b,
struct block_translation_pair *old_pair) {
return t->block_translation && b.b < t->smallest_never_used_blocknum.b &&
old_pair->u.diskoff == t->block_translation[b.b].u.diskoff;
}
void block_table::_realloc_on_disk_internal(BLOCKNUM b,
DISKOFF size,
DISKOFF *offset,
FT ft,
bool for_checkpoint) {
toku_mutex_assert_locked(&_mutex);
ft_set_dirty(ft, for_checkpoint);
struct translation *t = &_current;
struct block_translation_pair old_pair = t->block_translation[b.b];
// Free the old block if it is not still in use by the checkpoint in
// progress or the previous checkpoint
bool cannot_free =
(!for_checkpoint &&
_translation_prevents_freeing(&_inprogress, b, &old_pair)) ||
_translation_prevents_freeing(&_checkpointed, b, &old_pair);
if (!cannot_free && old_pair.u.diskoff != diskoff_unused) {
_bt_block_allocator->FreeBlock(old_pair.u.diskoff, old_pair.size);
}
uint64_t allocator_offset = diskoff_unused;
t->block_translation[b.b].size = size;
if (size > 0) {
// Allocate a new block if the size is greater than 0,
// if the size is just 0, offset will be set to diskoff_unused
_bt_block_allocator->AllocBlock(size, &allocator_offset);
}
t->block_translation[b.b].u.diskoff = allocator_offset;
*offset = allocator_offset;
// Update inprogress btt if appropriate (if called because Pending bit is
// set).
if (for_checkpoint) {
paranoid_invariant(b.b < _inprogress.length_of_array);
_inprogress.block_translation[b.b] = t->block_translation[b.b];
}
}
void block_table::_ensure_safe_write_unlocked(int fd,
DISKOFF block_size,
DISKOFF block_offset) {
// Requires: holding _mutex
uint64_t size_needed = block_size + block_offset;
if (size_needed > _safe_file_size) {
// Must hold _safe_file_size_lock to change _safe_file_size.
nb_mutex_lock(&_safe_file_size_lock, &_mutex);
if (size_needed > _safe_file_size) {
_mutex_unlock();
int64_t size_after;
toku_maybe_preallocate_in_file(
fd, size_needed, _safe_file_size, &size_after);
_mutex_lock();
_safe_file_size = size_after;
}
nb_mutex_unlock(&_safe_file_size_lock);
}
}
void block_table::realloc_on_disk(BLOCKNUM b,
DISKOFF size,
DISKOFF *offset,
FT ft,
int fd,
bool for_checkpoint) {
_mutex_lock();
struct translation *t = &_current;
_verify_valid_freeable_blocknum(t, b);
_realloc_on_disk_internal(b, size, offset, ft, for_checkpoint);
_ensure_safe_write_unlocked(fd, size, *offset);
_mutex_unlock();
}
bool block_table::_pair_is_unallocated(struct block_translation_pair *pair) {
return pair->size == 0 && pair->u.diskoff == diskoff_unused;
}
// Effect: figure out where to put the inprogress btt on disk, allocate space
// for it there.
// The space must be 512-byte aligned (both the starting address and the
// size).
// As a result, the allcoated space may be a little bit bigger (up to the next
// 512-byte boundary) than the actual btt.
void block_table::_alloc_inprogress_translation_on_disk_unlocked() {
toku_mutex_assert_locked(&_mutex);
struct translation *t = &_inprogress;
paranoid_invariant_notnull(t->block_translation);
BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_TRANSLATION);
// Each inprogress is allocated only once
paranoid_invariant(_pair_is_unallocated(&t->block_translation[b.b]));
// Allocate a new block
int64_t size = _calculate_size_on_disk(t);
uint64_t offset;
_bt_block_allocator->AllocBlock(size, &offset);
t->block_translation[b.b].u.diskoff = offset;
t->block_translation[b.b].size = size;
}
// Effect: Serializes the blocktable to a wbuf (which starts uninitialized)
// A clean shutdown runs checkpoint start so that current and inprogress are
// copies.
// The resulting wbuf buffer is guaranteed to be be 512-byte aligned and the
// total length is a multiple of 512 (so we pad with zeros at the end if
// needd)
// The address is guaranteed to be 512-byte aligned, but the size is not
// guaranteed.
// It *is* guaranteed that we can read up to the next 512-byte boundary,
// however
void block_table::serialize_translation_to_wbuf(int fd,
struct wbuf *w,
int64_t *address,
int64_t *size) {
_mutex_lock();
struct translation *t = &_inprogress;
BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_TRANSLATION);
_alloc_inprogress_translation_on_disk_unlocked(); // The allocated block
// must be 512-byte
// aligned to make
// O_DIRECT happy.
uint64_t size_translation = _calculate_size_on_disk(t);
uint64_t size_aligned = roundup_to_multiple(512, size_translation);
invariant((int64_t)size_translation == t->block_translation[b.b].size);
{
// Init wbuf
if (0)
printf(
"%s:%d writing translation table of size_translation %" PRIu64
" at %" PRId64 "\n",
__FILE__,
__LINE__,
size_translation,
t->block_translation[b.b].u.diskoff);
char *XMALLOC_N_ALIGNED(512, size_aligned, buf);
for (uint64_t i = size_translation; i < size_aligned; i++)
buf[i] = 0; // fill in the end of the buffer with zeros.
wbuf_init(w, buf, size_aligned);
}
wbuf_BLOCKNUM(w, t->smallest_never_used_blocknum);
wbuf_BLOCKNUM(w, t->blocknum_freelist_head);
int64_t i;
for (i = 0; i < t->smallest_never_used_blocknum.b; i++) {
if (0)
printf("%s:%d %" PRId64 ",%" PRId64 "\n",
__FILE__,
__LINE__,
t->block_translation[i].u.diskoff,
t->block_translation[i].size);
wbuf_DISKOFF(w, t->block_translation[i].u.diskoff);
wbuf_DISKOFF(w, t->block_translation[i].size);
}
uint32_t checksum = toku_x1764_finish(&w->checksum);
wbuf_int(w, checksum);
*address = t->block_translation[b.b].u.diskoff;
*size = size_translation;
invariant((*address) % 512 == 0);
_ensure_safe_write_unlocked(fd, size_aligned, *address);
_mutex_unlock();
}
// Perhaps rename: purpose is get disk address of a block, given its blocknum
// (blockid?)
void block_table::_translate_blocknum_to_offset_size_unlocked(BLOCKNUM b,
DISKOFF *offset,
DISKOFF *size) {
struct translation *t = &_current;
_verify_valid_blocknum(t, b);
if (offset) {
*offset = t->block_translation[b.b].u.diskoff;
}
if (size) {
*size = t->block_translation[b.b].size;
}
}
// Perhaps rename: purpose is get disk address of a block, given its blocknum
// (blockid?)
void block_table::translate_blocknum_to_offset_size(BLOCKNUM b,
DISKOFF *offset,
DISKOFF *size) {
_mutex_lock();
_translate_blocknum_to_offset_size_unlocked(b, offset, size);
_mutex_unlock();
}
// Only called by toku_allocate_blocknum
// Effect: expand the array to maintain size invariant
// given that one more never-used blocknum will soon be used.
void block_table::_maybe_expand_translation(struct translation *t) {
if (t->length_of_array <= t->smallest_never_used_blocknum.b) {
// expansion is necessary
uint64_t new_length = t->smallest_never_used_blocknum.b * 2;
XREALLOC_N(new_length, t->block_translation);
uint64_t i;
for (i = t->length_of_array; i < new_length; i++) {
t->block_translation[i].u.next_free_blocknum = freelist_null;
t->block_translation[i].size = size_is_free;
}
t->length_of_array = new_length;
}
}
void block_table::_allocate_blocknum_unlocked(BLOCKNUM *res, FT ft) {
toku_mutex_assert_locked(&_mutex);
BLOCKNUM result;
struct translation *t = &_current;
if (t->blocknum_freelist_head.b == freelist_null.b) {
// no previously used blocknums are available
// use a never used blocknum
_maybe_expand_translation(
t); // Ensure a never used blocknums is available
result = t->smallest_never_used_blocknum;
t->smallest_never_used_blocknum.b++;
} else { // reuse a previously used blocknum
result = t->blocknum_freelist_head;
BLOCKNUM next = t->block_translation[result.b].u.next_free_blocknum;
t->blocknum_freelist_head = next;
}
// Verify the blocknum is free
paranoid_invariant(t->block_translation[result.b].size == size_is_free);
// blocknum is not free anymore
t->block_translation[result.b].u.diskoff = diskoff_unused;
t->block_translation[result.b].size = 0;
_verify_valid_freeable_blocknum(t, result);
*res = result;
ft_set_dirty(ft, false);
}
void block_table::allocate_blocknum(BLOCKNUM *res, FT ft) {
_mutex_lock();
_allocate_blocknum_unlocked(res, ft);
_mutex_unlock();
}
void block_table::_free_blocknum_in_translation(struct translation *t,
BLOCKNUM b) {
_verify_valid_freeable_blocknum(t, b);
paranoid_invariant(t->block_translation[b.b].size != size_is_free);
t->block_translation[b.b].size = size_is_free;
t->block_translation[b.b].u.next_free_blocknum = t->blocknum_freelist_head;
t->blocknum_freelist_head = b;
}
// Effect: Free a blocknum.
// If the blocknum holds the only reference to a block on disk, free that block
void block_table::_free_blocknum_unlocked(BLOCKNUM *bp,
FT ft,
bool for_checkpoint) {
toku_mutex_assert_locked(&_mutex);
BLOCKNUM b = *bp;
bp->b = 0; // Remove caller's reference.
struct block_translation_pair old_pair = _current.block_translation[b.b];
_free_blocknum_in_translation(&_current, b);
if (for_checkpoint) {
paranoid_invariant(ft->checkpoint_header->type ==
FT_CHECKPOINT_INPROGRESS);
_free_blocknum_in_translation(&_inprogress, b);
}
// If the size is 0, no disk block has ever been assigned to this blocknum.
if (old_pair.size > 0) {
// Free the old block if it is not still in use by the checkpoint in
// progress or the previous checkpoint
bool cannot_free =
_translation_prevents_freeing(&_inprogress, b, &old_pair) ||
_translation_prevents_freeing(&_checkpointed, b, &old_pair);
if (!cannot_free) {
_bt_block_allocator->FreeBlock(old_pair.u.diskoff, old_pair.size);
}
} else {
paranoid_invariant(old_pair.size == 0);
paranoid_invariant(old_pair.u.diskoff == diskoff_unused);
}
ft_set_dirty(ft, for_checkpoint);
}
void block_table::free_blocknum(BLOCKNUM *bp, FT ft, bool for_checkpoint) {
_mutex_lock();
_free_blocknum_unlocked(bp, ft, for_checkpoint);
_mutex_unlock();
}
// Verify there are no free blocks.
void block_table::verify_no_free_blocknums() {
invariant(_current.blocknum_freelist_head.b == freelist_null.b);
}
// Frees blocknums that have a size of 0 and unused diskoff
// Currently used for eliminating unused cached rollback log nodes
void block_table::free_unused_blocknums(BLOCKNUM root) {
_mutex_lock();
int64_t smallest = _current.smallest_never_used_blocknum.b;
for (int64_t i = RESERVED_BLOCKNUMS; i < smallest; i++) {
if (i == root.b) {
continue;
}
BLOCKNUM b = make_blocknum(i);
if (_current.block_translation[b.b].size == 0) {
invariant(_current.block_translation[b.b].u.diskoff ==
diskoff_unused);
_free_blocknum_in_translation(&_current, b);
}
}
_mutex_unlock();
}
bool block_table::_no_data_blocks_except_root(BLOCKNUM root) {
bool ok = true;
_mutex_lock();
int64_t smallest = _current.smallest_never_used_blocknum.b;
if (root.b < RESERVED_BLOCKNUMS) {
ok = false;
goto cleanup;
}
for (int64_t i = RESERVED_BLOCKNUMS; i < smallest; i++) {
if (i == root.b) {
continue;
}
BLOCKNUM b = make_blocknum(i);
if (_current.block_translation[b.b].size != size_is_free) {
ok = false;
goto cleanup;
}
}
cleanup:
_mutex_unlock();
return ok;
}
// Verify there are no data blocks except root.
// TODO(leif): This actually takes a lock, but I don't want to fix all the
// callers right now.
void block_table::verify_no_data_blocks_except_root(BLOCKNUM UU(root)) {
paranoid_invariant(_no_data_blocks_except_root(root));
}
bool block_table::_blocknum_allocated(BLOCKNUM b) {
_mutex_lock();
struct translation *t = &_current;
_verify_valid_blocknum(t, b);
bool ok = t->block_translation[b.b].size != size_is_free;
_mutex_unlock();
return ok;
}
// Verify a blocknum is currently allocated.
void block_table::verify_blocknum_allocated(BLOCKNUM UU(b)) {
paranoid_invariant(_blocknum_allocated(b));
}
// Only used by toku_dump_translation table (debug info)
void block_table::_dump_translation_internal(FILE *f, struct translation *t) {
if (t->block_translation) {
BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_TRANSLATION);
fprintf(f, " length_of_array[%" PRId64 "]", t->length_of_array);
fprintf(f,
" smallest_never_used_blocknum[%" PRId64 "]",
t->smallest_never_used_blocknum.b);
fprintf(f,
" blocknum_free_list_head[%" PRId64 "]",
t->blocknum_freelist_head.b);
fprintf(
f, " size_on_disk[%" PRId64 "]", t->block_translation[b.b].size);
fprintf(f,
" location_on_disk[%" PRId64 "]\n",
t->block_translation[b.b].u.diskoff);
int64_t i;
for (i = 0; i < t->length_of_array; i++) {
fprintf(f,
" %" PRId64 ": %" PRId64 " %" PRId64 "\n",
i,
t->block_translation[i].u.diskoff,
t->block_translation[i].size);
}
fprintf(f, "\n");
} else {
fprintf(f, " does not exist\n");
}
}
// Only used by toku_ft_dump which is only for debugging purposes
// "pretty" just means we use tabs so we can parse output easier later
void block_table::dump_translation_table_pretty(FILE *f) {
_mutex_lock();
struct translation *t = &_checkpointed;
invariant(t->block_translation != nullptr);
for (int64_t i = 0; i < t->length_of_array; ++i) {
fprintf(f,
"%" PRId64 "\t%" PRId64 "\t%" PRId64 "\n",
i,
t->block_translation[i].u.diskoff,
t->block_translation[i].size);
}
_mutex_unlock();
}
// Only used by toku_ft_dump which is only for debugging purposes
void block_table::dump_translation_table(FILE *f) {
_mutex_lock();
fprintf(f, "Current block translation:");
_dump_translation_internal(f, &_current);
fprintf(f, "Checkpoint in progress block translation:");
_dump_translation_internal(f, &_inprogress);
fprintf(f, "Checkpointed block translation:");
_dump_translation_internal(f, &_checkpointed);
_mutex_unlock();
}
// Only used by ftdump
void block_table::blocknum_dump_translation(BLOCKNUM b) {
_mutex_lock();
struct translation *t = &_current;
if (b.b < t->length_of_array) {
struct block_translation_pair *bx = &t->block_translation[b.b];
printf("%" PRId64 ": %" PRId64 " %" PRId64 "\n",
b.b,
bx->u.diskoff,
bx->size);
}
_mutex_unlock();
}
// Must not call this function when anything else is using the blocktable.
// No one may use the blocktable afterwards.
void block_table::destroy(void) {
// TODO: translation.destroy();
toku_free(_current.block_translation);
toku_free(_inprogress.block_translation);
toku_free(_checkpointed.block_translation);
_bt_block_allocator->Destroy();
delete _bt_block_allocator;
toku_mutex_destroy(&_mutex);
nb_mutex_destroy(&_safe_file_size_lock);
}
int block_table::_translation_deserialize_from_buffer(
struct translation *t,
DISKOFF location_on_disk,
uint64_t size_on_disk,
// out: buffer with serialized translation
unsigned char *translation_buffer) {
int r = 0;
invariant(location_on_disk != 0);
t->type = TRANSLATION_CHECKPOINTED;
// check the checksum
uint32_t x1764 = toku_x1764_memory(translation_buffer, size_on_disk - 4);
uint64_t offset = size_on_disk - 4;
uint32_t stored_x1764 = toku_dtoh32(*(int *)(translation_buffer + offset));
if (x1764 != stored_x1764) {
fprintf(stderr,
"Translation table checksum failure: calc=0x%08x read=0x%08x\n",
x1764,
stored_x1764);
r = TOKUDB_BAD_CHECKSUM;
goto exit;
}
struct rbuf rb;
rb.buf = translation_buffer;
rb.ndone = 0;
rb.size = size_on_disk - 4; // 4==checksum
t->smallest_never_used_blocknum = rbuf_blocknum(&rb);
t->length_of_array = t->smallest_never_used_blocknum.b;
invariant(t->smallest_never_used_blocknum.b >= RESERVED_BLOCKNUMS);
t->blocknum_freelist_head = rbuf_blocknum(&rb);
XMALLOC_N(t->length_of_array, t->block_translation);
for (int64_t i = 0; i < t->length_of_array; i++) {
t->block_translation[i].u.diskoff = rbuf_DISKOFF(&rb);
t->block_translation[i].size = rbuf_DISKOFF(&rb);
}
invariant(_calculate_size_on_disk(t) == (int64_t)size_on_disk);
invariant(t->block_translation[RESERVED_BLOCKNUM_TRANSLATION].size ==
(int64_t)size_on_disk);
invariant(t->block_translation[RESERVED_BLOCKNUM_TRANSLATION].u.diskoff ==
location_on_disk);
exit:
return r;
}
int block_table::iterate(enum translation_type type,
BLOCKTABLE_CALLBACK f,
void *extra,
bool data_only,
bool used_only) {
struct translation *src;
int r = 0;
switch (type) {
case TRANSLATION_CURRENT:
src = &_current;
break;
case TRANSLATION_INPROGRESS:
src = &_inprogress;
break;
case TRANSLATION_CHECKPOINTED:
src = &_checkpointed;
break;
default:
r = EINVAL;
}
struct translation fakecurrent;
memset(&fakecurrent, 0, sizeof(struct translation));
struct translation *t = &fakecurrent;
if (r == 0) {
_mutex_lock();
_copy_translation(t, src, TRANSLATION_DEBUG);
t->block_translation[RESERVED_BLOCKNUM_TRANSLATION] =
src->block_translation[RESERVED_BLOCKNUM_TRANSLATION];
_mutex_unlock();
int64_t i;
for (i = 0; i < t->smallest_never_used_blocknum.b; i++) {
struct block_translation_pair pair = t->block_translation[i];
if (data_only && i < RESERVED_BLOCKNUMS)
continue;
if (used_only && pair.size <= 0)
continue;
r = f(make_blocknum(i), pair.size, pair.u.diskoff, extra);
if (r != 0)
break;
}
toku_free(t->block_translation);
}
return r;
}
typedef struct {
int64_t used_space;
int64_t total_space;
} frag_extra;
static int frag_helper(BLOCKNUM UU(b),
int64_t size,
int64_t address,
void *extra) {
frag_extra *info = (frag_extra *)extra;
if (size + address > info->total_space)
info->total_space = size + address;
info->used_space += size;
return 0;
}
void block_table::internal_fragmentation(int64_t *total_sizep,
int64_t *used_sizep) {
frag_extra info = {0, 0};
int r = iterate(TRANSLATION_CHECKPOINTED, frag_helper, &info, false, true);
invariant_zero(r);
if (total_sizep)
*total_sizep = info.total_space;
if (used_sizep)
*used_sizep = info.used_space;
}
void block_table::_realloc_descriptor_on_disk_unlocked(DISKOFF size,
DISKOFF *offset,
FT ft) {
toku_mutex_assert_locked(&_mutex);
BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_DESCRIPTOR);
_realloc_on_disk_internal(b, size, offset, ft, false);
}
void block_table::realloc_descriptor_on_disk(DISKOFF size,
DISKOFF *offset,
FT ft,
int fd) {
_mutex_lock();
_realloc_descriptor_on_disk_unlocked(size, offset, ft);
_ensure_safe_write_unlocked(fd, size, *offset);
_mutex_unlock();
}
void block_table::get_descriptor_offset_size(DISKOFF *offset, DISKOFF *size) {
_mutex_lock();
BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_DESCRIPTOR);
_translate_blocknum_to_offset_size_unlocked(b, offset, size);
_mutex_unlock();
}
void block_table::get_fragmentation_unlocked(TOKU_DB_FRAGMENTATION report) {
// Requires: blocktable lock is held.
// Requires: report->file_size_bytes is already filled in.
// Count the headers.
report->data_bytes = BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE;
report->data_blocks = 1;
report->checkpoint_bytes_additional =
BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE;
report->checkpoint_blocks_additional = 1;
struct translation *current = &_current;
for (int64_t i = 0; i < current->length_of_array; i++) {
struct block_translation_pair *pair = ¤t->block_translation[i];
if (pair->size > 0) {
report->data_bytes += pair->size;
report->data_blocks++;
}
}
struct translation *checkpointed = &_checkpointed;
for (int64_t i = 0; i < checkpointed->length_of_array; i++) {
struct block_translation_pair *pair =
&checkpointed->block_translation[i];
if (pair->size > 0 &&
!(i < current->length_of_array &&
current->block_translation[i].size > 0 &&
current->block_translation[i].u.diskoff == pair->u.diskoff)) {
report->checkpoint_bytes_additional += pair->size;
report->checkpoint_blocks_additional++;
}
}
struct translation *inprogress = &_inprogress;
for (int64_t i = 0; i < inprogress->length_of_array; i++) {
struct block_translation_pair *pair = &inprogress->block_translation[i];
if (pair->size > 0 &&
!(i < current->length_of_array &&
current->block_translation[i].size > 0 &&
current->block_translation[i].u.diskoff == pair->u.diskoff) &&
!(i < checkpointed->length_of_array &&
checkpointed->block_translation[i].size > 0 &&
checkpointed->block_translation[i].u.diskoff ==
pair->u.diskoff)) {
report->checkpoint_bytes_additional += pair->size;
report->checkpoint_blocks_additional++;
}
}
_bt_block_allocator->UnusedStatistics(report);
}
void block_table::get_info64(struct ftinfo64 *s) {
_mutex_lock();
struct translation *current = &_current;
s->num_blocks_allocated = current->length_of_array;
s->num_blocks_in_use = 0;
s->size_allocated = 0;
s->size_in_use = 0;
for (int64_t i = 0; i < current->length_of_array; ++i) {
struct block_translation_pair *block = ¤t->block_translation[i];
if (block->size != size_is_free) {
++s->num_blocks_in_use;
s->size_in_use += block->size;
if (block->u.diskoff != diskoff_unused) {
uint64_t limit = block->u.diskoff + block->size;
if (limit > s->size_allocated) {
s->size_allocated = limit;
}
}
}
}
_mutex_unlock();
}
int block_table::iterate_translation_tables(
uint64_t checkpoint_count,
int (*iter)(uint64_t checkpoint_count,
int64_t total_num_rows,
int64_t blocknum,
int64_t diskoff,
int64_t size,
void *extra),
void *iter_extra) {
int error = 0;
_mutex_lock();
int64_t total_num_rows =
_current.length_of_array + _checkpointed.length_of_array;
for (int64_t i = 0; error == 0 && i < _current.length_of_array; ++i) {
struct block_translation_pair *block = &_current.block_translation[i];
error = iter(checkpoint_count,
total_num_rows,
i,
block->u.diskoff,
block->size,
iter_extra);
}
for (int64_t i = 0; error == 0 && i < _checkpointed.length_of_array; ++i) {
struct block_translation_pair *block =
&_checkpointed.block_translation[i];
error = iter(checkpoint_count - 1,
total_num_rows,
i,
block->u.diskoff,
block->size,
iter_extra);
}
_mutex_unlock();
return error;
}
|