solana_accounts_db/in_mem_accounts_index.rs
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 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158
use {
crate::{
accounts_index::{
AccountMapEntry, AccountMapEntryInner, AccountMapEntryMeta, DiskIndexValue, IndexValue,
PreAllocatedAccountMapEntry, RefCount, SlotList, UpsertReclaim, ZeroLamport,
},
bucket_map_holder::{Age, AtomicAge, BucketMapHolder},
bucket_map_holder_stats::BucketMapHolderStats,
waitable_condvar::WaitableCondvar,
},
rand::{thread_rng, Rng},
solana_bucket_map::bucket_api::BucketApi,
solana_measure::measure::Measure,
solana_sdk::{clock::Slot, pubkey::Pubkey},
std::{
collections::{hash_map::Entry, HashMap, HashSet},
fmt::Debug,
ops::{Bound, RangeBounds, RangeInclusive},
sync::{
atomic::{AtomicBool, AtomicU64, Ordering},
Arc, Mutex, RwLock, RwLockWriteGuard,
},
},
};
type K = Pubkey;
type CacheRangesHeld = RwLock<Vec<RangeInclusive<Pubkey>>>;
type InMemMap<T> = HashMap<Pubkey, AccountMapEntry<T>>;
#[derive(Debug, Default)]
pub struct StartupStats {
pub copy_data_us: AtomicU64,
}
#[derive(Debug)]
pub struct PossibleEvictions<T: IndexValue> {
/// vec per age in the future, up to size 'ages_to_stay_in_cache'
possible_evictions: Vec<FlushScanResult<T>>,
/// next index to use into 'possible_evictions'
/// if 'index' >= 'possible_evictions.len()', then there are no available entries
index: usize,
}
impl<T: IndexValue> PossibleEvictions<T> {
fn new(max_ages: Age) -> Self {
Self {
possible_evictions: (0..max_ages).map(|_| FlushScanResult::default()).collect(),
index: max_ages as usize, // initially no data
}
}
/// remove the possible evictions. This is required because we need ownership of the Arc strong counts to transfer to caller so entries can be removed from the accounts index
fn get_possible_evictions(&mut self) -> Option<FlushScanResult<T>> {
self.possible_evictions.get_mut(self.index).map(|result| {
self.index += 1;
// remove the list from 'possible_evictions'
std::mem::take(result)
})
}
/// clear existing data and prepare to add 'entries' more ages of data
fn reset(&mut self, entries: Age) {
self.possible_evictions.iter_mut().for_each(|entry| {
entry.evictions_random.clear();
entry.evictions_age_possible.clear();
});
let entries = entries as usize;
assert!(
entries <= self.possible_evictions.len(),
"entries: {}, len: {}",
entries,
self.possible_evictions.len()
);
self.index = self.possible_evictions.len() - entries;
}
/// insert 'entry' at 'relative_age' in the future into 'possible_evictions'
fn insert(&mut self, relative_age: Age, key: Pubkey, entry: AccountMapEntry<T>, random: bool) {
let index = self.index + (relative_age as usize);
let list = &mut self.possible_evictions[index];
if random {
&mut list.evictions_random
} else {
&mut list.evictions_age_possible
}
.push((key, entry));
}
}
// one instance of this represents one bin of the accounts index.
pub struct InMemAccountsIndex<T: IndexValue, U: DiskIndexValue + From<T> + Into<T>> {
last_age_flushed: AtomicAge,
// backing store
map_internal: RwLock<InMemMap<T>>,
storage: Arc<BucketMapHolder<T, U>>,
bin: usize,
bucket: Option<Arc<BucketApi<(Slot, U)>>>,
// pubkey ranges that this bin must hold in the cache while the range is present in this vec
pub cache_ranges_held: CacheRangesHeld,
// incremented each time stop_evictions is changed
stop_evictions_changes: AtomicU64,
// true while ranges are being manipulated. Used to keep an async flush from removing things while a range is being held.
stop_evictions: AtomicU64,
// set to true while this bin is being actively flushed
flushing_active: AtomicBool,
/// info to streamline initial index generation
startup_info: StartupInfo<T, U>,
/// possible evictions for next few slots coming up
possible_evictions: RwLock<PossibleEvictions<T>>,
/// how many more ages to skip before this bucket is flushed (as opposed to being skipped).
/// When this reaches 0, this bucket is flushed.
remaining_ages_to_skip_flushing: AtomicAge,
/// an individual bucket will evict its entries and write to disk every 1/NUM_AGES_TO_DISTRIBUTE_FLUSHES ages
/// Higher numbers mean we flush less buckets/s
/// Lower numbers mean we flush more buckets/s
num_ages_to_distribute_flushes: Age,
/// stats related to starting up
pub(crate) startup_stats: Arc<StartupStats>,
}
impl<T: IndexValue, U: DiskIndexValue + From<T> + Into<T>> Debug for InMemAccountsIndex<T, U> {
fn fmt(&self, _f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
Ok(())
}
}
pub enum InsertNewEntryResults {
DidNotExist,
ExistedNewEntryZeroLamports,
ExistedNewEntryNonZeroLamports,
}
#[derive(Default, Debug)]
struct StartupInfoDuplicates<T: IndexValue> {
/// entries that were found to have duplicate index entries.
/// When all entries have been inserted, these can be resolved and held in memory.
duplicates: Vec<(Slot, Pubkey, T)>,
/// pubkeys that were already added to disk and later found to be duplicates,
duplicates_put_on_disk: HashSet<(Slot, Pubkey)>,
}
#[derive(Default, Debug)]
struct StartupInfo<T: IndexValue, U: DiskIndexValue + From<T> + Into<T>> {
/// entries to add next time we are flushing to disk
insert: Mutex<Vec<(Pubkey, (Slot, U))>>,
/// pubkeys with more than 1 entry
duplicates: Mutex<StartupInfoDuplicates<T>>,
}
#[derive(Default, Debug)]
/// result from scanning in-mem index during flush
struct FlushScanResult<T> {
/// pubkeys whose age indicates they may be evicted now, pending further checks.
evictions_age_possible: Vec<(Pubkey, AccountMapEntry<T>)>,
/// pubkeys chosen to evict based on random eviction
evictions_random: Vec<(Pubkey, AccountMapEntry<T>)>,
}
impl<T: IndexValue, U: DiskIndexValue + From<T> + Into<T>> InMemAccountsIndex<T, U> {
pub fn new(storage: &Arc<BucketMapHolder<T, U>>, bin: usize) -> Self {
let num_ages_to_distribute_flushes = Age::MAX - storage.ages_to_stay_in_cache;
Self {
map_internal: RwLock::default(),
storage: Arc::clone(storage),
bin,
bucket: storage
.disk
.as_ref()
.map(|disk| disk.get_bucket_from_index(bin))
.map(Arc::clone),
cache_ranges_held: CacheRangesHeld::default(),
stop_evictions_changes: AtomicU64::default(),
stop_evictions: AtomicU64::default(),
flushing_active: AtomicBool::default(),
// initialize this to max, to make it clear we have not flushed at age 0, the starting age
last_age_flushed: AtomicAge::new(Age::MAX),
startup_info: StartupInfo::default(),
possible_evictions: RwLock::new(PossibleEvictions::new(1)),
// Spread out the scanning across all ages within the window.
// This causes us to scan 1/N of the bins each 'Age'
remaining_ages_to_skip_flushing: AtomicAge::new(
thread_rng().gen_range(0..num_ages_to_distribute_flushes),
),
num_ages_to_distribute_flushes,
startup_stats: Arc::clone(&storage.startup_stats),
}
}
/// true if this bucket needs to call flush for the current age
/// we need to scan each bucket once per value of age
fn get_should_age(&self, age: Age) -> bool {
let last_age_flushed = self.last_age_flushed();
last_age_flushed != age
}
/// called after flush scans this bucket at the current age
fn set_has_aged(&self, age: Age, can_advance_age: bool) {
self.last_age_flushed.store(age, Ordering::Release);
self.storage.bucket_flushed_at_current_age(can_advance_age);
}
fn last_age_flushed(&self) -> Age {
self.last_age_flushed.load(Ordering::Acquire)
}
/// Release entire in-mem hashmap to free all memory associated with it.
/// Idea is that during startup we needed a larger map than we need during runtime.
/// When using disk-buckets, in-mem index grows over time with dynamic use and then shrinks, in theory back to 0.
pub fn shrink_to_fit(&self) {
// shrink_to_fit could be quite expensive on large map sizes, which 'no disk buckets' could produce, so avoid shrinking in case we end up here
if self.storage.is_disk_index_enabled() {
self.map_internal.write().unwrap().shrink_to_fit();
}
}
pub fn items<R>(&self, range: &R) -> Vec<(K, AccountMapEntry<T>)>
where
R: RangeBounds<Pubkey> + std::fmt::Debug,
{
let m = Measure::start("items");
self.hold_range_in_memory(range, true);
let map = self.map_internal.read().unwrap();
let mut result = Vec::with_capacity(map.len());
map.iter().for_each(|(k, v)| {
if range.contains(k) {
result.push((*k, Arc::clone(v)));
}
});
drop(map);
self.hold_range_in_memory(range, false);
Self::update_stat(&self.stats().items, 1);
Self::update_time_stat(&self.stats().items_us, m);
result
}
// only called in debug code paths
pub fn keys(&self) -> Vec<Pubkey> {
Self::update_stat(&self.stats().keys, 1);
// easiest implementation is to load everything from disk into cache and return the keys
let evictions_guard = EvictionsGuard::lock(self);
self.put_range_in_cache(&None::<&RangeInclusive<Pubkey>>, &evictions_guard);
let keys = self.map_internal.read().unwrap().keys().cloned().collect();
keys
}
fn load_from_disk(&self, pubkey: &Pubkey) -> Option<(SlotList<U>, RefCount)> {
self.bucket.as_ref().and_then(|disk| {
let m = Measure::start("load_disk_found_count");
let entry_disk = disk.read_value(pubkey);
match &entry_disk {
Some(_) => {
Self::update_time_stat(&self.stats().load_disk_found_us, m);
Self::update_stat(&self.stats().load_disk_found_count, 1);
}
None => {
Self::update_time_stat(&self.stats().load_disk_missing_us, m);
Self::update_stat(&self.stats().load_disk_missing_count, 1);
}
}
entry_disk
})
}
/// lookup 'pubkey' in disk map.
/// If it is found, convert it to a cache entry and return the cache entry.
/// Cache entries from this function will always not be dirty.
fn load_account_entry_from_disk(&self, pubkey: &Pubkey) -> Option<AccountMapEntry<T>> {
let entry_disk = self.load_from_disk(pubkey)?; // returns None if not on disk
let entry_cache = self.disk_to_cache_entry(entry_disk.0, entry_disk.1);
debug_assert!(!entry_cache.dirty());
Some(entry_cache)
}
/// lookup 'pubkey' by only looking in memory. Does not look on disk.
/// callback is called whether pubkey is found or not
fn get_only_in_mem<RT>(
&self,
pubkey: &K,
update_age: bool,
callback: impl for<'a> FnOnce(Option<&'a AccountMapEntry<T>>) -> RT,
) -> RT {
let mut found = true;
let mut m = Measure::start("get");
let result = {
let map = self.map_internal.read().unwrap();
let result = map.get(pubkey);
m.stop();
callback(if let Some(entry) = result {
if update_age {
self.set_age_to_future(entry, false);
}
Some(entry)
} else {
drop(map);
found = false;
None
})
};
let stats = self.stats();
let (count, time) = if found {
(&stats.gets_from_mem, &stats.get_mem_us)
} else {
(&stats.gets_missing, &stats.get_missing_us)
};
Self::update_stat(time, m.as_us());
Self::update_stat(count, 1);
result
}
/// lookup 'pubkey' in index (in mem or on disk)
pub fn get(&self, pubkey: &K) -> Option<AccountMapEntry<T>> {
self.get_internal(pubkey, |entry| (true, entry.map(Arc::clone)))
}
/// set age of 'entry' to the future
/// if 'is_cached', age will be set farther
fn set_age_to_future(&self, entry: &AccountMapEntry<T>, is_cached: bool) {
entry.set_age(self.storage.future_age_to_flush(is_cached));
}
/// lookup 'pubkey' in index (in_mem or disk).
/// call 'callback' whether found or not
pub fn get_internal<RT>(
&self,
pubkey: &K,
// return true if item should be added to in_mem cache
callback: impl for<'a> FnOnce(Option<&AccountMapEntry<T>>) -> (bool, RT),
) -> RT {
self.get_only_in_mem(pubkey, true, |entry| {
if let Some(entry) = entry {
callback(Some(entry)).1
} else {
// not in cache, look on disk
let stats = self.stats();
let disk_entry = self.load_account_entry_from_disk(pubkey);
if disk_entry.is_none() {
return callback(None).1;
}
let disk_entry = disk_entry.unwrap();
let mut map = self.map_internal.write().unwrap();
let entry = map.entry(*pubkey);
match entry {
Entry::Occupied(occupied) => callback(Some(occupied.get())).1,
Entry::Vacant(vacant) => {
debug_assert!(!disk_entry.dirty());
let (add_to_cache, rt) = callback(Some(&disk_entry));
// We are holding a write lock to the in-memory map.
// This pubkey is not in the in-memory map.
// If the entry is now dirty, then it must be put in the cache or the modifications will be lost.
if add_to_cache || disk_entry.dirty() {
stats.inc_mem_count(self.bin);
vacant.insert(disk_entry);
}
rt
}
}
}
})
}
fn remove_if_slot_list_empty_value(&self, is_empty: bool) -> bool {
if is_empty {
self.stats().inc_delete();
true
} else {
false
}
}
fn delete_disk_key(&self, pubkey: &Pubkey) {
if let Some(disk) = self.bucket.as_ref() {
disk.delete_key(pubkey)
}
}
/// return false if the entry is in the index (disk or memory) and has a slot list len > 0
/// return true in all other cases, including if the entry is NOT in the index at all
fn remove_if_slot_list_empty_entry(&self, entry: Entry<K, AccountMapEntry<T>>) -> bool {
match entry {
Entry::Occupied(occupied) => {
let result = self.remove_if_slot_list_empty_value(
occupied.get().slot_list.read().unwrap().is_empty(),
);
if result {
// note there is a potential race here that has existed.
// if someone else holds the arc,
// then they think the item is still in the index and can make modifications.
// We have to have a write lock to the map here, which means nobody else can get
// the arc, but someone may already have retrieved a clone of it.
// account index in_mem flushing is one such possibility
self.delete_disk_key(occupied.key());
self.stats().dec_mem_count(self.bin);
occupied.remove();
}
result
}
Entry::Vacant(vacant) => {
// not in cache, look on disk
let entry_disk = self.load_from_disk(vacant.key());
match entry_disk {
Some(entry_disk) => {
// on disk
if self.remove_if_slot_list_empty_value(entry_disk.0.is_empty()) {
// not in cache, but on disk, so just delete from disk
self.delete_disk_key(vacant.key());
true
} else {
// could insert into cache here, but not required for correctness and value is unclear
false
}
}
None => true, // not in cache or on disk, but slot list is 'empty' and entry is not in index, so return true
}
}
}
}
// If the slot list for pubkey exists in the index and is empty, remove the index entry for pubkey and return true.
// Return false otherwise.
pub fn remove_if_slot_list_empty(&self, pubkey: Pubkey) -> bool {
let mut m = Measure::start("entry");
let mut map = self.map_internal.write().unwrap();
let entry = map.entry(pubkey);
m.stop();
let found = matches!(entry, Entry::Occupied(_));
let result = self.remove_if_slot_list_empty_entry(entry);
drop(map);
self.update_entry_stats(m, found);
result
}
pub fn slot_list_mut<RT>(
&self,
pubkey: &Pubkey,
user: impl for<'a> FnOnce(&mut RwLockWriteGuard<'a, SlotList<T>>) -> RT,
) -> Option<RT> {
self.get_internal(pubkey, |entry| {
(
true,
entry.map(|entry| {
let result = user(&mut entry.slot_list.write().unwrap());
entry.set_dirty(true);
result
}),
)
})
}
/// update 'entry' with 'new_value'
fn update_slot_list_entry(
&self,
entry: &AccountMapEntry<T>,
new_value: PreAllocatedAccountMapEntry<T>,
other_slot: Option<Slot>,
reclaims: &mut SlotList<T>,
reclaim: UpsertReclaim,
) {
let new_value: (Slot, T) = new_value.into();
let mut upsert_cached = new_value.1.is_cached();
if Self::lock_and_update_slot_list(entry, new_value, other_slot, reclaims, reclaim) > 1 {
// if slot list > 1, then we are going to hold this entry in memory until it gets set back to 1
upsert_cached = true;
}
self.set_age_to_future(entry, upsert_cached);
}
pub fn upsert(
&self,
pubkey: &Pubkey,
new_value: PreAllocatedAccountMapEntry<T>,
other_slot: Option<Slot>,
reclaims: &mut SlotList<T>,
reclaim: UpsertReclaim,
) {
let mut updated_in_mem = true;
// try to get it just from memory first using only a read lock
self.get_only_in_mem(pubkey, false, |entry| {
if let Some(entry) = entry {
self.update_slot_list_entry(entry, new_value, other_slot, reclaims, reclaim);
} else {
let mut m = Measure::start("entry");
let mut map = self.map_internal.write().unwrap();
let entry = map.entry(*pubkey);
m.stop();
let found = matches!(entry, Entry::Occupied(_));
match entry {
Entry::Occupied(mut occupied) => {
let current = occupied.get_mut();
self.update_slot_list_entry(
current, new_value, other_slot, reclaims, reclaim,
);
}
Entry::Vacant(vacant) => {
// not in cache, look on disk
updated_in_mem = false;
// go to in-mem cache first
let disk_entry = self.load_account_entry_from_disk(vacant.key());
let new_value = if let Some(disk_entry) = disk_entry {
// on disk, so merge new_value with what was on disk
self.update_slot_list_entry(
&disk_entry,
new_value,
other_slot,
reclaims,
reclaim,
);
disk_entry
} else {
// not on disk, so insert new thing
self.stats().inc_insert();
new_value.into_account_map_entry(&self.storage)
};
assert!(new_value.dirty());
vacant.insert(new_value);
self.stats().inc_mem_count(self.bin);
}
};
drop(map);
self.update_entry_stats(m, found);
};
});
if updated_in_mem {
Self::update_stat(&self.stats().updates_in_mem, 1);
}
}
fn update_entry_stats(&self, stopped_measure: Measure, found: bool) {
let stats = self.stats();
let (count, time) = if found {
(&stats.entries_from_mem, &stats.entry_mem_us)
} else {
(&stats.entries_missing, &stats.entry_missing_us)
};
Self::update_stat(time, stopped_measure.as_us());
Self::update_stat(count, 1);
}
/// Try to update an item in the slot list the given `slot` If an item for the slot
/// already exists in the list, remove the older item, add it to `reclaims`, and insert
/// the new item.
/// if 'other_slot' is some, then also remove any entries in the slot list that are at 'other_slot'
/// return resulting len of slot list
pub fn lock_and_update_slot_list(
current: &AccountMapEntryInner<T>,
new_value: (Slot, T),
other_slot: Option<Slot>,
reclaims: &mut SlotList<T>,
reclaim: UpsertReclaim,
) -> usize {
let mut slot_list = current.slot_list.write().unwrap();
let (slot, new_entry) = new_value;
let addref = Self::update_slot_list(
&mut slot_list,
slot,
new_entry,
other_slot,
reclaims,
reclaim,
);
if addref {
current.addref();
}
current.set_dirty(true);
slot_list.len()
}
/// modifies slot_list
/// any entry at 'slot' or slot 'other_slot' is replaced with 'account_info'.
/// or, 'account_info' is appended to the slot list if the slot did not exist previously.
/// returns true if caller should addref
/// conditions when caller should addref:
/// 'account_info' does NOT represent a cached storage (the slot is being flushed from the cache)
/// AND
/// previous slot_list entry AT 'slot' did not exist (this is the first time this account was modified in this "slot"), or was previously cached (the storage is now being flushed from the cache)
/// Note that even if entry DID exist at 'other_slot', the above conditions apply.
fn update_slot_list(
slot_list: &mut SlotList<T>,
slot: Slot,
account_info: T,
mut other_slot: Option<Slot>,
reclaims: &mut SlotList<T>,
reclaim: UpsertReclaim,
) -> bool {
let mut addref = !account_info.is_cached();
if other_slot == Some(slot) {
other_slot = None; // redundant info, so ignore
}
// There may be 0..=2 dirty accounts found (one at 'slot' and one at 'other_slot')
// that are already in the slot list. Since the first one found will be swapped with the
// new account, if a second one is found, we cannot swap again. Instead, just remove it.
let mut found_slot = false;
let mut found_other_slot = false;
(0..slot_list.len())
.rev() // rev since we delete from the list in some cases
.for_each(|slot_list_index| {
let (cur_slot, cur_account_info) = &slot_list[slot_list_index];
let matched_slot = *cur_slot == slot;
if matched_slot || Some(*cur_slot) == other_slot {
// make sure neither 'slot' nor 'other_slot' are in the slot list more than once
let matched_other_slot = !matched_slot;
assert!(
!(found_slot && matched_slot || matched_other_slot && found_other_slot),
"{slot_list:?}, slot: {slot}, other_slot: {other_slot:?}"
);
let is_cur_account_cached = cur_account_info.is_cached();
let reclaim_item = if !(found_slot || found_other_slot) {
// first time we found an entry in 'slot' or 'other_slot', so replace it in-place.
// this may be the only instance we find
std::mem::replace(&mut slot_list[slot_list_index], (slot, account_info))
} else {
// already replaced one entry, so this one has to be removed
slot_list.remove(slot_list_index)
};
match reclaim {
UpsertReclaim::PopulateReclaims => {
reclaims.push(reclaim_item);
}
UpsertReclaim::PreviousSlotEntryWasCached => {
assert!(is_cur_account_cached);
}
UpsertReclaim::IgnoreReclaims => {
// do nothing. nothing to assert. nothing to return in reclaims
}
}
if matched_slot {
found_slot = true;
} else {
found_other_slot = true;
}
if !is_cur_account_cached {
// current info at 'slot' is NOT cached, so we should NOT addref. This slot already has a ref count for this pubkey.
addref = false;
}
}
});
if !found_slot && !found_other_slot {
// if we make it here, we did not find the slot in the list
slot_list.push((slot, account_info));
}
addref
}
// convert from raw data on disk to AccountMapEntry, set to age in future
fn disk_to_cache_entry(
&self,
slot_list: SlotList<U>,
ref_count: RefCount,
) -> AccountMapEntry<T> {
Arc::new(AccountMapEntryInner::new(
slot_list
.into_iter()
.map(|(slot, info)| (slot, info.into()))
.collect(),
ref_count,
AccountMapEntryMeta::new_clean(&self.storage),
))
}
pub fn len_for_stats(&self) -> usize {
self.stats().count_in_bucket(self.bin)
}
/// Queue up these insertions for when the flush thread is dealing with this bin.
/// This is very fast and requires no lookups or disk access.
pub fn startup_insert_only(&self, items: impl Iterator<Item = (Pubkey, (Slot, T))>) {
assert!(self.storage.get_startup());
assert!(self.bucket.is_some());
let mut insert = self.startup_info.insert.lock().unwrap();
let m = Measure::start("copy");
items
.into_iter()
.for_each(|(k, (slot, v))| insert.push((k, (slot, v.into()))));
self.startup_stats
.copy_data_us
.fetch_add(m.end_as_us(), Ordering::Relaxed);
}
pub fn insert_new_entry_if_missing_with_lock(
&self,
pubkey: Pubkey,
new_entry: PreAllocatedAccountMapEntry<T>,
) -> InsertNewEntryResults {
let mut m = Measure::start("entry");
let mut map = self.map_internal.write().unwrap();
let entry = map.entry(pubkey);
m.stop();
let new_entry_zero_lamports = new_entry.is_zero_lamport();
let (found_in_mem, already_existed) = match entry {
Entry::Occupied(occupied) => {
// in cache, so merge into cache
let (slot, account_info) = new_entry.into();
InMemAccountsIndex::<T, U>::lock_and_update_slot_list(
occupied.get(),
(slot, account_info),
None, // should be None because we don't expect a different slot # during index generation
&mut Vec::default(),
UpsertReclaim::PopulateReclaims, // this should be ignore?
);
(
true, /* found in mem */
true, /* already existed */
)
}
Entry::Vacant(vacant) => {
// not in cache, look on disk
let disk_entry = self.load_account_entry_from_disk(vacant.key());
self.stats().inc_mem_count(self.bin);
if let Some(disk_entry) = disk_entry {
let (slot, account_info) = new_entry.into();
InMemAccountsIndex::<T, U>::lock_and_update_slot_list(
&disk_entry,
(slot, account_info),
// None because we are inserting the first element in the slot list for this pubkey.
// There can be no 'other' slot in the list.
None,
&mut Vec::default(),
UpsertReclaim::PopulateReclaims,
);
vacant.insert(disk_entry);
(
false, /* found in mem */
true, /* already existed */
)
} else {
// not on disk, so insert new thing and we're done
let new_entry: AccountMapEntry<T> =
new_entry.into_account_map_entry(&self.storage);
assert!(new_entry.dirty());
vacant.insert(new_entry);
(false, false)
}
}
};
drop(map);
self.update_entry_stats(m, found_in_mem);
let stats = self.stats();
if !already_existed {
stats.inc_insert();
} else {
Self::update_stat(&stats.updates_in_mem, 1);
}
if !already_existed {
InsertNewEntryResults::DidNotExist
} else if new_entry_zero_lamports {
InsertNewEntryResults::ExistedNewEntryZeroLamports
} else {
InsertNewEntryResults::ExistedNewEntryNonZeroLamports
}
}
/// Look at the currently held ranges. If 'range' is already included in what is
/// being held, then add 'range' to the currently held list AND return true
/// If 'range' is NOT already included in what is being held, then return false
/// withOUT adding 'range' to the list of what is currently held
fn add_hold_range_in_memory_if_already_held<R>(
&self,
range: &R,
evictions_guard: &EvictionsGuard,
) -> bool
where
R: RangeBounds<Pubkey>,
{
let start_holding = true;
let only_add_if_already_held = true;
self.just_set_hold_range_in_memory_internal(
range,
start_holding,
only_add_if_already_held,
evictions_guard,
)
}
fn just_set_hold_range_in_memory<R>(
&self,
range: &R,
start_holding: bool,
evictions_guard: &EvictionsGuard,
) where
R: RangeBounds<Pubkey>,
{
let only_add_if_already_held = false;
let _ = self.just_set_hold_range_in_memory_internal(
range,
start_holding,
only_add_if_already_held,
evictions_guard,
);
}
/// if 'start_holding', then caller wants to add 'range' to the list of ranges being held
/// if !'start_holding', then caller wants to remove 'range' to the list
/// if 'only_add_if_already_held', caller intends to only add 'range' to the list if the range is already held
/// returns true iff start_holding=true and the range we're asked to hold was already being held
fn just_set_hold_range_in_memory_internal<R>(
&self,
range: &R,
start_holding: bool,
only_add_if_already_held: bool,
_evictions_guard: &EvictionsGuard,
) -> bool
where
R: RangeBounds<Pubkey>,
{
assert!(!only_add_if_already_held || start_holding);
let start = match range.start_bound() {
Bound::Included(bound) | Bound::Excluded(bound) => *bound,
Bound::Unbounded => Pubkey::from([0; 32]),
};
let end = match range.end_bound() {
Bound::Included(bound) | Bound::Excluded(bound) => *bound,
Bound::Unbounded => Pubkey::from([0xff; 32]),
};
// this becomes inclusive - that is ok - we are just roughly holding a range of items.
// inclusive is bigger than exclusive so we may hold 1 extra item worst case
let inclusive_range = start..=end;
let mut ranges = self.cache_ranges_held.write().unwrap();
let mut already_held = false;
if start_holding {
if only_add_if_already_held {
for r in ranges.iter() {
if r.contains(&start) && r.contains(&end) {
already_held = true;
break;
}
}
}
if already_held || !only_add_if_already_held {
ranges.push(inclusive_range);
}
} else {
// find the matching range and delete it since we don't want to hold it anymore
// search backwards, assuming LIFO ordering
for (i, r) in ranges.iter().enumerate().rev() {
if let (Bound::Included(start_found), Bound::Included(end_found)) =
(r.start_bound(), r.end_bound())
{
if start_found == &start && end_found == &end {
// found a match. There may be dups, that's ok, we expect another call to remove the dup.
ranges.remove(i);
break;
}
}
}
}
already_held
}
/// if 'start_holding'=true, then:
/// at the end of this function, cache_ranges_held will be updated to contain 'range'
/// and all pubkeys in that range will be in the in-mem cache
/// if 'start_holding'=false, then:
/// 'range' will be removed from cache_ranges_held
/// and all pubkeys will be eligible for being removed from in-mem cache in the bg if no other range is holding them
/// Any in-process flush will be aborted when it gets to evicting items from in-mem.
pub fn hold_range_in_memory<R>(&self, range: &R, start_holding: bool)
where
R: RangeBounds<Pubkey> + Debug,
{
let evictions_guard = EvictionsGuard::lock(self);
if !start_holding || !self.add_hold_range_in_memory_if_already_held(range, &evictions_guard)
{
if start_holding {
// put everything in the cache and it will be held there
self.put_range_in_cache(&Some(range), &evictions_guard);
}
// do this AFTER items have been put in cache - that way anyone who finds this range can know that the items are already in the cache
self.just_set_hold_range_in_memory(range, start_holding, &evictions_guard);
}
}
fn put_range_in_cache<R>(&self, range: &Option<&R>, _evictions_guard: &EvictionsGuard)
where
R: RangeBounds<Pubkey>,
{
assert!(self.get_stop_evictions()); // caller should be controlling the lifetime of how long this needs to be present
let m = Measure::start("range");
let mut added_to_mem = 0;
// load from disk
if let Some(disk) = self.bucket.as_ref() {
let mut map = self.map_internal.write().unwrap();
let items = disk.items_in_range(range); // map's lock has to be held while we are getting items from disk
let future_age = self.storage.future_age_to_flush(false);
for item in items {
let entry = map.entry(item.pubkey);
match entry {
Entry::Occupied(occupied) => {
// item already in cache, bump age to future. This helps the current age flush to succeed.
occupied.get().set_age(future_age);
}
Entry::Vacant(vacant) => {
vacant.insert(self.disk_to_cache_entry(item.slot_list, item.ref_count));
added_to_mem += 1;
}
}
}
}
self.stats().add_mem_count(self.bin, added_to_mem);
Self::update_time_stat(&self.stats().get_range_us, m);
}
/// returns true if there are active requests to stop evictions
fn get_stop_evictions(&self) -> bool {
self.stop_evictions.load(Ordering::Acquire) > 0
}
/// return count of calls to 'start_stop_evictions', indicating changes could have been made to eviction strategy
fn get_stop_evictions_changes(&self) -> u64 {
self.stop_evictions_changes.load(Ordering::Acquire)
}
pub fn flush(&self, can_advance_age: bool) {
if let Some(flush_guard) = FlushGuard::lock(&self.flushing_active) {
self.flush_internal(&flush_guard, can_advance_age)
}
}
/// returns true if a dice roll indicates this call should result in a random eviction.
/// This causes non-determinism in cache contents per validator.
fn random_chance_of_eviction() -> bool {
// random eviction
const N: usize = 1000;
// 1/N chance of eviction
thread_rng().gen_range(0..N) == 0
}
/// assumes 1 entry in the slot list. Ignores overhead of the HashMap and such
fn approx_size_of_one_entry() -> usize {
std::mem::size_of::<T>()
+ std::mem::size_of::<Pubkey>()
+ std::mem::size_of::<AccountMapEntry<T>>()
}
fn should_evict_based_on_age(
current_age: Age,
entry: &AccountMapEntry<T>,
startup: bool,
ages_flushing_now: Age,
) -> bool {
startup || current_age.wrapping_sub(entry.age()) <= ages_flushing_now
}
/// return true if 'entry' should be evicted from the in-mem index
fn should_evict_from_mem<'a>(
&self,
current_age: Age,
entry: &'a AccountMapEntry<T>,
startup: bool,
update_stats: bool,
exceeds_budget: bool,
ages_flushing_now: Age,
) -> (bool, Option<std::sync::RwLockReadGuard<'a, SlotList<T>>>) {
// this could be tunable dynamically based on memory pressure
// we could look at more ages or we could throw out more items we are choosing to keep in the cache
if Self::should_evict_based_on_age(current_age, entry, startup, ages_flushing_now) {
if exceeds_budget {
// if we are already holding too many items in-mem, then we need to be more aggressive at kicking things out
(true, None)
} else if entry.ref_count() != 1 {
Self::update_stat(&self.stats().held_in_mem.ref_count, 1);
(false, None)
} else {
// only read the slot list if we are planning to throw the item out
let slot_list = entry.slot_list.read().unwrap();
if slot_list.len() != 1 {
if update_stats {
Self::update_stat(&self.stats().held_in_mem.slot_list_len, 1);
}
(false, None) // keep 0 and > 1 slot lists in mem. They will be cleaned or shrunk soon.
} else {
// keep items with slot lists that contained cached items
let evict = !slot_list.iter().any(|(_, info)| info.is_cached());
if !evict && update_stats {
Self::update_stat(&self.stats().held_in_mem.slot_list_cached, 1);
}
(evict, if evict { Some(slot_list) } else { None })
}
}
} else {
(false, None)
}
}
/// fill in `possible_evictions` from `iter` by checking age
fn gather_possible_evictions<'a>(
iter: impl Iterator<Item = (&'a Pubkey, &'a Arc<AccountMapEntryInner<T>>)>,
possible_evictions: &mut PossibleEvictions<T>,
startup: bool,
current_age: Age,
ages_flushing_now: Age,
can_randomly_flush: bool,
) {
for (k, v) in iter {
let mut random = false;
if !startup && current_age.wrapping_sub(v.age()) > ages_flushing_now {
if !can_randomly_flush || !Self::random_chance_of_eviction() {
// not planning to evict this item from memory within 'ages_flushing_now' ages
continue;
}
random = true;
}
possible_evictions.insert(0, *k, Arc::clone(v), random);
}
}
/// scan loop
/// holds read lock
/// identifies items which are potential candidates to evict
fn flush_scan(
&self,
current_age: Age,
startup: bool,
_flush_guard: &FlushGuard,
ages_flushing_now: Age,
) -> FlushScanResult<T> {
let mut possible_evictions = self.possible_evictions.write().unwrap();
possible_evictions.reset(1);
let m;
{
let map = self.map_internal.read().unwrap();
m = Measure::start("flush_scan"); // we don't care about lock time in this metric - bg threads can wait
Self::gather_possible_evictions(
map.iter(),
&mut possible_evictions,
startup,
current_age,
ages_flushing_now,
true,
);
}
Self::update_time_stat(&self.stats().flush_scan_us, m);
possible_evictions.get_possible_evictions().unwrap()
}
fn write_startup_info_to_disk(&self) {
let insert = std::mem::take(&mut *self.startup_info.insert.lock().unwrap());
if insert.is_empty() {
// nothing to insert for this bin
return;
}
// during startup, nothing should be in the in-mem map
let map_internal = self.map_internal.read().unwrap();
assert!(
map_internal.is_empty(),
"len: {}, first: {:?}",
map_internal.len(),
map_internal.iter().take(1).collect::<Vec<_>>()
);
drop(map_internal);
// this fn should only be called from a single thread, so holding the lock is fine
let mut duplicates = self.startup_info.duplicates.lock().unwrap();
// merge all items into the disk index now
let disk = self.bucket.as_ref().unwrap();
let mut count = insert.len() as u64;
for (i, duplicate_entry) in disk.batch_insert_non_duplicates(&insert) {
let (k, entry) = &insert[i];
duplicates.duplicates.push((entry.0, *k, entry.1.into()));
// accurately account for there being a duplicate for the first entry that was previously added to the disk index.
// That entry could not have known yet that it was a duplicate.
// It is important to capture each slot with a duplicate because of slot limits applied to clean.
duplicates
.duplicates_put_on_disk
.insert((duplicate_entry.0, *k));
count -= 1;
}
self.stats().inc_insert_count(count);
}
/// pull out all duplicate pubkeys from 'startup_info'
/// duplicate pubkeys have a slot list with len > 1
/// These were collected for this bin when we did batch inserts in the bg flush threads.
/// Insert these into the in-mem index, then return the duplicate (Slot, Pubkey)
pub fn populate_and_retrieve_duplicate_keys_from_startup(&self) -> Vec<(Slot, Pubkey)> {
// in order to return accurate and complete duplicates, we must have nothing left remaining to insert
assert!(self.startup_info.insert.lock().unwrap().is_empty());
let mut duplicate_items = self.startup_info.duplicates.lock().unwrap();
let duplicates = std::mem::take(&mut duplicate_items.duplicates);
let duplicates_put_on_disk = std::mem::take(&mut duplicate_items.duplicates_put_on_disk);
drop(duplicate_items);
duplicates_put_on_disk
.into_iter()
.chain(duplicates.into_iter().map(|(slot, key, info)| {
let entry = PreAllocatedAccountMapEntry::new(slot, info, &self.storage, true);
self.insert_new_entry_if_missing_with_lock(key, entry);
(slot, key)
}))
.collect()
}
/// synchronize the in-mem index with the disk index
fn flush_internal(&self, flush_guard: &FlushGuard, can_advance_age: bool) {
let current_age = self.storage.current_age();
let iterate_for_age = self.get_should_age(current_age);
let startup = self.storage.get_startup();
if !iterate_for_age && !startup {
// no need to age, so no need to flush this bucket
// but, at startup we want to evict from buckets as fast as possible if any items exist
return;
}
if startup {
self.write_startup_info_to_disk();
}
let ages_flushing_now = if iterate_for_age && !startup {
let old_value = self
.remaining_ages_to_skip_flushing
.fetch_sub(1, Ordering::AcqRel);
if old_value == 0 {
self.remaining_ages_to_skip_flushing
.store(self.num_ages_to_distribute_flushes, Ordering::Release);
} else {
// skipping iteration of the buckets at the current age, but mark the bucket as having aged
assert_eq!(current_age, self.storage.current_age());
self.set_has_aged(current_age, can_advance_age);
return;
}
self.num_ages_to_distribute_flushes
} else {
// just 1 age to flush. 0 means age == age
0
};
Self::update_stat(&self.stats().buckets_scanned, 1);
// scan in-mem map for items that we may evict
let FlushScanResult {
mut evictions_age_possible,
mut evictions_random,
} = self.flush_scan(current_age, startup, flush_guard, ages_flushing_now);
// write to disk outside in-mem map read lock
{
let mut evictions_age = Vec::with_capacity(evictions_age_possible.len());
if !evictions_age_possible.is_empty() || !evictions_random.is_empty() {
let disk = self.bucket.as_ref().unwrap();
let mut flush_entries_updated_on_disk = 0;
let exceeds_budget = self.get_exceeds_budget();
let mut flush_should_evict_us = 0;
// we don't care about lock time in this metric - bg threads can wait
let m = Measure::start("flush_update");
// consider whether to write to disk for all the items we may evict, whether evicting due to age or random
for (is_random, check_for_eviction_and_dirty) in [
(false, &mut evictions_age_possible),
(true, &mut evictions_random),
] {
for (k, v) in check_for_eviction_and_dirty.drain(..) {
let mut slot_list = None;
if !is_random {
let mut mse = Measure::start("flush_should_evict");
let (evict_for_age, slot_list_temp) = self.should_evict_from_mem(
current_age,
&v,
startup,
true,
exceeds_budget,
ages_flushing_now,
);
slot_list = slot_list_temp;
mse.stop();
flush_should_evict_us += mse.as_us();
if evict_for_age {
evictions_age.push(k);
} else {
// not evicting, so don't write, even if dirty
continue;
}
} else if v.ref_count() != 1 {
continue;
}
if is_random && v.dirty() {
// Don't randomly evict dirty entries. Evicting dirty entries results in us writing entries with many slot list elements for example, unnecessarily.
// So, only randomly evict entries that lru would say don't throw away and were just read (or were dirty and written, but could not be evicted).
continue;
}
// if we are evicting it, then we need to update disk if we're dirty
if v.clear_dirty() {
// step 1: clear the dirty flag
// step 2: perform the update on disk based on the fields in the entry
// If a parallel operation dirties the item again - even while this flush is occurring,
// the last thing the writer will do, after updating contents, is set_dirty(true)
// That prevents dropping an item from cache before disk is updated to latest in mem.
// It is possible that the item in the cache is marked as dirty while these updates are happening. That is ok.
// The dirty will be picked up and the item will be prevented from being evicted.
// may have to loop if disk has to grow and we have to retry the write
loop {
let disk_resize = {
let slot_list = slot_list
.take()
.unwrap_or_else(|| v.slot_list.read().unwrap());
disk.try_write(
&k,
(
&slot_list
.iter()
.map(|(slot, info)| (*slot, (*info).into()))
.collect::<Vec<_>>(),
v.ref_count(),
),
)
};
match disk_resize {
Ok(_) => {
// successfully written to disk
flush_entries_updated_on_disk += 1;
break;
}
Err(err) => {
// disk needs to resize. This item did not get written. Resize and try again.
let m = Measure::start("flush_grow");
disk.grow(err);
Self::update_time_stat(&self.stats().flush_grow_us, m);
}
}
}
}
}
}
Self::update_time_stat(&self.stats().flush_update_us, m);
Self::update_stat(&self.stats().flush_should_evict_us, flush_should_evict_us);
Self::update_stat(
&self.stats().flush_entries_updated_on_disk,
flush_entries_updated_on_disk,
);
// remove the 'v'
let evictions_random = evictions_random
.into_iter()
.map(|(k, _v)| k)
.collect::<Vec<_>>();
let m = Measure::start("flush_evict");
self.evict_from_cache(
evictions_age,
current_age,
startup,
false,
ages_flushing_now,
);
self.evict_from_cache(
evictions_random,
current_age,
startup,
true,
ages_flushing_now,
);
Self::update_time_stat(&self.stats().flush_evict_us, m);
}
if iterate_for_age {
// completed iteration of the buckets at the current age
assert_eq!(current_age, self.storage.current_age());
self.set_has_aged(current_age, can_advance_age);
}
}
}
/// calculate the estimated size of the in-mem index
/// return whether the size exceeds the specified budget
fn get_exceeds_budget(&self) -> bool {
let in_mem_count = self.stats().count_in_mem.load(Ordering::Relaxed);
let limit = self.storage.mem_budget_mb;
let estimate_mem = in_mem_count * Self::approx_size_of_one_entry();
let exceeds_budget = limit
.map(|limit| estimate_mem >= limit * 1024 * 1024)
.unwrap_or_default();
self.stats()
.estimate_mem
.store(estimate_mem as u64, Ordering::Relaxed);
exceeds_budget
}
/// for each key in 'keys', look up in map, set age to the future
fn move_ages_to_future(&self, next_age: Age, current_age: Age, keys: &[Pubkey]) {
let map = self.map_internal.read().unwrap();
keys.iter().for_each(|key| {
if let Some(entry) = map.get(key) {
entry.try_exchange_age(next_age, current_age);
}
});
}
// evict keys in 'evictions' from in-mem cache, likely due to age
fn evict_from_cache(
&self,
mut evictions: Vec<Pubkey>,
current_age: Age,
startup: bool,
randomly_evicted: bool,
ages_flushing_now: Age,
) {
if evictions.is_empty() {
return;
}
let stop_evictions_changes_at_start = self.get_stop_evictions_changes();
let next_age_on_failure = self.storage.future_age_to_flush(false);
if self.get_stop_evictions() {
// ranges were changed
self.move_ages_to_future(next_age_on_failure, current_age, &evictions);
return;
}
let mut failed = 0;
// skip any keys that are held in memory because of ranges being held
let ranges = self.cache_ranges_held.read().unwrap().clone();
if !ranges.is_empty() {
let mut move_age = Vec::default();
evictions.retain(|k| {
if ranges.iter().any(|range| range.contains(k)) {
// this item is held in mem by range, so don't evict
move_age.push(*k);
false
} else {
true
}
});
if !move_age.is_empty() {
failed += move_age.len();
self.move_ages_to_future(next_age_on_failure, current_age, &move_age);
}
}
let mut evicted = 0;
// chunk these so we don't hold the write lock too long
for evictions in evictions.chunks(50) {
let mut map = self.map_internal.write().unwrap();
for k in evictions {
if let Entry::Occupied(occupied) = map.entry(*k) {
let v = occupied.get();
if Arc::strong_count(v) > 1 {
// someone is holding the value arc's ref count and could modify it, so do not evict
failed += 1;
v.try_exchange_age(next_age_on_failure, current_age);
continue;
}
if v.dirty()
|| (!randomly_evicted
&& !Self::should_evict_based_on_age(
current_age,
v,
startup,
ages_flushing_now,
))
{
// marked dirty or bumped in age after we looked above
// these evictions will be handled in later passes (at later ages)
// but, at startup, everything is ready to age out if it isn't dirty
failed += 1;
continue;
}
if stop_evictions_changes_at_start != self.get_stop_evictions_changes() {
// ranges were changed
failed += 1;
v.try_exchange_age(next_age_on_failure, current_age);
continue;
}
// all conditions for eviction succeeded, so really evict item from in-mem cache
evicted += 1;
occupied.remove();
}
}
if map.is_empty() {
map.shrink_to_fit();
}
}
self.stats().sub_mem_count(self.bin, evicted);
Self::update_stat(&self.stats().flush_entries_evicted_from_mem, evicted as u64);
Self::update_stat(&self.stats().failed_to_evict, failed as u64);
}
pub fn stats(&self) -> &BucketMapHolderStats {
&self.storage.stats
}
fn update_stat(stat: &AtomicU64, value: u64) {
if value != 0 {
stat.fetch_add(value, Ordering::Relaxed);
}
}
pub fn update_time_stat(stat: &AtomicU64, mut m: Measure) {
m.stop();
let value = m.as_us();
Self::update_stat(stat, value);
}
}
/// An RAII implementation of a scoped lock for the `flushing_active` atomic flag in
/// `InMemAccountsIndex`. When this structure is dropped (falls out of scope), the flag will be
/// cleared (set to false).
///
/// After successfully locking (calling `FlushGuard::lock()`), pass a reference to the `FlashGuard`
/// instance to any function/code that requires the `flushing_active` flag has been set (to true).
#[derive(Debug)]
struct FlushGuard<'a> {
flushing: &'a AtomicBool,
}
impl<'a> FlushGuard<'a> {
/// Set the `flushing` atomic flag to true. If the flag was already true, then return `None`
/// (so as to not clear the flag erroneously). Otherwise return `Some(FlushGuard)`.
#[must_use = "if unused, the `flushing` flag will immediately clear"]
fn lock(flushing: &'a AtomicBool) -> Option<Self> {
let already_flushing = flushing.swap(true, Ordering::AcqRel);
// Eager evaluation here would result in dropping Self and clearing flushing flag
#[allow(clippy::unnecessary_lazy_evaluations)]
(!already_flushing).then(|| Self { flushing })
}
}
impl Drop for FlushGuard<'_> {
fn drop(&mut self) {
self.flushing.store(false, Ordering::Release);
}
}
/// Disable (and safely enable) the background flusher from evicting entries from the in-mem
/// accounts index. When disabled, no entries may be evicted. When enabled, only eligible entries
/// may be evicted (i.e. those not in a held range).
///
/// An RAII implementation of a scoped lock for the `stop_evictions` atomic flag/counter in
/// `InMemAccountsIndex`. When this structure is dropped (falls out of scope), the counter will
/// decrement and conditionally notify its storage.
///
/// After successfully locking (calling `EvictionsGuard::lock()`), pass a reference to the
/// `EvictionsGuard` instance to any function/code that requires `stop_evictions` to be
/// incremented/decremented correctly.
#[derive(Debug)]
struct EvictionsGuard<'a> {
/// The number of active callers disabling evictions
stop_evictions: &'a AtomicU64,
/// The number of times that evictions have been disabled or enabled
num_state_changes: &'a AtomicU64,
/// Who will be notified after the evictions are re-enabled
storage_notifier: &'a WaitableCondvar,
}
impl<'a> EvictionsGuard<'a> {
#[must_use = "if unused, this evictions lock will be immediately unlocked"]
fn lock<T: IndexValue, U: DiskIndexValue + From<T> + Into<T>>(
in_mem_accounts_index: &'a InMemAccountsIndex<T, U>,
) -> Self {
Self::lock_with(
&in_mem_accounts_index.stop_evictions,
&in_mem_accounts_index.stop_evictions_changes,
&in_mem_accounts_index.storage.wait_dirty_or_aged,
)
}
#[must_use = "if unused, this evictions lock will be immediately unlocked"]
fn lock_with(
stop_evictions: &'a AtomicU64,
num_state_changes: &'a AtomicU64,
storage_notifier: &'a WaitableCondvar,
) -> Self {
num_state_changes.fetch_add(1, Ordering::Release);
stop_evictions.fetch_add(1, Ordering::Release);
Self {
stop_evictions,
num_state_changes,
storage_notifier,
}
}
}
impl Drop for EvictionsGuard<'_> {
fn drop(&mut self) {
let previous_value = self.stop_evictions.fetch_sub(1, Ordering::AcqRel);
debug_assert!(previous_value > 0);
let should_notify = previous_value == 1;
if should_notify {
// stop_evictions went to 0, so this bucket could now be ready to be aged
self.storage_notifier.notify_one();
}
self.num_state_changes.fetch_add(1, Ordering::Release);
}
}
#[cfg(test)]
mod tests {
use {
super::*,
crate::accounts_index::{AccountsIndexConfig, IndexLimitMb, BINS_FOR_TESTING},
assert_matches::assert_matches,
itertools::Itertools,
};
fn new_for_test<T: IndexValue>() -> InMemAccountsIndex<T, T> {
let holder = Arc::new(BucketMapHolder::new(
BINS_FOR_TESTING,
&Some(AccountsIndexConfig::default()),
1,
));
let bin = 0;
InMemAccountsIndex::new(&holder, bin)
}
fn new_disk_buckets_for_test<T: IndexValue>() -> InMemAccountsIndex<T, T> {
let holder = Arc::new(BucketMapHolder::new(
BINS_FOR_TESTING,
&Some(AccountsIndexConfig {
index_limit_mb: IndexLimitMb::Limit(1),
..AccountsIndexConfig::default()
}),
1,
));
let bin = 0;
let bucket = InMemAccountsIndex::new(&holder, bin);
assert!(bucket.storage.is_disk_index_enabled());
bucket
}
#[test]
fn test_should_evict_from_mem_ref_count() {
for ref_count in [0, 1, 2] {
let bucket = new_for_test::<u64>();
let startup = false;
let current_age = 0;
let one_element_slot_list = vec![(0, 0)];
let one_element_slot_list_entry = Arc::new(AccountMapEntryInner::new(
one_element_slot_list,
ref_count,
AccountMapEntryMeta::default(),
));
// exceeded budget
assert_eq!(
bucket
.should_evict_from_mem(
current_age,
&one_element_slot_list_entry,
startup,
false,
false,
1,
)
.0,
ref_count == 1
);
}
}
#[test]
fn test_gather_possible_evictions() {
solana_logger::setup();
let startup = false;
let ref_count = 1;
let pks = (0..=255)
.map(|i| Pubkey::from([i as u8; 32]))
.collect::<Vec<_>>();
let accounts = (0..=255)
.map(|age| {
let one_element_slot_list = vec![(0, 0)];
let one_element_slot_list_entry = Arc::new(AccountMapEntryInner::new(
one_element_slot_list,
ref_count,
AccountMapEntryMeta::default(),
));
one_element_slot_list_entry.set_age(age);
one_element_slot_list_entry
})
.collect::<Vec<_>>();
let both = pks.iter().zip(accounts.iter()).collect::<Vec<_>>();
for current_age in 0..=255 {
for ages_flushing_now in 0..=255 {
let mut possible_evictions = PossibleEvictions::new(1);
possible_evictions.reset(1);
InMemAccountsIndex::<u64, u64>::gather_possible_evictions(
both.iter().cloned(),
&mut possible_evictions,
startup,
current_age,
ages_flushing_now,
false, // true=can_randomly_flush
);
let evictions = possible_evictions.possible_evictions.pop().unwrap();
assert_eq!(
evictions.evictions_age_possible.len(),
1 + ages_flushing_now as usize
);
evictions.evictions_age_possible.iter().for_each(|(_k, v)| {
assert!(
InMemAccountsIndex::<u64, u64>::should_evict_based_on_age(
current_age,
v,
startup,
ages_flushing_now,
),
"current_age: {}, age: {}, ages_flushing_now: {}",
current_age,
v.age(),
ages_flushing_now
);
});
}
}
}
#[test]
fn test_should_evict_from_mem() {
solana_logger::setup();
let bucket = new_for_test::<u64>();
let mut startup = false;
let mut current_age = 0;
let ref_count = 1;
let one_element_slot_list = vec![(0, 0)];
let one_element_slot_list_entry = Arc::new(AccountMapEntryInner::new(
one_element_slot_list,
ref_count,
AccountMapEntryMeta::default(),
));
// exceeded budget
assert!(
bucket
.should_evict_from_mem(
current_age,
&Arc::new(AccountMapEntryInner::new(
vec![],
ref_count,
AccountMapEntryMeta::default()
)),
startup,
false,
true,
0,
)
.0
);
// empty slot list
assert!(
!bucket
.should_evict_from_mem(
current_age,
&Arc::new(AccountMapEntryInner::new(
vec![],
ref_count,
AccountMapEntryMeta::default()
)),
startup,
false,
false,
0,
)
.0
);
// 1 element slot list
assert!(
bucket
.should_evict_from_mem(
current_age,
&one_element_slot_list_entry,
startup,
false,
false,
0,
)
.0
);
// 2 element slot list
assert!(
!bucket
.should_evict_from_mem(
current_age,
&Arc::new(AccountMapEntryInner::new(
vec![(0, 0), (1, 1)],
ref_count,
AccountMapEntryMeta::default()
)),
startup,
false,
false,
0,
)
.0
);
{
let bucket = new_for_test::<f64>();
// 1 element slot list with a CACHED item - f64 acts like cached
assert!(
!bucket
.should_evict_from_mem(
current_age,
&Arc::new(AccountMapEntryInner::new(
vec![(0, 0.0)],
ref_count,
AccountMapEntryMeta::default()
)),
startup,
false,
false,
0,
)
.0
);
}
// 1 element slot list, age is now
assert!(
bucket
.should_evict_from_mem(
current_age,
&one_element_slot_list_entry,
startup,
false,
false,
0,
)
.0
);
// 1 element slot list, but not current age
current_age = 1;
assert!(
!bucket
.should_evict_from_mem(
current_age,
&one_element_slot_list_entry,
startup,
false,
false,
0,
)
.0
);
// 1 element slot list, but at startup and age not current
startup = true;
assert!(
bucket
.should_evict_from_mem(
current_age,
&one_element_slot_list_entry,
startup,
false,
false,
0,
)
.0
);
}
#[test]
fn test_hold_range_in_memory() {
let bucket = new_disk_buckets_for_test::<u64>();
// 0x81 is just some other range
let all = Pubkey::from([0; 32])..=Pubkey::from([0xff; 32]);
let ranges = [
all.clone(),
Pubkey::from([0x81; 32])..=Pubkey::from([0xff; 32]),
];
for range in ranges.clone() {
assert!(bucket.cache_ranges_held.read().unwrap().is_empty());
bucket.hold_range_in_memory(&range, true);
assert_eq!(
bucket.cache_ranges_held.read().unwrap().to_vec(),
vec![range.clone()]
);
{
let evictions_guard = EvictionsGuard::lock(&bucket);
assert!(bucket.add_hold_range_in_memory_if_already_held(&range, &evictions_guard));
bucket.hold_range_in_memory(&range, false);
}
bucket.hold_range_in_memory(&range, false);
assert!(bucket.cache_ranges_held.read().unwrap().is_empty());
bucket.hold_range_in_memory(&range, true);
assert_eq!(
bucket.cache_ranges_held.read().unwrap().to_vec(),
vec![range.clone()]
);
bucket.hold_range_in_memory(&range, true);
assert_eq!(
bucket.cache_ranges_held.read().unwrap().to_vec(),
vec![range.clone(), range.clone()]
);
bucket.hold_range_in_memory(&ranges[0], true);
assert_eq!(
bucket.cache_ranges_held.read().unwrap().to_vec(),
vec![range.clone(), range.clone(), ranges[0].clone()]
);
bucket.hold_range_in_memory(&range, false);
assert_eq!(
bucket.cache_ranges_held.read().unwrap().to_vec(),
vec![range.clone(), ranges[0].clone()]
);
bucket.hold_range_in_memory(&range, false);
assert_eq!(
bucket.cache_ranges_held.read().unwrap().to_vec(),
vec![ranges[0].clone()]
);
bucket.hold_range_in_memory(&ranges[0].clone(), false);
assert!(bucket.cache_ranges_held.read().unwrap().is_empty());
// hold all in mem first
assert!(bucket.cache_ranges_held.read().unwrap().is_empty());
bucket.hold_range_in_memory(&all, true);
let evictions_guard = EvictionsGuard::lock(&bucket);
assert!(bucket.add_hold_range_in_memory_if_already_held(&range, &evictions_guard));
bucket.hold_range_in_memory(&range, false);
bucket.hold_range_in_memory(&all, false);
}
}
#[test]
fn test_age() {
solana_logger::setup();
let test = new_for_test::<u64>();
assert!(test.get_should_age(test.storage.current_age()));
assert_eq!(test.storage.count_buckets_flushed(), 0);
test.set_has_aged(0, true);
assert!(!test.get_should_age(test.storage.current_age()));
assert_eq!(test.storage.count_buckets_flushed(), 1);
// simulate rest of buckets aging
for _ in 1..BINS_FOR_TESTING {
assert!(!test.storage.all_buckets_flushed_at_current_age());
test.storage.bucket_flushed_at_current_age(true);
}
assert!(test.storage.all_buckets_flushed_at_current_age());
// advance age
test.storage.increment_age();
assert_eq!(test.storage.current_age(), 1);
assert!(!test.storage.all_buckets_flushed_at_current_age());
assert!(test.get_should_age(test.storage.current_age()));
assert_eq!(test.storage.count_buckets_flushed(), 0);
}
#[test]
fn test_update_slot_list_other() {
solana_logger::setup();
let reclaim = UpsertReclaim::PopulateReclaims;
let new_slot = 0;
let info = 1;
let other_value = info + 1;
let at_new_slot = (new_slot, info);
let unique_other_slot = new_slot + 1;
for other_slot in [Some(new_slot), Some(unique_other_slot), None] {
let mut reclaims = Vec::default();
let mut slot_list = Vec::default();
// upserting into empty slot_list, so always addref
assert!(
InMemAccountsIndex::<u64, u64>::update_slot_list(
&mut slot_list,
new_slot,
info,
other_slot,
&mut reclaims,
reclaim
),
"other_slot: {other_slot:?}"
);
assert_eq!(slot_list, vec![at_new_slot]);
assert!(reclaims.is_empty());
}
// replace other
let mut slot_list = vec![(unique_other_slot, other_value)];
let expected_reclaims = slot_list.clone();
let other_slot = Some(unique_other_slot);
let mut reclaims = Vec::default();
assert!(
// upserting into slot_list that does NOT contain an entry at 'new_slot'
// but, it DOES contain an entry at other_slot, so we do NOT add-ref. The assumption is that 'other_slot' is going away
// and that the previously held add-ref is now used by 'new_slot'
!InMemAccountsIndex::<u64, u64>::update_slot_list(
&mut slot_list,
new_slot,
info,
other_slot,
&mut reclaims,
reclaim
),
"other_slot: {other_slot:?}"
);
assert_eq!(slot_list, vec![at_new_slot]);
assert_eq!(reclaims, expected_reclaims);
// replace other and new_slot
let mut slot_list = vec![(unique_other_slot, other_value), (new_slot, other_value)];
let expected_reclaims = slot_list.clone();
let other_slot = Some(unique_other_slot);
// upserting into slot_list that already contain an entry at 'new-slot', so do NOT addref
let mut reclaims = Vec::default();
assert!(
!InMemAccountsIndex::<u64, u64>::update_slot_list(
&mut slot_list,
new_slot,
info,
other_slot,
&mut reclaims,
reclaim
),
"other_slot: {other_slot:?}"
);
assert_eq!(slot_list, vec![at_new_slot]);
assert_eq!(
reclaims,
expected_reclaims.into_iter().rev().collect::<Vec<_>>()
);
// nothing will exist at this slot
let missing_other_slot = unique_other_slot + 1;
let ignored_slot = 10; // bigger than is used elsewhere in the test
let ignored_value = info + 10;
let mut possible_initial_slot_list_contents;
// build a list of possible contents in the slot_list prior to calling 'update_slot_list'
{
// up to 3 ignored slot account_info (ignored means not 'new_slot', not 'other_slot', but different slot #s which could exist in the slot_list initially)
possible_initial_slot_list_contents = (0..3)
.map(|i| (ignored_slot + i, ignored_value + i))
.collect::<Vec<_>>();
// account_info that already exists in the slot_list AT 'new_slot'
possible_initial_slot_list_contents.push(at_new_slot);
// account_info that already exists in the slot_list AT 'other_slot'
possible_initial_slot_list_contents.push((unique_other_slot, other_value));
}
/*
* loop over all possible permutations of 'possible_initial_slot_list_contents'
* some examples:
* []
* [other]
* [other, new_slot]
* [new_slot, other]
* [dummy0, new_slot, dummy1, other] (and all permutation of this order)
* [other, dummy1, new_slot] (and all permutation of this order)
* ...
* [dummy0, new_slot, dummy1, other_slot, dummy2] (and all permutation of this order)
*/
let mut attempts = 0;
// loop over each initial size of 'slot_list'
for initial_slot_list_len in 0..=possible_initial_slot_list_contents.len() {
// loop over every permutation of possible_initial_slot_list_contents within a list of len 'initial_slot_list_len'
for content_source_indexes in
(0..possible_initial_slot_list_contents.len()).permutations(initial_slot_list_len)
{
// loop over each possible parameter for 'other_slot'
for other_slot in [
Some(new_slot),
Some(unique_other_slot),
Some(missing_other_slot),
None,
] {
attempts += 1;
// initialize slot_list prior to call to 'InMemAccountsIndex::update_slot_list'
// by inserting each possible entry at each possible position
let mut slot_list = content_source_indexes
.iter()
.map(|i| possible_initial_slot_list_contents[*i])
.collect::<Vec<_>>();
let mut expected = slot_list.clone();
let original = slot_list.clone();
let mut reclaims = Vec::default();
let result = InMemAccountsIndex::<u64, u64>::update_slot_list(
&mut slot_list,
new_slot,
info,
other_slot,
&mut reclaims,
reclaim,
);
// calculate expected results
let mut expected_reclaims = Vec::default();
// addref iff the slot_list did NOT previously contain an entry at 'new_slot' and it also did not contain an entry at 'other_slot'
let expected_result = !expected
.iter()
.any(|(slot, _info)| slot == &new_slot || Some(*slot) == other_slot);
{
// this is the logical equivalent of 'InMemAccountsIndex::update_slot_list', but slower (and ignoring addref)
expected.retain(|(slot, info)| {
let retain = slot != &new_slot && Some(*slot) != other_slot;
if !retain {
expected_reclaims.push((*slot, *info));
}
retain
});
expected.push((new_slot, info));
}
assert_eq!(
expected_result, result,
"return value different. other: {other_slot:?}, {expected:?}, {slot_list:?}, original: {original:?}"
);
// sort for easy comparison
expected_reclaims.sort_unstable();
reclaims.sort_unstable();
assert_eq!(
expected_reclaims, reclaims,
"reclaims different. other: {other_slot:?}, {expected:?}, {slot_list:?}, original: {original:?}"
);
// sort for easy comparison
slot_list.sort_unstable();
expected.sort_unstable();
assert_eq!(
slot_list, expected,
"slot_list different. other: {other_slot:?}, {expected:?}, {slot_list:?}, original: {original:?}"
);
}
}
}
assert_eq!(attempts, 1304); // complicated permutations, so make sure we ran the right #
}
#[test]
fn test_flush_guard() {
let flushing_active = AtomicBool::new(false);
{
let flush_guard = FlushGuard::lock(&flushing_active);
assert!(flush_guard.is_some());
assert!(flushing_active.load(Ordering::Acquire));
{
// Trying to lock the FlushGuard again will not succeed.
let flush_guard2 = FlushGuard::lock(&flushing_active);
assert!(flush_guard2.is_none());
}
// The `flushing_active` flag will remain true, even after `flush_guard2` goes out of
// scope (and is dropped). This ensures `lock()` and `drop()` work harmoniously.
assert!(flushing_active.load(Ordering::Acquire));
}
// After the FlushGuard is dropped, the flag will be cleared.
assert!(!flushing_active.load(Ordering::Acquire));
}
#[test]
fn test_remove_if_slot_list_empty_entry() {
let key = solana_sdk::pubkey::new_rand();
let unknown_key = solana_sdk::pubkey::new_rand();
let test = new_for_test::<u64>();
let mut map = test.map_internal.write().unwrap();
{
// item is NOT in index at all, still return true from remove_if_slot_list_empty_entry
// make sure not initially in index
let entry = map.entry(unknown_key);
assert_matches!(entry, Entry::Vacant(_));
let entry = map.entry(unknown_key);
assert!(test.remove_if_slot_list_empty_entry(entry));
// make sure still not in index
let entry = map.entry(unknown_key);
assert_matches!(entry, Entry::Vacant(_));
}
{
// add an entry with an empty slot list
let val = Arc::new(AccountMapEntryInner::<u64>::default());
map.insert(key, val);
let entry = map.entry(key);
assert_matches!(entry, Entry::Occupied(_));
// should have removed it since it had an empty slot list
assert!(test.remove_if_slot_list_empty_entry(entry));
let entry = map.entry(key);
assert_matches!(entry, Entry::Vacant(_));
// return true - item is not in index at all now
assert!(test.remove_if_slot_list_empty_entry(entry));
}
{
// add an entry with a NON empty slot list - it will NOT get removed
let val = Arc::new(AccountMapEntryInner::<u64>::default());
val.slot_list.write().unwrap().push((1, 1));
map.insert(key, val);
// does NOT remove it since it has a non-empty slot list
let entry = map.entry(key);
assert!(!test.remove_if_slot_list_empty_entry(entry));
let entry = map.entry(key);
assert_matches!(entry, Entry::Occupied(_));
}
}
#[test]
fn test_lock_and_update_slot_list() {
let test = AccountMapEntryInner::<u64>::default();
let info = 65;
let mut reclaims = Vec::default();
// first upsert, should increase
let len = InMemAccountsIndex::<u64, u64>::lock_and_update_slot_list(
&test,
(1, info),
None,
&mut reclaims,
UpsertReclaim::IgnoreReclaims,
);
assert_eq!(test.slot_list.read().unwrap().len(), len);
assert_eq!(len, 1);
// update to different slot, should increase
let len = InMemAccountsIndex::<u64, u64>::lock_and_update_slot_list(
&test,
(2, info),
None,
&mut reclaims,
UpsertReclaim::IgnoreReclaims,
);
assert_eq!(test.slot_list.read().unwrap().len(), len);
assert_eq!(len, 2);
// update to same slot, should not increase
let len = InMemAccountsIndex::<u64, u64>::lock_and_update_slot_list(
&test,
(2, info),
None,
&mut reclaims,
UpsertReclaim::IgnoreReclaims,
);
assert_eq!(test.slot_list.read().unwrap().len(), len);
assert_eq!(len, 2);
}
}