Loading...
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 | /*
* GPL HEADER START
*
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 only,
* as published by the Free Software Foundation.
*
* This program 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 version 2 for more details (a copy is included
* in the LICENSE file that accompanied this code).
*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
* http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
*
* Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
* CA 95054 USA or visit www.sun.com if you need additional information or
* have any questions.
*
* GPL HEADER END
*/
/*
* Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
* Copyright (c) 2011, 2012, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
* Lustre is a trademark of Sun Microsystems, Inc.
*
* libcfs/libcfs/hash.c
*
* Implement a hash class for hash process in lustre system.
*
* Author: YuZhangyong <yzy@clusterfs.com>
*
* 2008-08-15: Brian Behlendorf <behlendorf1@llnl.gov>
* - Simplified API and improved documentation
* - Added per-hash feature flags:
* * CFS_HASH_DEBUG additional validation
* * CFS_HASH_REHASH dynamic rehashing
* - Added per-hash statistics
* - General performance enhancements
*
* 2009-07-31: Liang Zhen <zhen.liang@sun.com>
* - move all stuff to libcfs
* - don't allow cur_bits != max_bits without setting of CFS_HASH_REHASH
* - ignore hs_rwlock if without CFS_HASH_REHASH setting
* - buckets are allocated one by one(instead of contiguous memory),
* to avoid unnecessary cacheline conflict
*
* 2010-03-01: Liang Zhen <zhen.liang@sun.com>
* - "bucket" is a group of hlist_head now, user can specify bucket size
* by bkt_bits of cfs_hash_create(), all hlist_heads in a bucket share
* one lock for reducing memory overhead.
*
* - support lockless hash, caller will take care of locks:
* avoid lock overhead for hash tables that are already protected
* by locking in the caller for another reason
*
* - support both spin_lock/rwlock for bucket:
* overhead of spinlock contention is lower than read/write
* contention of rwlock, so using spinlock to serialize operations on
* bucket is more reasonable for those frequently changed hash tables
*
* - support one-single lock mode:
* one lock to protect all hash operations to avoid overhead of
* multiple locks if hash table is always small
*
* - removed a lot of unnecessary addref & decref on hash element:
* addref & decref are atomic operations in many use-cases which
* are expensive.
*
* - support non-blocking cfs_hash_add() and cfs_hash_findadd():
* some lustre use-cases require these functions to be strictly
* non-blocking, we need to schedule required rehash on a different
* thread on those cases.
*
* - safer rehash on large hash table
* In old implementation, rehash function will exclusively lock the
* hash table and finish rehash in one batch, it's dangerous on SMP
* system because rehash millions of elements could take long time.
* New implemented rehash can release lock and relax CPU in middle
* of rehash, it's safe for another thread to search/change on the
* hash table even it's in rehasing.
*
* - support two different refcount modes
* . hash table has refcount on element
* . hash table doesn't change refcount on adding/removing element
*
* - support long name hash table (for param-tree)
*
* - fix a bug for cfs_hash_rehash_key:
* in old implementation, cfs_hash_rehash_key could screw up the
* hash-table because @key is overwritten without any protection.
* Now we need user to define hs_keycpy for those rehash enabled
* hash tables, cfs_hash_rehash_key will overwrite hash-key
* inside lock by calling hs_keycpy.
*
* - better hash iteration:
* Now we support both locked iteration & lockless iteration of hash
* table. Also, user can break the iteration by return 1 in callback.
*/
#include "../../include/linux/libcfs/libcfs.h"
#include <linux/seq_file.h>
#if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1
static unsigned int warn_on_depth = 8;
module_param(warn_on_depth, uint, 0644);
MODULE_PARM_DESC(warn_on_depth, "warning when hash depth is high.");
#endif
struct cfs_wi_sched *cfs_sched_rehash;
static inline void
cfs_hash_nl_lock(union cfs_hash_lock *lock, int exclusive) {}
static inline void
cfs_hash_nl_unlock(union cfs_hash_lock *lock, int exclusive) {}
static inline void
cfs_hash_spin_lock(union cfs_hash_lock *lock, int exclusive)
{
spin_lock(&lock->spin);
}
static inline void
cfs_hash_spin_unlock(union cfs_hash_lock *lock, int exclusive)
{
spin_unlock(&lock->spin);
}
static inline void
cfs_hash_rw_lock(union cfs_hash_lock *lock, int exclusive)
{
if (!exclusive)
read_lock(&lock->rw);
else
write_lock(&lock->rw);
}
static inline void
cfs_hash_rw_unlock(union cfs_hash_lock *lock, int exclusive)
{
if (!exclusive)
read_unlock(&lock->rw);
else
write_unlock(&lock->rw);
}
/** No lock hash */
static cfs_hash_lock_ops_t cfs_hash_nl_lops = {
.hs_lock = cfs_hash_nl_lock,
.hs_unlock = cfs_hash_nl_unlock,
.hs_bkt_lock = cfs_hash_nl_lock,
.hs_bkt_unlock = cfs_hash_nl_unlock,
};
/** no bucket lock, one spinlock to protect everything */
static cfs_hash_lock_ops_t cfs_hash_nbl_lops = {
.hs_lock = cfs_hash_spin_lock,
.hs_unlock = cfs_hash_spin_unlock,
.hs_bkt_lock = cfs_hash_nl_lock,
.hs_bkt_unlock = cfs_hash_nl_unlock,
};
/** spin bucket lock, rehash is enabled */
static cfs_hash_lock_ops_t cfs_hash_bkt_spin_lops = {
.hs_lock = cfs_hash_rw_lock,
.hs_unlock = cfs_hash_rw_unlock,
.hs_bkt_lock = cfs_hash_spin_lock,
.hs_bkt_unlock = cfs_hash_spin_unlock,
};
/** rw bucket lock, rehash is enabled */
static cfs_hash_lock_ops_t cfs_hash_bkt_rw_lops = {
.hs_lock = cfs_hash_rw_lock,
.hs_unlock = cfs_hash_rw_unlock,
.hs_bkt_lock = cfs_hash_rw_lock,
.hs_bkt_unlock = cfs_hash_rw_unlock,
};
/** spin bucket lock, rehash is disabled */
static cfs_hash_lock_ops_t cfs_hash_nr_bkt_spin_lops = {
.hs_lock = cfs_hash_nl_lock,
.hs_unlock = cfs_hash_nl_unlock,
.hs_bkt_lock = cfs_hash_spin_lock,
.hs_bkt_unlock = cfs_hash_spin_unlock,
};
/** rw bucket lock, rehash is disabled */
static cfs_hash_lock_ops_t cfs_hash_nr_bkt_rw_lops = {
.hs_lock = cfs_hash_nl_lock,
.hs_unlock = cfs_hash_nl_unlock,
.hs_bkt_lock = cfs_hash_rw_lock,
.hs_bkt_unlock = cfs_hash_rw_unlock,
};
static void
cfs_hash_lock_setup(struct cfs_hash *hs)
{
if (cfs_hash_with_no_lock(hs)) {
hs->hs_lops = &cfs_hash_nl_lops;
} else if (cfs_hash_with_no_bktlock(hs)) {
hs->hs_lops = &cfs_hash_nbl_lops;
spin_lock_init(&hs->hs_lock.spin);
} else if (cfs_hash_with_rehash(hs)) {
rwlock_init(&hs->hs_lock.rw);
if (cfs_hash_with_rw_bktlock(hs))
hs->hs_lops = &cfs_hash_bkt_rw_lops;
else if (cfs_hash_with_spin_bktlock(hs))
hs->hs_lops = &cfs_hash_bkt_spin_lops;
else
LBUG();
} else {
if (cfs_hash_with_rw_bktlock(hs))
hs->hs_lops = &cfs_hash_nr_bkt_rw_lops;
else if (cfs_hash_with_spin_bktlock(hs))
hs->hs_lops = &cfs_hash_nr_bkt_spin_lops;
else
LBUG();
}
}
/**
* Simple hash head without depth tracking
* new element is always added to head of hlist
*/
typedef struct {
struct hlist_head hh_head; /**< entries list */
} cfs_hash_head_t;
static int
cfs_hash_hh_hhead_size(struct cfs_hash *hs)
{
return sizeof(cfs_hash_head_t);
}
static struct hlist_head *
cfs_hash_hh_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
{
cfs_hash_head_t *head = (cfs_hash_head_t *)&bd->bd_bucket->hsb_head[0];
return &head[bd->bd_offset].hh_head;
}
static int
cfs_hash_hh_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
hlist_add_head(hnode, cfs_hash_hh_hhead(hs, bd));
return -1; /* unknown depth */
}
static int
cfs_hash_hh_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
hlist_del_init(hnode);
return -1; /* unknown depth */
}
/**
* Simple hash head with depth tracking
* new element is always added to head of hlist
*/
typedef struct {
struct hlist_head hd_head; /**< entries list */
unsigned int hd_depth; /**< list length */
} cfs_hash_head_dep_t;
static int
cfs_hash_hd_hhead_size(struct cfs_hash *hs)
{
return sizeof(cfs_hash_head_dep_t);
}
static struct hlist_head *
cfs_hash_hd_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
{
cfs_hash_head_dep_t *head;
head = (cfs_hash_head_dep_t *)&bd->bd_bucket->hsb_head[0];
return &head[bd->bd_offset].hd_head;
}
static int
cfs_hash_hd_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
cfs_hash_head_dep_t *hh = container_of(cfs_hash_hd_hhead(hs, bd),
cfs_hash_head_dep_t, hd_head);
hlist_add_head(hnode, &hh->hd_head);
return ++hh->hd_depth;
}
static int
cfs_hash_hd_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
cfs_hash_head_dep_t *hh = container_of(cfs_hash_hd_hhead(hs, bd),
cfs_hash_head_dep_t, hd_head);
hlist_del_init(hnode);
return --hh->hd_depth;
}
/**
* double links hash head without depth tracking
* new element is always added to tail of hlist
*/
typedef struct {
struct hlist_head dh_head; /**< entries list */
struct hlist_node *dh_tail; /**< the last entry */
} cfs_hash_dhead_t;
static int
cfs_hash_dh_hhead_size(struct cfs_hash *hs)
{
return sizeof(cfs_hash_dhead_t);
}
static struct hlist_head *
cfs_hash_dh_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
{
cfs_hash_dhead_t *head;
head = (cfs_hash_dhead_t *)&bd->bd_bucket->hsb_head[0];
return &head[bd->bd_offset].dh_head;
}
static int
cfs_hash_dh_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
cfs_hash_dhead_t *dh = container_of(cfs_hash_dh_hhead(hs, bd),
cfs_hash_dhead_t, dh_head);
if (dh->dh_tail != NULL) /* not empty */
hlist_add_behind(hnode, dh->dh_tail);
else /* empty list */
hlist_add_head(hnode, &dh->dh_head);
dh->dh_tail = hnode;
return -1; /* unknown depth */
}
static int
cfs_hash_dh_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnd)
{
cfs_hash_dhead_t *dh = container_of(cfs_hash_dh_hhead(hs, bd),
cfs_hash_dhead_t, dh_head);
if (hnd->next == NULL) { /* it's the tail */
dh->dh_tail = (hnd->pprev == &dh->dh_head.first) ? NULL :
container_of(hnd->pprev, struct hlist_node, next);
}
hlist_del_init(hnd);
return -1; /* unknown depth */
}
/**
* double links hash head with depth tracking
* new element is always added to tail of hlist
*/
typedef struct {
struct hlist_head dd_head; /**< entries list */
struct hlist_node *dd_tail; /**< the last entry */
unsigned int dd_depth; /**< list length */
} cfs_hash_dhead_dep_t;
static int
cfs_hash_dd_hhead_size(struct cfs_hash *hs)
{
return sizeof(cfs_hash_dhead_dep_t);
}
static struct hlist_head *
cfs_hash_dd_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
{
cfs_hash_dhead_dep_t *head;
head = (cfs_hash_dhead_dep_t *)&bd->bd_bucket->hsb_head[0];
return &head[bd->bd_offset].dd_head;
}
static int
cfs_hash_dd_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
cfs_hash_dhead_dep_t *dh = container_of(cfs_hash_dd_hhead(hs, bd),
cfs_hash_dhead_dep_t, dd_head);
if (dh->dd_tail != NULL) /* not empty */
hlist_add_behind(hnode, dh->dd_tail);
else /* empty list */
hlist_add_head(hnode, &dh->dd_head);
dh->dd_tail = hnode;
return ++dh->dd_depth;
}
static int
cfs_hash_dd_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnd)
{
cfs_hash_dhead_dep_t *dh = container_of(cfs_hash_dd_hhead(hs, bd),
cfs_hash_dhead_dep_t, dd_head);
if (hnd->next == NULL) { /* it's the tail */
dh->dd_tail = (hnd->pprev == &dh->dd_head.first) ? NULL :
container_of(hnd->pprev, struct hlist_node, next);
}
hlist_del_init(hnd);
return --dh->dd_depth;
}
static cfs_hash_hlist_ops_t cfs_hash_hh_hops = {
.hop_hhead = cfs_hash_hh_hhead,
.hop_hhead_size = cfs_hash_hh_hhead_size,
.hop_hnode_add = cfs_hash_hh_hnode_add,
.hop_hnode_del = cfs_hash_hh_hnode_del,
};
static cfs_hash_hlist_ops_t cfs_hash_hd_hops = {
.hop_hhead = cfs_hash_hd_hhead,
.hop_hhead_size = cfs_hash_hd_hhead_size,
.hop_hnode_add = cfs_hash_hd_hnode_add,
.hop_hnode_del = cfs_hash_hd_hnode_del,
};
static cfs_hash_hlist_ops_t cfs_hash_dh_hops = {
.hop_hhead = cfs_hash_dh_hhead,
.hop_hhead_size = cfs_hash_dh_hhead_size,
.hop_hnode_add = cfs_hash_dh_hnode_add,
.hop_hnode_del = cfs_hash_dh_hnode_del,
};
static cfs_hash_hlist_ops_t cfs_hash_dd_hops = {
.hop_hhead = cfs_hash_dd_hhead,
.hop_hhead_size = cfs_hash_dd_hhead_size,
.hop_hnode_add = cfs_hash_dd_hnode_add,
.hop_hnode_del = cfs_hash_dd_hnode_del,
};
static void
cfs_hash_hlist_setup(struct cfs_hash *hs)
{
if (cfs_hash_with_add_tail(hs)) {
hs->hs_hops = cfs_hash_with_depth(hs) ?
&cfs_hash_dd_hops : &cfs_hash_dh_hops;
} else {
hs->hs_hops = cfs_hash_with_depth(hs) ?
&cfs_hash_hd_hops : &cfs_hash_hh_hops;
}
}
static void
cfs_hash_bd_from_key(struct cfs_hash *hs, struct cfs_hash_bucket **bkts,
unsigned int bits, const void *key, struct cfs_hash_bd *bd)
{
unsigned int index = cfs_hash_id(hs, key, (1U << bits) - 1);
LASSERT(bits == hs->hs_cur_bits || bits == hs->hs_rehash_bits);
bd->bd_bucket = bkts[index & ((1U << (bits - hs->hs_bkt_bits)) - 1)];
bd->bd_offset = index >> (bits - hs->hs_bkt_bits);
}
void
cfs_hash_bd_get(struct cfs_hash *hs, const void *key, struct cfs_hash_bd *bd)
{
/* NB: caller should hold hs->hs_rwlock if REHASH is set */
if (likely(hs->hs_rehash_buckets == NULL)) {
cfs_hash_bd_from_key(hs, hs->hs_buckets,
hs->hs_cur_bits, key, bd);
} else {
LASSERT(hs->hs_rehash_bits != 0);
cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets,
hs->hs_rehash_bits, key, bd);
}
}
EXPORT_SYMBOL(cfs_hash_bd_get);
static inline void
cfs_hash_bd_dep_record(struct cfs_hash *hs, struct cfs_hash_bd *bd, int dep_cur)
{
if (likely(dep_cur <= bd->bd_bucket->hsb_depmax))
return;
bd->bd_bucket->hsb_depmax = dep_cur;
# if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1
if (likely(warn_on_depth == 0 ||
max(warn_on_depth, hs->hs_dep_max) >= dep_cur))
return;
spin_lock(&hs->hs_dep_lock);
hs->hs_dep_max = dep_cur;
hs->hs_dep_bkt = bd->bd_bucket->hsb_index;
hs->hs_dep_off = bd->bd_offset;
hs->hs_dep_bits = hs->hs_cur_bits;
spin_unlock(&hs->hs_dep_lock);
cfs_wi_schedule(cfs_sched_rehash, &hs->hs_dep_wi);
# endif
}
void
cfs_hash_bd_add_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
int rc;
rc = hs->hs_hops->hop_hnode_add(hs, bd, hnode);
cfs_hash_bd_dep_record(hs, bd, rc);
bd->bd_bucket->hsb_version++;
if (unlikely(bd->bd_bucket->hsb_version == 0))
bd->bd_bucket->hsb_version++;
bd->bd_bucket->hsb_count++;
if (cfs_hash_with_counter(hs))
atomic_inc(&hs->hs_count);
if (!cfs_hash_with_no_itemref(hs))
cfs_hash_get(hs, hnode);
}
EXPORT_SYMBOL(cfs_hash_bd_add_locked);
void
cfs_hash_bd_del_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode)
{
hs->hs_hops->hop_hnode_del(hs, bd, hnode);
LASSERT(bd->bd_bucket->hsb_count > 0);
bd->bd_bucket->hsb_count--;
bd->bd_bucket->hsb_version++;
if (unlikely(bd->bd_bucket->hsb_version == 0))
bd->bd_bucket->hsb_version++;
if (cfs_hash_with_counter(hs)) {
LASSERT(atomic_read(&hs->hs_count) > 0);
atomic_dec(&hs->hs_count);
}
if (!cfs_hash_with_no_itemref(hs))
cfs_hash_put_locked(hs, hnode);
}
EXPORT_SYMBOL(cfs_hash_bd_del_locked);
void
cfs_hash_bd_move_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd_old,
struct cfs_hash_bd *bd_new, struct hlist_node *hnode)
{
struct cfs_hash_bucket *obkt = bd_old->bd_bucket;
struct cfs_hash_bucket *nbkt = bd_new->bd_bucket;
int rc;
if (cfs_hash_bd_compare(bd_old, bd_new) == 0)
return;
/* use cfs_hash_bd_hnode_add/del, to avoid atomic & refcount ops
* in cfs_hash_bd_del/add_locked */
hs->hs_hops->hop_hnode_del(hs, bd_old, hnode);
rc = hs->hs_hops->hop_hnode_add(hs, bd_new, hnode);
cfs_hash_bd_dep_record(hs, bd_new, rc);
LASSERT(obkt->hsb_count > 0);
obkt->hsb_count--;
obkt->hsb_version++;
if (unlikely(obkt->hsb_version == 0))
obkt->hsb_version++;
nbkt->hsb_count++;
nbkt->hsb_version++;
if (unlikely(nbkt->hsb_version == 0))
nbkt->hsb_version++;
}
EXPORT_SYMBOL(cfs_hash_bd_move_locked);
enum {
/** always set, for sanity (avoid ZERO intent) */
CFS_HS_LOOKUP_MASK_FIND = 1 << 0,
/** return entry with a ref */
CFS_HS_LOOKUP_MASK_REF = 1 << 1,
/** add entry if not existing */
CFS_HS_LOOKUP_MASK_ADD = 1 << 2,
/** delete entry, ignore other masks */
CFS_HS_LOOKUP_MASK_DEL = 1 << 3,
};
typedef enum cfs_hash_lookup_intent {
/** return item w/o refcount */
CFS_HS_LOOKUP_IT_PEEK = CFS_HS_LOOKUP_MASK_FIND,
/** return item with refcount */
CFS_HS_LOOKUP_IT_FIND = (CFS_HS_LOOKUP_MASK_FIND |
CFS_HS_LOOKUP_MASK_REF),
/** return item w/o refcount if existed, otherwise add */
CFS_HS_LOOKUP_IT_ADD = (CFS_HS_LOOKUP_MASK_FIND |
CFS_HS_LOOKUP_MASK_ADD),
/** return item with refcount if existed, otherwise add */
CFS_HS_LOOKUP_IT_FINDADD = (CFS_HS_LOOKUP_IT_FIND |
CFS_HS_LOOKUP_MASK_ADD),
/** delete if existed */
CFS_HS_LOOKUP_IT_FINDDEL = (CFS_HS_LOOKUP_MASK_FIND |
CFS_HS_LOOKUP_MASK_DEL)
} cfs_hash_lookup_intent_t;
static struct hlist_node *
cfs_hash_bd_lookup_intent(struct cfs_hash *hs, struct cfs_hash_bd *bd,
const void *key, struct hlist_node *hnode,
cfs_hash_lookup_intent_t intent)
{
struct hlist_head *hhead = cfs_hash_bd_hhead(hs, bd);
struct hlist_node *ehnode;
struct hlist_node *match;
int intent_add = (intent & CFS_HS_LOOKUP_MASK_ADD) != 0;
/* with this function, we can avoid a lot of useless refcount ops,
* which are expensive atomic operations most time. */
match = intent_add ? NULL : hnode;
hlist_for_each(ehnode, hhead) {
if (!cfs_hash_keycmp(hs, key, ehnode))
continue;
if (match != NULL && match != ehnode) /* can't match */
continue;
/* match and ... */
if ((intent & CFS_HS_LOOKUP_MASK_DEL) != 0) {
cfs_hash_bd_del_locked(hs, bd, ehnode);
return ehnode;
}
/* caller wants refcount? */
if ((intent & CFS_HS_LOOKUP_MASK_REF) != 0)
cfs_hash_get(hs, ehnode);
return ehnode;
}
/* no match item */
if (!intent_add)
return NULL;
LASSERT(hnode != NULL);
cfs_hash_bd_add_locked(hs, bd, hnode);
return hnode;
}
struct hlist_node *
cfs_hash_bd_lookup_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd, const void *key)
{
return cfs_hash_bd_lookup_intent(hs, bd, key, NULL,
CFS_HS_LOOKUP_IT_FIND);
}
EXPORT_SYMBOL(cfs_hash_bd_lookup_locked);
struct hlist_node *
cfs_hash_bd_peek_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd, const void *key)
{
return cfs_hash_bd_lookup_intent(hs, bd, key, NULL,
CFS_HS_LOOKUP_IT_PEEK);
}
EXPORT_SYMBOL(cfs_hash_bd_peek_locked);
struct hlist_node *
cfs_hash_bd_findadd_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
const void *key, struct hlist_node *hnode,
int noref)
{
return cfs_hash_bd_lookup_intent(hs, bd, key, hnode,
CFS_HS_LOOKUP_IT_ADD |
(!noref * CFS_HS_LOOKUP_MASK_REF));
}
EXPORT_SYMBOL(cfs_hash_bd_findadd_locked);
struct hlist_node *
cfs_hash_bd_finddel_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
const void *key, struct hlist_node *hnode)
{
/* hnode can be NULL, we find the first item with @key */
return cfs_hash_bd_lookup_intent(hs, bd, key, hnode,
CFS_HS_LOOKUP_IT_FINDDEL);
}
EXPORT_SYMBOL(cfs_hash_bd_finddel_locked);
static void
cfs_hash_multi_bd_lock(struct cfs_hash *hs, struct cfs_hash_bd *bds,
unsigned n, int excl)
{
struct cfs_hash_bucket *prev = NULL;
int i;
/**
* bds must be ascendantly ordered by bd->bd_bucket->hsb_index.
* NB: it's possible that several bds point to the same bucket but
* have different bd::bd_offset, so need take care of deadlock.
*/
cfs_hash_for_each_bd(bds, n, i) {
if (prev == bds[i].bd_bucket)
continue;
LASSERT(prev == NULL ||
prev->hsb_index < bds[i].bd_bucket->hsb_index);
cfs_hash_bd_lock(hs, &bds[i], excl);
prev = bds[i].bd_bucket;
}
}
static void
cfs_hash_multi_bd_unlock(struct cfs_hash *hs, struct cfs_hash_bd *bds,
unsigned n, int excl)
{
struct cfs_hash_bucket *prev = NULL;
int i;
cfs_hash_for_each_bd(bds, n, i) {
if (prev != bds[i].bd_bucket) {
cfs_hash_bd_unlock(hs, &bds[i], excl);
prev = bds[i].bd_bucket;
}
}
}
static struct hlist_node *
cfs_hash_multi_bd_lookup_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
unsigned n, const void *key)
{
struct hlist_node *ehnode;
unsigned i;
cfs_hash_for_each_bd(bds, n, i) {
ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key, NULL,
CFS_HS_LOOKUP_IT_FIND);
if (ehnode != NULL)
return ehnode;
}
return NULL;
}
static struct hlist_node *
cfs_hash_multi_bd_findadd_locked(struct cfs_hash *hs,
struct cfs_hash_bd *bds, unsigned n, const void *key,
struct hlist_node *hnode, int noref)
{
struct hlist_node *ehnode;
int intent;
unsigned i;
LASSERT(hnode != NULL);
intent = CFS_HS_LOOKUP_IT_PEEK | (!noref * CFS_HS_LOOKUP_MASK_REF);
cfs_hash_for_each_bd(bds, n, i) {
ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key,
NULL, intent);
if (ehnode != NULL)
return ehnode;
}
if (i == 1) { /* only one bucket */
cfs_hash_bd_add_locked(hs, &bds[0], hnode);
} else {
struct cfs_hash_bd mybd;
cfs_hash_bd_get(hs, key, &mybd);
cfs_hash_bd_add_locked(hs, &mybd, hnode);
}
return hnode;
}
static struct hlist_node *
cfs_hash_multi_bd_finddel_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
unsigned n, const void *key,
struct hlist_node *hnode)
{
struct hlist_node *ehnode;
unsigned i;
cfs_hash_for_each_bd(bds, n, i) {
ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key, hnode,
CFS_HS_LOOKUP_IT_FINDDEL);
if (ehnode != NULL)
return ehnode;
}
return NULL;
}
static void
cfs_hash_bd_order(struct cfs_hash_bd *bd1, struct cfs_hash_bd *bd2)
{
int rc;
if (bd2->bd_bucket == NULL)
return;
if (bd1->bd_bucket == NULL) {
*bd1 = *bd2;
bd2->bd_bucket = NULL;
return;
}
rc = cfs_hash_bd_compare(bd1, bd2);
if (rc == 0) {
bd2->bd_bucket = NULL;
} else if (rc > 0) { /* swab bd1 and bd2 */
struct cfs_hash_bd tmp;
tmp = *bd2;
*bd2 = *bd1;
*bd1 = tmp;
}
}
void
cfs_hash_dual_bd_get(struct cfs_hash *hs, const void *key, struct cfs_hash_bd *bds)
{
/* NB: caller should hold hs_lock.rw if REHASH is set */
cfs_hash_bd_from_key(hs, hs->hs_buckets,
hs->hs_cur_bits, key, &bds[0]);
if (likely(hs->hs_rehash_buckets == NULL)) {
/* no rehash or not rehashing */
bds[1].bd_bucket = NULL;
return;
}
LASSERT(hs->hs_rehash_bits != 0);
cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets,
hs->hs_rehash_bits, key, &bds[1]);
cfs_hash_bd_order(&bds[0], &bds[1]);
}
EXPORT_SYMBOL(cfs_hash_dual_bd_get);
void
cfs_hash_dual_bd_lock(struct cfs_hash *hs, struct cfs_hash_bd *bds, int excl)
{
cfs_hash_multi_bd_lock(hs, bds, 2, excl);
}
EXPORT_SYMBOL(cfs_hash_dual_bd_lock);
void
cfs_hash_dual_bd_unlock(struct cfs_hash *hs, struct cfs_hash_bd *bds, int excl)
{
cfs_hash_multi_bd_unlock(hs, bds, 2, excl);
}
EXPORT_SYMBOL(cfs_hash_dual_bd_unlock);
struct hlist_node *
cfs_hash_dual_bd_lookup_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
const void *key)
{
return cfs_hash_multi_bd_lookup_locked(hs, bds, 2, key);
}
EXPORT_SYMBOL(cfs_hash_dual_bd_lookup_locked);
struct hlist_node *
cfs_hash_dual_bd_findadd_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
const void *key, struct hlist_node *hnode,
int noref)
{
return cfs_hash_multi_bd_findadd_locked(hs, bds, 2, key,
hnode, noref);
}
EXPORT_SYMBOL(cfs_hash_dual_bd_findadd_locked);
struct hlist_node *
cfs_hash_dual_bd_finddel_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
const void *key, struct hlist_node *hnode)
{
return cfs_hash_multi_bd_finddel_locked(hs, bds, 2, key, hnode);
}
EXPORT_SYMBOL(cfs_hash_dual_bd_finddel_locked);
static void
cfs_hash_buckets_free(struct cfs_hash_bucket **buckets,
int bkt_size, int prev_size, int size)
{
int i;
for (i = prev_size; i < size; i++) {
if (buckets[i] != NULL)
LIBCFS_FREE(buckets[i], bkt_size);
}
LIBCFS_FREE(buckets, sizeof(buckets[0]) * size);
}
/*
* Create or grow bucket memory. Return old_buckets if no allocation was
* needed, the newly allocated buckets if allocation was needed and
* successful, and NULL on error.
*/
static struct cfs_hash_bucket **
cfs_hash_buckets_realloc(struct cfs_hash *hs, struct cfs_hash_bucket **old_bkts,
unsigned int old_size, unsigned int new_size)
{
struct cfs_hash_bucket **new_bkts;
int i;
LASSERT(old_size == 0 || old_bkts != NULL);
if (old_bkts != NULL && old_size == new_size)
return old_bkts;
LIBCFS_ALLOC(new_bkts, sizeof(new_bkts[0]) * new_size);
if (new_bkts == NULL)
return NULL;
if (old_bkts != NULL) {
memcpy(new_bkts, old_bkts,
min(old_size, new_size) * sizeof(*old_bkts));
}
for (i = old_size; i < new_size; i++) {
struct hlist_head *hhead;
struct cfs_hash_bd bd;
LIBCFS_ALLOC(new_bkts[i], cfs_hash_bkt_size(hs));
if (new_bkts[i] == NULL) {
cfs_hash_buckets_free(new_bkts, cfs_hash_bkt_size(hs),
old_size, new_size);
return NULL;
}
new_bkts[i]->hsb_index = i;
new_bkts[i]->hsb_version = 1; /* shouldn't be zero */
new_bkts[i]->hsb_depmax = -1; /* unknown */
bd.bd_bucket = new_bkts[i];
cfs_hash_bd_for_each_hlist(hs, &bd, hhead)
INIT_HLIST_HEAD(hhead);
if (cfs_hash_with_no_lock(hs) ||
cfs_hash_with_no_bktlock(hs))
continue;
if (cfs_hash_with_rw_bktlock(hs))
rwlock_init(&new_bkts[i]->hsb_lock.rw);
else if (cfs_hash_with_spin_bktlock(hs))
spin_lock_init(&new_bkts[i]->hsb_lock.spin);
else
LBUG(); /* invalid use-case */
}
return new_bkts;
}
/**
* Initialize new libcfs hash, where:
* @name - Descriptive hash name
* @cur_bits - Initial hash table size, in bits
* @max_bits - Maximum allowed hash table resize, in bits
* @ops - Registered hash table operations
* @flags - CFS_HASH_REHASH enable synamic hash resizing
* - CFS_HASH_SORT enable chained hash sort
*/
static int cfs_hash_rehash_worker(cfs_workitem_t *wi);
#if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1
static int cfs_hash_dep_print(cfs_workitem_t *wi)
{
struct cfs_hash *hs = container_of(wi, struct cfs_hash, hs_dep_wi);
int dep;
int bkt;
int off;
int bits;
spin_lock(&hs->hs_dep_lock);
dep = hs->hs_dep_max;
bkt = hs->hs_dep_bkt;
off = hs->hs_dep_off;
bits = hs->hs_dep_bits;
spin_unlock(&hs->hs_dep_lock);
LCONSOLE_WARN("#### HASH %s (bits: %d): max depth %d at bucket %d/%d\n",
hs->hs_name, bits, dep, bkt, off);
spin_lock(&hs->hs_dep_lock);
hs->hs_dep_bits = 0; /* mark as workitem done */
spin_unlock(&hs->hs_dep_lock);
return 0;
}
static void cfs_hash_depth_wi_init(struct cfs_hash *hs)
{
spin_lock_init(&hs->hs_dep_lock);
cfs_wi_init(&hs->hs_dep_wi, hs, cfs_hash_dep_print);
}
static void cfs_hash_depth_wi_cancel(struct cfs_hash *hs)
{
if (cfs_wi_deschedule(cfs_sched_rehash, &hs->hs_dep_wi))
return;
spin_lock(&hs->hs_dep_lock);
while (hs->hs_dep_bits != 0) {
spin_unlock(&hs->hs_dep_lock);
cond_resched();
spin_lock(&hs->hs_dep_lock);
}
spin_unlock(&hs->hs_dep_lock);
}
#else /* CFS_HASH_DEBUG_LEVEL < CFS_HASH_DEBUG_1 */
static inline void cfs_hash_depth_wi_init(struct cfs_hash *hs) {}
static inline void cfs_hash_depth_wi_cancel(struct cfs_hash *hs) {}
#endif /* CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1 */
struct cfs_hash *
cfs_hash_create(char *name, unsigned cur_bits, unsigned max_bits,
unsigned bkt_bits, unsigned extra_bytes,
unsigned min_theta, unsigned max_theta,
cfs_hash_ops_t *ops, unsigned flags)
{
struct cfs_hash *hs;
int len;
CLASSERT(CFS_HASH_THETA_BITS < 15);
LASSERT(name != NULL);
LASSERT(ops != NULL);
LASSERT(ops->hs_key);
LASSERT(ops->hs_hash);
LASSERT(ops->hs_object);
LASSERT(ops->hs_keycmp);
LASSERT(ops->hs_get != NULL);
LASSERT(ops->hs_put_locked != NULL);
if ((flags & CFS_HASH_REHASH) != 0)
flags |= CFS_HASH_COUNTER; /* must have counter */
LASSERT(cur_bits > 0);
LASSERT(cur_bits >= bkt_bits);
LASSERT(max_bits >= cur_bits && max_bits < 31);
LASSERT(ergo((flags & CFS_HASH_REHASH) == 0, cur_bits == max_bits));
LASSERT(ergo((flags & CFS_HASH_REHASH) != 0,
(flags & CFS_HASH_NO_LOCK) == 0));
LASSERT(ergo((flags & CFS_HASH_REHASH_KEY) != 0,
ops->hs_keycpy != NULL));
len = (flags & CFS_HASH_BIGNAME) == 0 ?
CFS_HASH_NAME_LEN : CFS_HASH_BIGNAME_LEN;
LIBCFS_ALLOC(hs, offsetof(struct cfs_hash, hs_name[len]));
if (hs == NULL)
return NULL;
strncpy(hs->hs_name, name, len);
hs->hs_name[len - 1] = '\0';
hs->hs_flags = flags;
atomic_set(&hs->hs_refcount, 1);
atomic_set(&hs->hs_count, 0);
cfs_hash_lock_setup(hs);
cfs_hash_hlist_setup(hs);
hs->hs_cur_bits = (__u8)cur_bits;
hs->hs_min_bits = (__u8)cur_bits;
hs->hs_max_bits = (__u8)max_bits;
hs->hs_bkt_bits = (__u8)bkt_bits;
hs->hs_ops = ops;
hs->hs_extra_bytes = extra_bytes;
hs->hs_rehash_bits = 0;
cfs_wi_init(&hs->hs_rehash_wi, hs, cfs_hash_rehash_worker);
cfs_hash_depth_wi_init(hs);
if (cfs_hash_with_rehash(hs))
__cfs_hash_set_theta(hs, min_theta, max_theta);
hs->hs_buckets = cfs_hash_buckets_realloc(hs, NULL, 0,
CFS_HASH_NBKT(hs));
if (hs->hs_buckets != NULL)
return hs;
LIBCFS_FREE(hs, offsetof(struct cfs_hash, hs_name[len]));
return NULL;
}
EXPORT_SYMBOL(cfs_hash_create);
/**
* Cleanup libcfs hash @hs.
*/
static void
cfs_hash_destroy(struct cfs_hash *hs)
{
struct hlist_node *hnode;
struct hlist_node *pos;
struct cfs_hash_bd bd;
int i;
LASSERT(hs != NULL);
LASSERT(!cfs_hash_is_exiting(hs) &&
!cfs_hash_is_iterating(hs));
/**
* prohibit further rehashes, don't need any lock because
* I'm the only (last) one can change it.
*/
hs->hs_exiting = 1;
if (cfs_hash_with_rehash(hs))
cfs_hash_rehash_cancel(hs);
cfs_hash_depth_wi_cancel(hs);
/* rehash should be done/canceled */
LASSERT(hs->hs_buckets != NULL &&
hs->hs_rehash_buckets == NULL);
cfs_hash_for_each_bucket(hs, &bd, i) {
struct hlist_head *hhead;
LASSERT(bd.bd_bucket != NULL);
/* no need to take this lock, just for consistent code */
cfs_hash_bd_lock(hs, &bd, 1);
cfs_hash_bd_for_each_hlist(hs, &bd, hhead) {
hlist_for_each_safe(hnode, pos, hhead) {
LASSERTF(!cfs_hash_with_assert_empty(hs),
"hash %s bucket %u(%u) is not "
" empty: %u items left\n",
hs->hs_name, bd.bd_bucket->hsb_index,
bd.bd_offset, bd.bd_bucket->hsb_count);
/* can't assert key valicate, because we
* can interrupt rehash */
cfs_hash_bd_del_locked(hs, &bd, hnode);
cfs_hash_exit(hs, hnode);
}
}
LASSERT(bd.bd_bucket->hsb_count == 0);
cfs_hash_bd_unlock(hs, &bd, 1);
cond_resched();
}
LASSERT(atomic_read(&hs->hs_count) == 0);
cfs_hash_buckets_free(hs->hs_buckets, cfs_hash_bkt_size(hs),
0, CFS_HASH_NBKT(hs));
i = cfs_hash_with_bigname(hs) ?
CFS_HASH_BIGNAME_LEN : CFS_HASH_NAME_LEN;
LIBCFS_FREE(hs, offsetof(struct cfs_hash, hs_name[i]));
}
struct cfs_hash *cfs_hash_getref(struct cfs_hash *hs)
{
if (atomic_inc_not_zero(&hs->hs_refcount))
return hs;
return NULL;
}
EXPORT_SYMBOL(cfs_hash_getref);
void cfs_hash_putref(struct cfs_hash *hs)
{
if (atomic_dec_and_test(&hs->hs_refcount))
cfs_hash_destroy(hs);
}
EXPORT_SYMBOL(cfs_hash_putref);
static inline int
cfs_hash_rehash_bits(struct cfs_hash *hs)
{
if (cfs_hash_with_no_lock(hs) ||
!cfs_hash_with_rehash(hs))
return -EOPNOTSUPP;
if (unlikely(cfs_hash_is_exiting(hs)))
return -ESRCH;
if (unlikely(cfs_hash_is_rehashing(hs)))
return -EALREADY;
if (unlikely(cfs_hash_is_iterating(hs)))
return -EAGAIN;
/* XXX: need to handle case with max_theta != 2.0
* and the case with min_theta != 0.5 */
if ((hs->hs_cur_bits < hs->hs_max_bits) &&
(__cfs_hash_theta(hs) > hs->hs_max_theta))
return hs->hs_cur_bits + 1;
if (!cfs_hash_with_shrink(hs))
return 0;
if ((hs->hs_cur_bits > hs->hs_min_bits) &&
(__cfs_hash_theta(hs) < hs->hs_min_theta))
return hs->hs_cur_bits - 1;
return 0;
}
/**
* don't allow inline rehash if:
* - user wants non-blocking change (add/del) on hash table
* - too many elements
*/
static inline int
cfs_hash_rehash_inline(struct cfs_hash *hs)
{
return !cfs_hash_with_nblk_change(hs) &&
atomic_read(&hs->hs_count) < CFS_HASH_LOOP_HOG;
}
/**
* Add item @hnode to libcfs hash @hs using @key. The registered
* ops->hs_get function will be called when the item is added.
*/
void
cfs_hash_add(struct cfs_hash *hs, const void *key, struct hlist_node *hnode)
{
struct cfs_hash_bd bd;
int bits;
LASSERT(hlist_unhashed(hnode));
cfs_hash_lock(hs, 0);
cfs_hash_bd_get_and_lock(hs, key, &bd, 1);
cfs_hash_key_validate(hs, key, hnode);
cfs_hash_bd_add_locked(hs, &bd, hnode);
cfs_hash_bd_unlock(hs, &bd, 1);
bits = cfs_hash_rehash_bits(hs);
cfs_hash_unlock(hs, 0);
if (bits > 0)
cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs));
}
EXPORT_SYMBOL(cfs_hash_add);
static struct hlist_node *
cfs_hash_find_or_add(struct cfs_hash *hs, const void *key,
struct hlist_node *hnode, int noref)
{
struct hlist_node *ehnode;
struct cfs_hash_bd bds[2];
int bits = 0;
LASSERT(hlist_unhashed(hnode));
cfs_hash_lock(hs, 0);
cfs_hash_dual_bd_get_and_lock(hs, key, bds, 1);
cfs_hash_key_validate(hs, key, hnode);
ehnode = cfs_hash_dual_bd_findadd_locked(hs, bds, key,
hnode, noref);
cfs_hash_dual_bd_unlock(hs, bds, 1);
if (ehnode == hnode) /* new item added */
bits = cfs_hash_rehash_bits(hs);
cfs_hash_unlock(hs, 0);
if (bits > 0)
cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs));
return ehnode;
}
/**
* Add item @hnode to libcfs hash @hs using @key. The registered
* ops->hs_get function will be called if the item was added.
* Returns 0 on success or -EALREADY on key collisions.
*/
int
cfs_hash_add_unique(struct cfs_hash *hs, const void *key, struct hlist_node *hnode)
{
return cfs_hash_find_or_add(hs, key, hnode, 1) != hnode ?
-EALREADY : 0;
}
EXPORT_SYMBOL(cfs_hash_add_unique);
/**
* Add item @hnode to libcfs hash @hs using @key. If this @key
* already exists in the hash then ops->hs_get will be called on the
* conflicting entry and that entry will be returned to the caller.
* Otherwise ops->hs_get is called on the item which was added.
*/
void *
cfs_hash_findadd_unique(struct cfs_hash *hs, const void *key,
struct hlist_node *hnode)
{
hnode = cfs_hash_find_or_add(hs, key, hnode, 0);
return cfs_hash_object(hs, hnode);
}
EXPORT_SYMBOL(cfs_hash_findadd_unique);
/**
* Delete item @hnode from the libcfs hash @hs using @key. The @key
* is required to ensure the correct hash bucket is locked since there
* is no direct linkage from the item to the bucket. The object
* removed from the hash will be returned and obs->hs_put is called
* on the removed object.
*/
void *
cfs_hash_del(struct cfs_hash *hs, const void *key, struct hlist_node *hnode)
{
void *obj = NULL;
int bits = 0;
struct cfs_hash_bd bds[2];
cfs_hash_lock(hs, 0);
cfs_hash_dual_bd_get_and_lock(hs, key, bds, 1);
/* NB: do nothing if @hnode is not in hash table */
if (hnode == NULL || !hlist_unhashed(hnode)) {
if (bds[1].bd_bucket == NULL && hnode != NULL) {
cfs_hash_bd_del_locked(hs, &bds[0], hnode);
} else {
hnode = cfs_hash_dual_bd_finddel_locked(hs, bds,
key, hnode);
}
}
if (hnode != NULL) {
obj = cfs_hash_object(hs, hnode);
bits = cfs_hash_rehash_bits(hs);
}
cfs_hash_dual_bd_unlock(hs, bds, 1);
cfs_hash_unlock(hs, 0);
if (bits > 0)
cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs));
return obj;
}
EXPORT_SYMBOL(cfs_hash_del);
/**
* Delete item given @key in libcfs hash @hs. The first @key found in
* the hash will be removed, if the key exists multiple times in the hash
* @hs this function must be called once per key. The removed object
* will be returned and ops->hs_put is called on the removed object.
*/
void *
cfs_hash_del_key(struct cfs_hash *hs, const void *key)
{
return cfs_hash_del(hs, key, NULL);
}
EXPORT_SYMBOL(cfs_hash_del_key);
/**
* Lookup an item using @key in the libcfs hash @hs and return it.
* If the @key is found in the hash hs->hs_get() is called and the
* matching objects is returned. It is the callers responsibility
* to call the counterpart ops->hs_put using the cfs_hash_put() macro
* when when finished with the object. If the @key was not found
* in the hash @hs NULL is returned.
*/
void *
cfs_hash_lookup(struct cfs_hash *hs, const void *key)
{
void *obj = NULL;
struct hlist_node *hnode;
struct cfs_hash_bd bds[2];
cfs_hash_lock(hs, 0);
cfs_hash_dual_bd_get_and_lock(hs, key, bds, 0);
hnode = cfs_hash_dual_bd_lookup_locked(hs, bds, key);
if (hnode != NULL)
obj = cfs_hash_object(hs, hnode);
cfs_hash_dual_bd_unlock(hs, bds, 0);
cfs_hash_unlock(hs, 0);
return obj;
}
EXPORT_SYMBOL(cfs_hash_lookup);
static void
cfs_hash_for_each_enter(struct cfs_hash *hs)
{
LASSERT(!cfs_hash_is_exiting(hs));
if (!cfs_hash_with_rehash(hs))
return;
/*
* NB: it's race on cfs_has_t::hs_iterating, but doesn't matter
* because it's just an unreliable signal to rehash-thread,
* rehash-thread will try to finish rehash ASAP when seeing this.
*/
hs->hs_iterating = 1;
cfs_hash_lock(hs, 1);
hs->hs_iterators++;
/* NB: iteration is mostly called by service thread,
* we tend to cancel pending rehash-request, instead of
* blocking service thread, we will relaunch rehash request
* after iteration */
if (cfs_hash_is_rehashing(hs))
cfs_hash_rehash_cancel_locked(hs);
cfs_hash_unlock(hs, 1);
}
static void
cfs_hash_for_each_exit(struct cfs_hash *hs)
{
int remained;
int bits;
if (!cfs_hash_with_rehash(hs))
return;
cfs_hash_lock(hs, 1);
remained = --hs->hs_iterators;
bits = cfs_hash_rehash_bits(hs);
cfs_hash_unlock(hs, 1);
/* NB: it's race on cfs_has_t::hs_iterating, see above */
if (remained == 0)
hs->hs_iterating = 0;
if (bits > 0) {
cfs_hash_rehash(hs, atomic_read(&hs->hs_count) <
CFS_HASH_LOOP_HOG);
}
}
/**
* For each item in the libcfs hash @hs call the passed callback @func
* and pass to it as an argument each hash item and the private @data.
*
* a) the function may sleep!
* b) during the callback:
* . the bucket lock is held so the callback must never sleep.
* . if @removal_safe is true, use can remove current item by
* cfs_hash_bd_del_locked
*/
static __u64
cfs_hash_for_each_tight(struct cfs_hash *hs, cfs_hash_for_each_cb_t func,
void *data, int remove_safe)
{
struct hlist_node *hnode;
struct hlist_node *pos;
struct cfs_hash_bd bd;
__u64 count = 0;
int excl = !!remove_safe;
int loop = 0;
int i;
cfs_hash_for_each_enter(hs);
cfs_hash_lock(hs, 0);
LASSERT(!cfs_hash_is_rehashing(hs));
cfs_hash_for_each_bucket(hs, &bd, i) {
struct hlist_head *hhead;
cfs_hash_bd_lock(hs, &bd, excl);
if (func == NULL) { /* only glimpse size */
count += bd.bd_bucket->hsb_count;
cfs_hash_bd_unlock(hs, &bd, excl);
continue;
}
cfs_hash_bd_for_each_hlist(hs, &bd, hhead) {
hlist_for_each_safe(hnode, pos, hhead) {
cfs_hash_bucket_validate(hs, &bd, hnode);
count++;
loop++;
if (func(hs, &bd, hnode, data)) {
cfs_hash_bd_unlock(hs, &bd, excl);
goto out;
}
}
}
cfs_hash_bd_unlock(hs, &bd, excl);
if (loop < CFS_HASH_LOOP_HOG)
continue;
loop = 0;
cfs_hash_unlock(hs, 0);
cond_resched();
cfs_hash_lock(hs, 0);
}
out:
cfs_hash_unlock(hs, 0);
cfs_hash_for_each_exit(hs);
return count;
}
typedef struct {
cfs_hash_cond_opt_cb_t func;
void *arg;
} cfs_hash_cond_arg_t;
static int
cfs_hash_cond_del_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode, void *data)
{
cfs_hash_cond_arg_t *cond = data;
if (cond->func(cfs_hash_object(hs, hnode), cond->arg))
cfs_hash_bd_del_locked(hs, bd, hnode);
return 0;
}
/**
* Delete item from the libcfs hash @hs when @func return true.
* The write lock being hold during loop for each bucket to avoid
* any object be reference.
*/
void
cfs_hash_cond_del(struct cfs_hash *hs, cfs_hash_cond_opt_cb_t func, void *data)
{
cfs_hash_cond_arg_t arg = {
.func = func,
.arg = data,
};
cfs_hash_for_each_tight(hs, cfs_hash_cond_del_locked, &arg, 1);
}
EXPORT_SYMBOL(cfs_hash_cond_del);
void
cfs_hash_for_each(struct cfs_hash *hs,
cfs_hash_for_each_cb_t func, void *data)
{
cfs_hash_for_each_tight(hs, func, data, 0);
}
EXPORT_SYMBOL(cfs_hash_for_each);
void
cfs_hash_for_each_safe(struct cfs_hash *hs,
cfs_hash_for_each_cb_t func, void *data)
{
cfs_hash_for_each_tight(hs, func, data, 1);
}
EXPORT_SYMBOL(cfs_hash_for_each_safe);
static int
cfs_hash_peek(struct cfs_hash *hs, struct cfs_hash_bd *bd,
struct hlist_node *hnode, void *data)
{
*(int *)data = 0;
return 1; /* return 1 to break the loop */
}
int
cfs_hash_is_empty(struct cfs_hash *hs)
{
int empty = 1;
cfs_hash_for_each_tight(hs, cfs_hash_peek, &empty, 0);
return empty;
}
EXPORT_SYMBOL(cfs_hash_is_empty);
__u64
cfs_hash_size_get(struct cfs_hash *hs)
{
return cfs_hash_with_counter(hs) ?
atomic_read(&hs->hs_count) :
cfs_hash_for_each_tight(hs, NULL, NULL, 0);
}
EXPORT_SYMBOL(cfs_hash_size_get);
/*
* cfs_hash_for_each_relax:
* Iterate the hash table and call @func on each item without
* any lock. This function can't guarantee to finish iteration
* if these features are enabled:
*
* a. if rehash_key is enabled, an item can be moved from
* one bucket to another bucket
* b. user can remove non-zero-ref item from hash-table,
* so the item can be removed from hash-table, even worse,
* it's possible that user changed key and insert to another
* hash bucket.
* there's no way for us to finish iteration correctly on previous
* two cases, so iteration has to be stopped on change.
*/
static int
cfs_hash_for_each_relax(struct cfs_hash *hs, cfs_hash_for_each_cb_t func, void *data)
{
struct hlist_node *hnode;
struct hlist_node *tmp;
struct cfs_hash_bd bd;
__u32 version;
int count = 0;
int stop_on_change;
int rc;
int i;
stop_on_change = cfs_hash_with_rehash_key(hs) ||
!cfs_hash_with_no_itemref(hs) ||
CFS_HOP(hs, put_locked) == NULL;
cfs_hash_lock(hs, 0);
LASSERT(!cfs_hash_is_rehashing(hs));
cfs_hash_for_each_bucket(hs, &bd, i) {
struct hlist_head *hhead;
cfs_hash_bd_lock(hs, &bd, 0);
version = cfs_hash_bd_version_get(&bd);
cfs_hash_bd_for_each_hlist(hs, &bd, hhead) {
for (hnode = hhead->first; hnode != NULL;) {
cfs_hash_bucket_validate(hs, &bd, hnode);
cfs_hash_get(hs, hnode);
cfs_hash_bd_unlock(hs, &bd, 0);
cfs_hash_unlock(hs, 0);
rc = func(hs, &bd, hnode, data);
if (stop_on_change)
cfs_hash_put(hs, hnode);
cond_resched();
count++;
cfs_hash_lock(hs, 0);
cfs_hash_bd_lock(hs, &bd, 0);
if (!stop_on_change) {
tmp = hnode->next;
cfs_hash_put_locked(hs, hnode);
hnode = tmp;
} else { /* bucket changed? */
if (version !=
cfs_hash_bd_version_get(&bd))
break;
/* safe to continue because no change */
hnode = hnode->next;
}
if (rc) /* callback wants to break iteration */
break;
}
}
cfs_hash_bd_unlock(hs, &bd, 0);
}
cfs_hash_unlock(hs, 0);
return count;
}
int
cfs_hash_for_each_nolock(struct cfs_hash *hs,
cfs_hash_for_each_cb_t func, void *data)
{
if (cfs_hash_with_no_lock(hs) ||
cfs_hash_with_rehash_key(hs) ||
!cfs_hash_with_no_itemref(hs))
return -EOPNOTSUPP;
if (CFS_HOP(hs, get) == NULL ||
(CFS_HOP(hs, put) == NULL &&
CFS_HOP(hs, put_locked) == NULL))
return -EOPNOTSUPP;
cfs_hash_for_each_enter(hs);
cfs_hash_for_each_relax(hs, func, data);
cfs_hash_for_each_exit(hs);
return 0;
}
EXPORT_SYMBOL(cfs_hash_for_each_nolock);
/**
* For each hash bucket in the libcfs hash @hs call the passed callback
* @func until all the hash buckets are empty. The passed callback @func
* or the previously registered callback hs->hs_put must remove the item
* from the hash. You may either use the cfs_hash_del() or hlist_del()
* functions. No rwlocks will be held during the callback @func it is
* safe to sleep if needed. This function will not terminate until the
* hash is empty. Note it is still possible to concurrently add new
* items in to the hash. It is the callers responsibility to ensure
* the required locking is in place to prevent concurrent insertions.
*/
int
cfs_hash_for_each_empty(struct cfs_hash *hs,
cfs_hash_for_each_cb_t func, void *data)
{
unsigned i = 0;
if (cfs_hash_with_no_lock(hs))
return -EOPNOTSUPP;
if (CFS_HOP(hs, get) == NULL ||
(CFS_HOP(hs, put) == NULL &&
CFS_HOP(hs, put_locked) == NULL))
return -EOPNOTSUPP;
cfs_hash_for_each_enter(hs);
while (cfs_hash_for_each_relax(hs, func, data)) {
CDEBUG(D_INFO, "Try to empty hash: %s, loop: %u\n",
hs->hs_name, i++);
}
cfs_hash_for_each_exit(hs);
return 0;
}
EXPORT_SYMBOL(cfs_hash_for_each_empty);
void
cfs_hash_hlist_for_each(struct cfs_hash *hs, unsigned hindex,
cfs_hash_for_each_cb_t func, void *data)
{
struct hlist_head *hhead;
struct hlist_node *hnode;
struct cfs_hash_bd bd;
cfs_hash_for_each_enter(hs);
cfs_hash_lock(hs, 0);
if (hindex >= CFS_HASH_NHLIST(hs))
goto out;
cfs_hash_bd_index_set(hs, hindex, &bd);
cfs_hash_bd_lock(hs, &bd, 0);
hhead = cfs_hash_bd_hhead(hs, &bd);
hlist_for_each(hnode, hhead) {
if (func(hs, &bd, hnode, data))
break;
}
cfs_hash_bd_unlock(hs, &bd, 0);
out:
cfs_hash_unlock(hs, 0);
cfs_hash_for_each_exit(hs);
}
EXPORT_SYMBOL(cfs_hash_hlist_for_each);
/*
* For each item in the libcfs hash @hs which matches the @key call
* the passed callback @func and pass to it as an argument each hash
* item and the private @data. During the callback the bucket lock
* is held so the callback must never sleep.
*/
void
cfs_hash_for_each_key(struct cfs_hash *hs, const void *key,
cfs_hash_for_each_cb_t func, void *data)
{
struct hlist_node *hnode;
struct cfs_hash_bd bds[2];
unsigned i;
cfs_hash_lock(hs, 0);
cfs_hash_dual_bd_get_and_lock(hs, key, bds, 0);
cfs_hash_for_each_bd(bds, 2, i) {
struct hlist_head *hlist = cfs_hash_bd_hhead(hs, &bds[i]);
hlist_for_each(hnode, hlist) {
cfs_hash_bucket_validate(hs, &bds[i], hnode);
if (cfs_hash_keycmp(hs, key, hnode)) {
if (func(hs, &bds[i], hnode, data))
break;
}
}
}
cfs_hash_dual_bd_unlock(hs, bds, 0);
cfs_hash_unlock(hs, 0);
}
EXPORT_SYMBOL(cfs_hash_for_each_key);
/**
* Rehash the libcfs hash @hs to the given @bits. This can be used
* to grow the hash size when excessive chaining is detected, or to
* shrink the hash when it is larger than needed. When the CFS_HASH_REHASH
* flag is set in @hs the libcfs hash may be dynamically rehashed
* during addition or removal if the hash's theta value exceeds
* either the hs->hs_min_theta or hs->max_theta values. By default
* these values are tuned to keep the chained hash depth small, and
* this approach assumes a reasonably uniform hashing function. The
* theta thresholds for @hs are tunable via cfs_hash_set_theta().
*/
void
cfs_hash_rehash_cancel_locked(struct cfs_hash *hs)
{
int i;
/* need hold cfs_hash_lock(hs, 1) */
LASSERT(cfs_hash_with_rehash(hs) &&
!cfs_hash_with_no_lock(hs));
if (!cfs_hash_is_rehashing(hs))
return;
if (cfs_wi_deschedule(cfs_sched_rehash, &hs->hs_rehash_wi)) {
hs->hs_rehash_bits = 0;
return;
}
for (i = 2; cfs_hash_is_rehashing(hs); i++) {
cfs_hash_unlock(hs, 1);
/* raise console warning while waiting too long */
CDEBUG(IS_PO2(i >> 3) ? D_WARNING : D_INFO,
"hash %s is still rehashing, rescheded %d\n",
hs->hs_name, i - 1);
cond_resched();
cfs_hash_lock(hs, 1);
}
}
EXPORT_SYMBOL(cfs_hash_rehash_cancel_locked);
void
cfs_hash_rehash_cancel(struct cfs_hash *hs)
{
cfs_hash_lock(hs, 1);
cfs_hash_rehash_cancel_locked(hs);
cfs_hash_unlock(hs, 1);
}
EXPORT_SYMBOL(cfs_hash_rehash_cancel);
int
cfs_hash_rehash(struct cfs_hash *hs, int do_rehash)
{
int rc;
LASSERT(cfs_hash_with_rehash(hs) && !cfs_hash_with_no_lock(hs));
cfs_hash_lock(hs, 1);
rc = cfs_hash_rehash_bits(hs);
if (rc <= 0) {
cfs_hash_unlock(hs, 1);
return rc;
}
hs->hs_rehash_bits = rc;
if (!do_rehash) {
/* launch and return */
cfs_wi_schedule(cfs_sched_rehash, &hs->hs_rehash_wi);
cfs_hash_unlock(hs, 1);
return 0;
}
/* rehash right now */
cfs_hash_unlock(hs, 1);
return cfs_hash_rehash_worker(&hs->hs_rehash_wi);
}
EXPORT_SYMBOL(cfs_hash_rehash);
static int
cfs_hash_rehash_bd(struct cfs_hash *hs, struct cfs_hash_bd *old)
{
struct cfs_hash_bd new;
struct hlist_head *hhead;
struct hlist_node *hnode;
struct hlist_node *pos;
void *key;
int c = 0;
/* hold cfs_hash_lock(hs, 1), so don't need any bucket lock */
cfs_hash_bd_for_each_hlist(hs, old, hhead) {
hlist_for_each_safe(hnode, pos, hhead) {
key = cfs_hash_key(hs, hnode);
LASSERT(key != NULL);
/* Validate hnode is in the correct bucket. */
cfs_hash_bucket_validate(hs, old, hnode);
/*
* Delete from old hash bucket; move to new bucket.
* ops->hs_key must be defined.
*/
cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets,
hs->hs_rehash_bits, key, &new);
cfs_hash_bd_move_locked(hs, old, &new, hnode);
c++;
}
}
return c;
}
static int
cfs_hash_rehash_worker(cfs_workitem_t *wi)
{
struct cfs_hash *hs = container_of(wi, struct cfs_hash, hs_rehash_wi);
struct cfs_hash_bucket **bkts;
struct cfs_hash_bd bd;
unsigned int old_size;
unsigned int new_size;
int bsize;
int count = 0;
int rc = 0;
int i;
LASSERT (hs != NULL && cfs_hash_with_rehash(hs));
cfs_hash_lock(hs, 0);
LASSERT(cfs_hash_is_rehashing(hs));
old_size = CFS_HASH_NBKT(hs);
new_size = CFS_HASH_RH_NBKT(hs);
cfs_hash_unlock(hs, 0);
/*
* don't need hs::hs_rwlock for hs::hs_buckets,
* because nobody can change bkt-table except me.
*/
bkts = cfs_hash_buckets_realloc(hs, hs->hs_buckets,
old_size, new_size);
cfs_hash_lock(hs, 1);
if (bkts == NULL) {
rc = -ENOMEM;
goto out;
}
if (bkts == hs->hs_buckets) {
bkts = NULL; /* do nothing */
goto out;
}
rc = __cfs_hash_theta(hs);
if ((rc >= hs->hs_min_theta) && (rc <= hs->hs_max_theta)) {
/* free the new allocated bkt-table */
old_size = new_size;
new_size = CFS_HASH_NBKT(hs);
rc = -EALREADY;
goto out;
}
LASSERT(hs->hs_rehash_buckets == NULL);
hs->hs_rehash_buckets = bkts;
rc = 0;
cfs_hash_for_each_bucket(hs, &bd, i) {
if (cfs_hash_is_exiting(hs)) {
rc = -ESRCH;
/* someone wants to destroy the hash, abort now */
if (old_size < new_size) /* OK to free old bkt-table */
break;
/* it's shrinking, need free new bkt-table */
hs->hs_rehash_buckets = NULL;
old_size = new_size;
new_size = CFS_HASH_NBKT(hs);
goto out;
}
count += cfs_hash_rehash_bd(hs, &bd);
if (count < CFS_HASH_LOOP_HOG ||
cfs_hash_is_iterating(hs)) { /* need to finish ASAP */
continue;
}
count = 0;
cfs_hash_unlock(hs, 1);
cond_resched();
cfs_hash_lock(hs, 1);
}
hs->hs_rehash_count++;
bkts = hs->hs_buckets;
hs->hs_buckets = hs->hs_rehash_buckets;
hs->hs_rehash_buckets = NULL;
hs->hs_cur_bits = hs->hs_rehash_bits;
out:
hs->hs_rehash_bits = 0;
if (rc == -ESRCH) /* never be scheduled again */
cfs_wi_exit(cfs_sched_rehash, wi);
bsize = cfs_hash_bkt_size(hs);
cfs_hash_unlock(hs, 1);
/* can't refer to @hs anymore because it could be destroyed */
if (bkts != NULL)
cfs_hash_buckets_free(bkts, bsize, new_size, old_size);
if (rc != 0)
CDEBUG(D_INFO, "early quit of rehashing: %d\n", rc);
/* return 1 only if cfs_wi_exit is called */
return rc == -ESRCH;
}
/**
* Rehash the object referenced by @hnode in the libcfs hash @hs. The
* @old_key must be provided to locate the objects previous location
* in the hash, and the @new_key will be used to reinsert the object.
* Use this function instead of a cfs_hash_add() + cfs_hash_del()
* combo when it is critical that there is no window in time where the
* object is missing from the hash. When an object is being rehashed
* the registered cfs_hash_get() and cfs_hash_put() functions will
* not be called.
*/
void cfs_hash_rehash_key(struct cfs_hash *hs, const void *old_key,
void *new_key, struct hlist_node *hnode)
{
struct cfs_hash_bd bds[3];
struct cfs_hash_bd old_bds[2];
struct cfs_hash_bd new_bd;
LASSERT(!hlist_unhashed(hnode));
cfs_hash_lock(hs, 0);
cfs_hash_dual_bd_get(hs, old_key, old_bds);
cfs_hash_bd_get(hs, new_key, &new_bd);
bds[0] = old_bds[0];
bds[1] = old_bds[1];
bds[2] = new_bd;
/* NB: bds[0] and bds[1] are ordered already */
cfs_hash_bd_order(&bds[1], &bds[2]);
cfs_hash_bd_order(&bds[0], &bds[1]);
cfs_hash_multi_bd_lock(hs, bds, 3, 1);
if (likely(old_bds[1].bd_bucket == NULL)) {
cfs_hash_bd_move_locked(hs, &old_bds[0], &new_bd, hnode);
} else {
cfs_hash_dual_bd_finddel_locked(hs, old_bds, old_key, hnode);
cfs_hash_bd_add_locked(hs, &new_bd, hnode);
}
/* overwrite key inside locks, otherwise may screw up with
* other operations, i.e: rehash */
cfs_hash_keycpy(hs, new_key, hnode);
cfs_hash_multi_bd_unlock(hs, bds, 3, 1);
cfs_hash_unlock(hs, 0);
}
EXPORT_SYMBOL(cfs_hash_rehash_key);
int cfs_hash_debug_header(struct seq_file *m)
{
return seq_printf(m, "%-*s%6s%6s%6s%6s%6s%6s%6s%7s%8s%8s%8s%s\n",
CFS_HASH_BIGNAME_LEN,
"name", "cur", "min", "max", "theta", "t-min", "t-max",
"flags", "rehash", "count", "maxdep", "maxdepb",
" distribution");
}
EXPORT_SYMBOL(cfs_hash_debug_header);
static struct cfs_hash_bucket **
cfs_hash_full_bkts(struct cfs_hash *hs)
{
/* NB: caller should hold hs->hs_rwlock if REHASH is set */
if (hs->hs_rehash_buckets == NULL)
return hs->hs_buckets;
LASSERT(hs->hs_rehash_bits != 0);
return hs->hs_rehash_bits > hs->hs_cur_bits ?
hs->hs_rehash_buckets : hs->hs_buckets;
}
static unsigned int
cfs_hash_full_nbkt(struct cfs_hash *hs)
{
/* NB: caller should hold hs->hs_rwlock if REHASH is set */
if (hs->hs_rehash_buckets == NULL)
return CFS_HASH_NBKT(hs);
LASSERT(hs->hs_rehash_bits != 0);
return hs->hs_rehash_bits > hs->hs_cur_bits ?
CFS_HASH_RH_NBKT(hs) : CFS_HASH_NBKT(hs);
}
int cfs_hash_debug_str(struct cfs_hash *hs, struct seq_file *m)
{
int dist[8] = { 0, };
int maxdep = -1;
int maxdepb = -1;
int total = 0;
int theta;
int i;
cfs_hash_lock(hs, 0);
theta = __cfs_hash_theta(hs);
seq_printf(m, "%-*s %5d %5d %5d %d.%03d %d.%03d %d.%03d 0x%02x %6d ",
CFS_HASH_BIGNAME_LEN, hs->hs_name,
1 << hs->hs_cur_bits, 1 << hs->hs_min_bits,
1 << hs->hs_max_bits,
__cfs_hash_theta_int(theta), __cfs_hash_theta_frac(theta),
__cfs_hash_theta_int(hs->hs_min_theta),
__cfs_hash_theta_frac(hs->hs_min_theta),
__cfs_hash_theta_int(hs->hs_max_theta),
__cfs_hash_theta_frac(hs->hs_max_theta),
hs->hs_flags, hs->hs_rehash_count);
/*
* The distribution is a summary of the chained hash depth in
* each of the libcfs hash buckets. Each buckets hsb_count is
* divided by the hash theta value and used to generate a
* histogram of the hash distribution. A uniform hash will
* result in all hash buckets being close to the average thus
* only the first few entries in the histogram will be non-zero.
* If you hash function results in a non-uniform hash the will
* be observable by outlier bucks in the distribution histogram.
*
* Uniform hash distribution: 128/128/0/0/0/0/0/0
* Non-Uniform hash distribution: 128/125/0/0/0/0/2/1
*/
for (i = 0; i < cfs_hash_full_nbkt(hs); i++) {
struct cfs_hash_bd bd;
bd.bd_bucket = cfs_hash_full_bkts(hs)[i];
cfs_hash_bd_lock(hs, &bd, 0);
if (maxdep < bd.bd_bucket->hsb_depmax) {
maxdep = bd.bd_bucket->hsb_depmax;
maxdepb = ffz(~maxdep);
}
total += bd.bd_bucket->hsb_count;
dist[min(fls(bd.bd_bucket->hsb_count / max(theta, 1)), 7)]++;
cfs_hash_bd_unlock(hs, &bd, 0);
}
seq_printf(m, "%7d %7d %7d ", total, maxdep, maxdepb);
for (i = 0; i < 8; i++)
seq_printf(m, "%d%c", dist[i], (i == 7) ? '\n' : '/');
cfs_hash_unlock(hs, 0);
return 0;
}
EXPORT_SYMBOL(cfs_hash_debug_str);
|