/*===========================================================================*/ /* */ /* Mesa-3.0 DirectX 6 Driver */ /* */ /* By Leigh McRae */ /* */ /* http://www.altsoftware.com/ */ /* */ /* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ /*===========================================================================*/ #include "Debug.h" /*===========================================================================*/ /* Global variables. */ /*===========================================================================*/ DWORD g_DBGMask = DBG_ALL_ERROR; /*===========================================================================*/ /* This is your basic DPF function with printf like support. The function */ /* also works with a global debug mask variable. I have written support that*/ /* allows for the user's enviroment variable space to be read and set the */ /* masks. This is done when the dll starts and is only in the debug version.*/ /*===========================================================================*/ /* RETURN: */ /*===========================================================================*/ void _cdecl DebugPrint( int mask, char *pszFormat, ... ) { char buffer[512]; va_list args; /* A mask of 0 will always pass. Easy to remeber. */ if ( (mask == 0) || (mask & g_DBGMask) ) { va_start( args, pszFormat ); if ( mask & DBG_ALL_ERROR ) OutputDebugString( "MesaD3D: (ERROR)" ); else OutputDebugString( "MesaD3D: " ); vsprintf( buffer, pszFormat, args ); strcat( buffer, "\n" ); OutputDebugString( buffer ); va_end( args ); } } /*===========================================================================*/ /* This call reads the users enviroment variables and sets any debug mask */ /* that they have set to TRUE. Now the value must be "TRUE". */ /*===========================================================================*/ /* RETURN: */ /*===========================================================================*/ void ReadDBGEnv( void ) { g_DBGMask = DBG_ALL_ERROR; #define IS_VAR_SET(v) if ( getenv( # v ) && !strcmp(getenv( # v ),"TRUE") ) g_DBGMask |= v; IS_VAR_SET( DBG_FUNC ); IS_VAR_SET( DBG_STATES ); IS_VAR_SET( DBG_CNTX_INFO ); IS_VAR_SET( DBG_CNTX_WARN ); IS_VAR_SET( DBG_CNTX_PROFILE ); IS_VAR_SET( DBG_CNTX_ERROR ); IS_VAR_SET( DBG_CNTX_ALL ); IS_VAR_SET( DBG_PRIM_INFO ); IS_VAR_SET( DBG_PRIM_WARN ); IS_VAR_SET( DBG_PRIM_PROFILE ); IS_VAR_SET( DBG_PRIM_ERROR ); IS_VAR_SET( DBG_PRIM_ALL ); IS_VAR_SET( DBG_TXT_INFO ); IS_VAR_SET( DBG_TXT_WARN ); IS_VAR_SET( DBG_TXT_PROFILE ); IS_VAR_SET( DBG_TXT_ERROR ); IS_VAR_SET( DBG_TXT_ALL ); IS_VAR_SET( DBG_ALL_INFO ); IS_VAR_SET( DBG_ALL_WARN ); IS_VAR_SET( DBG_ALL_PROFILE ); IS_VAR_SET( DBG_ALL_ERROR ); IS_VAR_SET( DBG_ALL ); #undef IS_VAR_SET } /*===========================================================================*/ /* This function will take a pointer to a DDSURFACEDESC2 structure & display*/ /* the parsed information using a DPF call. */ /*===========================================================================*/ /* RETURN: */ /*===========================================================================*/ void DebugPixelFormat( char *pszSurfaceName, DDPIXELFORMAT *pddpf ) { char buffer[256]; /* Parse the flag type and write the string equivalent. */ if ( pddpf->dwFlags & DDPF_ALPHA ) strcat( buffer, "DDPF_ALPHA " ); if ( pddpf->dwFlags & DDPF_ALPHAPIXELS ) strcat( buffer, "DDPF_ALPHAPIXELS " ); if ( pddpf->dwFlags & DDPF_ALPHAPREMULT ) strcat( buffer, "DDPF_ALPHAPREMULT " ); if ( pddpf->dwFlags & DDPF_BUMPLUMINANCE ) strcat( buffer, "DDPF_BUMPLUMINANCE " ); if ( pddpf->dwFlags & DDPF_BUMPDUDV ) strcat( buffer, "DDPF_BUMPDUDV " ); if ( pddpf->dwFlags & DDPF_COMPRESSED ) strcat( buffer, "DDPF_COMPRESSED " ); if ( pddpf->dwFlags & DDPF_FOURCC ) strcat( buffer, "DDPF_FOURCC " ); if ( pddpf->dwFlags & DDPF_LUMINANCE ) strcat( buffer, "DDPF_LUMINANCE " ); if ( pddpf->dwFlags & DDPF_PALETTEINDEXED1 ) strcat( buffer, "DDPF_PALETTEINDEXED1 " ); if ( pddpf->dwFlags & DDPF_PALETTEINDEXED2 ) strcat( buffer, "DDPF_PALETTEINDEXED2 " ); if ( pddpf->dwFlags & DDPF_PALETTEINDEXED4 ) strcat( buffer, "DDPF_PALETTEINDEXED4 " ); if ( pddpf->dwFlags & DDPF_PALETTEINDEXED8 ) strcat( buffer, "DDPF_PALETTEINDEXED8 " ); if ( pddpf->dwFlags & DDPF_PALETTEINDEXEDTO8 ) strcat( buffer, "DDPF_PALETTEINDEXEDTO8 " ); if ( pddpf->dwFlags & DDPF_RGB ) strcat( buffer, "DDPF_RGB " ); if ( pddpf->dwFlags & DDPF_RGBTOYUV ) strcat( buffer, "DDPF_RGBTOYUV " ); if ( pddpf->dwFlags & DDPF_STENCILBUFFER ) strcat( buffer, "DDPF_STENCILBUFFER " ); if ( pddpf->dwFlags & DDPF_YUV ) strcat( buffer, "DDPF_YUV " ); if ( pddpf->dwFlags & DDPF_ZBUFFER ) strcat( buffer, "DDPF_ZBUFFER " ); if ( pddpf->dwFlags & DDPF_ZPIXELS ) strcat( buffer, "DDPF_ZPIXELS " ); DPF(( (DBG_TXT_INFO|DBG_CNTX_INFO),"%s", buffer )); } a> 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 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
 * Copyright 2016 Gary Mills
 * Copyright (c) 2017 Datto Inc.
 * Copyright 2017 Joyent, Inc.
 */

#include <sys/dsl_scan.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_synctask.h>
#include <sys/dnode.h>
#include <sys/dmu_tx.h>
#include <sys/dmu_objset.h>
#include <sys/arc.h>
#include <sys/zap.h>
#include <sys/zio.h>
#include <sys/zfs_context.h>
#include <sys/fs/zfs.h>
#include <sys/zfs_znode.h>
#include <sys/spa_impl.h>
#include <sys/vdev_impl.h>
#include <sys/zil_impl.h>
#include <sys/zio_checksum.h>
#include <sys/ddt.h>
#include <sys/sa.h>
#include <sys/sa_impl.h>
#include <sys/zfeature.h>
#include <sys/abd.h>
#include <sys/range_tree.h>
#ifdef _KERNEL
#include <sys/zfs_vfsops.h>
#endif

/*
 * Grand theory statement on scan queue sorting
 *
 * Scanning is implemented by recursively traversing all indirection levels
 * in an object and reading all blocks referenced from said objects. This
 * results in us approximately traversing the object from lowest logical
 * offset to the highest. For best performance, we would want the logical
 * blocks to be physically contiguous. However, this is frequently not the
 * case with pools given the allocation patterns of copy-on-write filesystems.
 * So instead, we put the I/Os into a reordering queue and issue them in a
 * way that will most benefit physical disks (LBA-order).
 *
 * Queue management:
 *
 * Ideally, we would want to scan all metadata and queue up all block I/O
 * prior to starting to issue it, because that allows us to do an optimal
 * sorting job. This can however consume large amounts of memory. Therefore
 * we continuously monitor the size of the queues and constrain them to 5%
 * (zfs_scan_mem_lim_fact) of physmem. If the queues grow larger than this
 * limit, we clear out a few of the largest extents at the head of the queues
 * to make room for more scanning. Hopefully, these extents will be fairly
 * large and contiguous, allowing us to approach sequential I/O throughput
 * even without a fully sorted tree.
 *
 * Metadata scanning takes place in dsl_scan_visit(), which is called from
 * dsl_scan_sync() every spa_sync(). If we have either fully scanned all
 * metadata on the pool, or we need to make room in memory because our
 * queues are too large, dsl_scan_visit() is postponed and
 * scan_io_queues_run() is called from dsl_scan_sync() instead. This implies
 * that metadata scanning and queued I/O issuing are mutually exclusive. This
 * allows us to provide maximum sequential I/O throughput for the majority of
 * I/O's issued since sequential I/O performance is significantly negatively
 * impacted if it is interleaved with random I/O.
 *
 * Implementation Notes
 *
 * One side effect of the queued scanning algorithm is that the scanning code
 * needs to be notified whenever a block is freed. This is needed to allow
 * the scanning code to remove these I/Os from the issuing queue. Additionally,
 * we do not attempt to queue gang blocks to be issued sequentially since this
 * is very hard to do and would have an extremely limited performance benefit.
 * Instead, we simply issue gang I/Os as soon as we find them using the legacy
 * algorithm.
 *
 * Backwards compatibility
 *
 * This new algorithm is backwards compatible with the legacy on-disk data
 * structures (and therefore does not require a new feature flag).
 * Periodically during scanning (see zfs_scan_checkpoint_intval), the scan
 * will stop scanning metadata (in logical order) and wait for all outstanding
 * sorted I/O to complete. Once this is done, we write out a checkpoint
 * bookmark, indicating that we have scanned everything logically before it.
 * If the pool is imported on a machine without the new sorting algorithm,
 * the scan simply resumes from the last checkpoint using the legacy algorithm.
 */

typedef int (scan_cb_t)(dsl_pool_t *, const blkptr_t *,
    const zbookmark_phys_t *);

static scan_cb_t dsl_scan_scrub_cb;

static int scan_ds_queue_compare(const void *a, const void *b);
static int scan_prefetch_queue_compare(const void *a, const void *b);
static void scan_ds_queue_clear(dsl_scan_t *scn);
static void scan_ds_prefetch_queue_clear(dsl_scan_t *scn);
static boolean_t scan_ds_queue_contains(dsl_scan_t *scn, uint64_t dsobj,
    uint64_t *txg);
static void scan_ds_queue_insert(dsl_scan_t *scn, uint64_t dsobj, uint64_t txg);
static void scan_ds_queue_remove(dsl_scan_t *scn, uint64_t dsobj);
static void scan_ds_queue_sync(dsl_scan_t *scn, dmu_tx_t *tx);
static uint64_t dsl_scan_count_leaves(vdev_t *vd);

extern int zfs_vdev_async_write_active_min_dirty_percent;

/*
 * By default zfs will check to ensure it is not over the hard memory
 * limit before each txg. If finer-grained control of this is needed
 * this value can be set to 1 to enable checking before scanning each
 * block.
 */
int zfs_scan_strict_mem_lim = B_FALSE;

/*
 * Maximum number of parallelly executed bytes per leaf vdev. We attempt
 * to strike a balance here between keeping the vdev queues full of I/Os
 * at all times and not overflowing the queues to cause long latency,
 * which would cause long txg sync times. No matter what, we will not
 * overload the drives with I/O, since that is protected by
 * zfs_vdev_scrub_max_active.
 */
unsigned long zfs_scan_vdev_limit = 4 << 20;

int zfs_scan_issue_strategy = 0;
int zfs_scan_legacy = B_FALSE; /* don't queue & sort zios, go direct */
unsigned long zfs_scan_max_ext_gap = 2 << 20; /* in bytes */

/*
 * fill_weight is non-tunable at runtime, so we copy it at module init from
 * zfs_scan_fill_weight. Runtime adjustments to zfs_scan_fill_weight would
 * break queue sorting.
 */
int zfs_scan_fill_weight = 3;
static uint64_t fill_weight;

/* See dsl_scan_should_clear() for details on the memory limit tunables */
uint64_t zfs_scan_mem_lim_min = 16 << 20;	/* bytes */
uint64_t zfs_scan_mem_lim_soft_max = 128 << 20;	/* bytes */
int zfs_scan_mem_lim_fact = 20;		/* fraction of physmem */
int zfs_scan_mem_lim_soft_fact = 20;	/* fraction of mem lim above */

int zfs_scrub_min_time_ms = 1000; /* min millisecs to scrub per txg */
int zfs_obsolete_min_time_ms = 500; /* min millisecs to obsolete per txg */
int zfs_free_min_time_ms = 1000; /* min millisecs to free per txg */
int zfs_resilver_min_time_ms = 3000; /* min millisecs to resilver per txg */
int zfs_scan_checkpoint_intval = 7200; /* in seconds */
int zfs_scan_suspend_progress = 0; /* set to prevent scans from progressing */
int zfs_no_scrub_io = B_FALSE; /* set to disable scrub i/o */
int zfs_no_scrub_prefetch = B_FALSE; /* set to disable scrub prefetch */
enum ddt_class zfs_scrub_ddt_class_max = DDT_CLASS_DUPLICATE;
/* max number of blocks to free in a single TXG */
unsigned long zfs_async_block_max_blocks = 100000;

int zfs_resilver_disable_defer = 0; /* set to disable resilver deferring */

/*
 * We wait a few txgs after importing a pool to begin scanning so that
 * the import / mounting code isn't held up by scrub / resilver IO.
 * Unfortunately, it is a bit difficult to determine exactly how long
 * this will take since userspace will trigger fs mounts asynchronously
 * and the kernel will create zvol minors asynchronously. As a result,
 * the value provided here is a bit arbitrary, but represents a
 * reasonable estimate of how many txgs it will take to finish fully
 * importing a pool
 */
#define	SCAN_IMPORT_WAIT_TXGS 		5

#define	DSL_SCAN_IS_SCRUB_RESILVER(scn) \
	((scn)->scn_phys.scn_func == POOL_SCAN_SCRUB || \
	(scn)->scn_phys.scn_func == POOL_SCAN_RESILVER)

/*
 * Enable/disable the processing of the free_bpobj object.
 */
int zfs_free_bpobj_enabled = 1;

/* the order has to match pool_scan_type */
static scan_cb_t *scan_funcs[POOL_SCAN_FUNCS] = {
	NULL,
	dsl_scan_scrub_cb,	/* POOL_SCAN_SCRUB */
	dsl_scan_scrub_cb,	/* POOL_SCAN_RESILVER */
};

/* In core node for the scn->scn_queue. Represents a dataset to be scanned */
typedef struct {
	uint64_t	sds_dsobj;
	uint64_t	sds_txg;
	avl_node_t	sds_node;
} scan_ds_t;

/*
 * This controls what conditions are placed on dsl_scan_sync_state():
 * SYNC_OPTIONAL) write out scn_phys iff scn_bytes_pending == 0
 * SYNC_MANDATORY) write out scn_phys always. scn_bytes_pending must be 0.
 * SYNC_CACHED) if scn_bytes_pending == 0, write out scn_phys. Otherwise
 *	write out the scn_phys_cached version.
 * See dsl_scan_sync_state for details.
 */
typedef enum {
	SYNC_OPTIONAL,
	SYNC_MANDATORY,
	SYNC_CACHED
} state_sync_type_t;

/*
 * This struct represents the minimum information needed to reconstruct a
 * zio for sequential scanning. This is useful because many of these will
 * accumulate in the sequential IO queues before being issued, so saving
 * memory matters here.
 */
typedef struct scan_io {
	/* fields from blkptr_t */
	uint64_t		sio_blk_prop;
	uint64_t		sio_phys_birth;
	uint64_t		sio_birth;
	zio_cksum_t		sio_cksum;
	uint32_t		sio_nr_dvas;

	/* fields from zio_t */
	uint32_t		sio_flags;
	zbookmark_phys_t	sio_zb;

	/* members for queue sorting */
	union {
		avl_node_t	sio_addr_node; /* link into issuing queue */
		list_node_t	sio_list_node; /* link for issuing to disk */
	} sio_nodes;

	/*
	 * There may be up to SPA_DVAS_PER_BP DVAs here from the bp,
	 * depending on how many were in the original bp. Only the
	 * first DVA is really used for sorting and issuing purposes.
	 * The other DVAs (if provided) simply exist so that the zio
	 * layer can find additional copies to repair from in the
	 * event of an error. This array must go at the end of the
	 * struct to allow this for the variable number of elements.
	 */
	dva_t			sio_dva[0];
} scan_io_t;

#define	SIO_SET_OFFSET(sio, x)		DVA_SET_OFFSET(&(sio)->sio_dva[0], x)
#define	SIO_SET_ASIZE(sio, x)		DVA_SET_ASIZE(&(sio)->sio_dva[0], x)
#define	SIO_GET_OFFSET(sio)		DVA_GET_OFFSET(&(sio)->sio_dva[0])
#define	SIO_GET_ASIZE(sio)		DVA_GET_ASIZE(&(sio)->sio_dva[0])
#define	SIO_GET_END_OFFSET(sio)		\
	(SIO_GET_OFFSET(sio) + SIO_GET_ASIZE(sio))
#define	SIO_GET_MUSED(sio)		\
	(sizeof (scan_io_t) + ((sio)->sio_nr_dvas * sizeof (dva_t)))

struct dsl_scan_io_queue {
	dsl_scan_t	*q_scn; /* associated dsl_scan_t */
	vdev_t		*q_vd; /* top-level vdev that this queue represents */

	/* trees used for sorting I/Os and extents of I/Os */
	range_tree_t	*q_exts_by_addr;
	avl_tree_t	q_exts_by_size;
	avl_tree_t	q_sios_by_addr;
	uint64_t	q_sio_memused;

	/* members for zio rate limiting */
	uint64_t	q_maxinflight_bytes;
	uint64_t	q_inflight_bytes;
	kcondvar_t	q_zio_cv; /* used under vd->vdev_scan_io_queue_lock */

	/* per txg statistics */
	uint64_t	q_total_seg_size_this_txg;
	uint64_t	q_segs_this_txg;
	uint64_t	q_total_zio_size_this_txg;
	uint64_t	q_zios_this_txg;
};

/* private data for dsl_scan_prefetch_cb() */
typedef struct scan_prefetch_ctx {
	zfs_refcount_t spc_refcnt;	/* refcount for memory management */
	dsl_scan_t *spc_scn;		/* dsl_scan_t for the pool */
	boolean_t spc_root;		/* is this prefetch for an objset? */
	uint8_t spc_indblkshift;	/* dn_indblkshift of current dnode */
	uint16_t spc_datablkszsec;	/* dn_idatablkszsec of current dnode */
} scan_prefetch_ctx_t;

/* private data for dsl_scan_prefetch() */
typedef struct scan_prefetch_issue_ctx {
	avl_node_t spic_avl_node;	/* link into scn->scn_prefetch_queue */
	scan_prefetch_ctx_t *spic_spc;	/* spc for the callback */
	blkptr_t spic_bp;		/* bp to prefetch */
	zbookmark_phys_t spic_zb;	/* bookmark to prefetch */
} scan_prefetch_issue_ctx_t;

static void scan_exec_io(dsl_pool_t *dp, const blkptr_t *bp, int zio_flags,
    const zbookmark_phys_t *zb, dsl_scan_io_queue_t *queue);
static void scan_io_queue_insert_impl(dsl_scan_io_queue_t *queue,
    scan_io_t *sio);

static dsl_scan_io_queue_t *scan_io_queue_create(vdev_t *vd);
static void scan_io_queues_destroy(dsl_scan_t *scn);

static kmem_cache_t *sio_cache[SPA_DVAS_PER_BP];

/* sio->sio_nr_dvas must be set so we know which cache to free from */
static void
sio_free(scan_io_t *sio)
{
	ASSERT3U(sio->sio_nr_dvas, >, 0);
	ASSERT3U(sio->sio_nr_dvas, <=, SPA_DVAS_PER_BP);

	kmem_cache_free(sio_cache[sio->sio_nr_dvas - 1], sio);
}

/* It is up to the caller to set sio->sio_nr_dvas for freeing */
static scan_io_t *
sio_alloc(unsigned short nr_dvas)
{
	ASSERT3U(nr_dvas, >, 0);
	ASSERT3U(nr_dvas, <=, SPA_DVAS_PER_BP);

	return (kmem_cache_alloc(sio_cache[nr_dvas - 1], KM_SLEEP));
}

void
scan_init(void)
{
	/*
	 * This is used in ext_size_compare() to weight segments
	 * based on how sparse they are. This cannot be changed
	 * mid-scan and the tree comparison functions don't currently
	 * have a mechanism for passing additional context to the
	 * compare functions. Thus we store this value globally and
	 * we only allow it to be set at module initialization time
	 */
	fill_weight = zfs_scan_fill_weight;

	for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
		char name[36];

		(void) sprintf(name, "sio_cache_%d", i);
		sio_cache[i] = kmem_cache_create(name,
		    (sizeof (scan_io_t) + ((i + 1) * sizeof (dva_t))),
		    0, NULL, NULL, NULL, NULL, NULL, 0);
	}
}

void
scan_fini(void)
{
	for (int i = 0; i < SPA_DVAS_PER_BP; i++) {
		kmem_cache_destroy(sio_cache[i]);
	}
}

static inline boolean_t
dsl_scan_is_running(const dsl_scan_t *scn)
{
	return (scn->scn_phys.scn_state == DSS_SCANNING);
}

boolean_t
dsl_scan_resilvering(dsl_pool_t *dp)
{
	return (dsl_scan_is_running(dp->dp_scan) &&
	    dp->dp_scan->scn_phys.scn_func == POOL_SCAN_RESILVER);
}

static inline void
sio2bp(const scan_io_t *sio, blkptr_t *bp)
{
	bzero(bp, sizeof (*bp));
	bp->blk_prop = sio->sio_blk_prop;
	bp->blk_phys_birth = sio->sio_phys_birth;
	bp->blk_birth = sio->sio_birth;
	bp->blk_fill = 1;	/* we always only work with data pointers */
	bp->blk_cksum = sio->sio_cksum;

	ASSERT3U(sio->sio_nr_dvas, >, 0);
	ASSERT3U(sio->sio_nr_dvas, <=, SPA_DVAS_PER_BP);

	bcopy(sio->sio_dva, bp->blk_dva, sio->sio_nr_dvas * sizeof (dva_t));
}

static inline void
bp2sio(const blkptr_t *bp, scan_io_t *sio, int dva_i)
{
	sio->sio_blk_prop = bp->blk_prop;
	sio->sio_phys_birth = bp->blk_phys_birth;
	sio->sio_birth = bp->blk_birth;
	sio->sio_cksum = bp->blk_cksum;
	sio->sio_nr_dvas = BP_GET_NDVAS(bp);

	/*
	 * Copy the DVAs to the sio. We need all copies of the block so
	 * that the self healing code can use the alternate copies if the
	 * first is corrupted. We want the DVA at index dva_i to be first
	 * in the sio since this is the primary one that we want to issue.
	 */
	for (int i = 0, j = dva_i; i < sio->sio_nr_dvas; i++, j++) {
		sio->sio_dva[i] = bp->blk_dva[j % sio->sio_nr_dvas];
	}
}

int
dsl_scan_init(dsl_pool_t *dp, uint64_t txg)
{
	int err;
	dsl_scan_t *scn;
	spa_t *spa = dp->dp_spa;
	uint64_t f;

	scn = dp->dp_scan = kmem_zalloc(sizeof (dsl_scan_t), KM_SLEEP);
	scn->scn_dp = dp;

	/*
	 * It's possible that we're resuming a scan after a reboot so
	 * make sure that the scan_async_destroying flag is initialized
	 * appropriately.
	 */
	ASSERT(!scn->scn_async_destroying);
	scn->scn_async_destroying = spa_feature_is_active(dp->dp_spa,
	    SPA_FEATURE_ASYNC_DESTROY);

	/*
	 * Calculate the max number of in-flight bytes for pool-wide
	 * scanning operations (minimum 1MB). Limits for the issuing
	 * phase are done per top-level vdev and are handled separately.
	 */
	scn->scn_maxinflight_bytes = MAX(zfs_scan_vdev_limit *
	    dsl_scan_count_leaves(spa->spa_root_vdev), 1ULL << 20);

	avl_create(&scn->scn_queue, scan_ds_queue_compare, sizeof (scan_ds_t),
	    offsetof(scan_ds_t, sds_node));
	avl_create(&scn->scn_prefetch_queue, scan_prefetch_queue_compare,
	    sizeof (scan_prefetch_issue_ctx_t),
	    offsetof(scan_prefetch_issue_ctx_t, spic_avl_node));

	err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	    "scrub_func", sizeof (uint64_t), 1, &f);
	if (err == 0) {
		/*
		 * There was an old-style scrub in progress.  Restart a
		 * new-style scrub from the beginning.
		 */
		scn->scn_restart_txg = txg;
		zfs_dbgmsg("old-style scrub was in progress; "
		    "restarting new-style scrub in txg %llu",
		    (longlong_t)scn->scn_restart_txg);

		/*
		 * Load the queue obj from the old location so that it
		 * can be freed by dsl_scan_done().
		 */
		(void) zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
		    "scrub_queue", sizeof (uint64_t), 1,
		    &scn->scn_phys.scn_queue_obj);
	} else {
		err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
		    DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
		    &scn->scn_phys);
		/*
		 * Detect if the pool contains the signature of #2094.  If it
		 * does properly update the scn->scn_phys structure and notify
		 * the administrator by setting an errata for the pool.
		 */
		if (err == EOVERFLOW) {
			uint64_t zaptmp[SCAN_PHYS_NUMINTS + 1];
			VERIFY3S(SCAN_PHYS_NUMINTS, ==, 24);
			VERIFY3S(offsetof(dsl_scan_phys_t, scn_flags), ==,
			    (23 * sizeof (uint64_t)));

			err = zap_lookup(dp->dp_meta_objset,
			    DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SCAN,
			    sizeof (uint64_t), SCAN_PHYS_NUMINTS + 1, &zaptmp);
			if (err == 0) {
				uint64_t overflow = zaptmp[SCAN_PHYS_NUMINTS];

				if (overflow & ~DSL_SCAN_FLAGS_MASK ||
				    scn->scn_async_destroying) {
					spa->spa_errata =
					    ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY;
					return (EOVERFLOW);
				}

				bcopy(zaptmp, &scn->scn_phys,
				    SCAN_PHYS_NUMINTS * sizeof (uint64_t));
				scn->scn_phys.scn_flags = overflow;

				/* Required scrub already in progress. */
				if (scn->scn_phys.scn_state == DSS_FINISHED ||
				    scn->scn_phys.scn_state == DSS_CANCELED)
					spa->spa_errata =
					    ZPOOL_ERRATA_ZOL_2094_SCRUB;
			}
		}

		if (err == ENOENT)
			return (0);
		else if (err)
			return (err);

		/*
		 * We might be restarting after a reboot, so jump the issued
		 * counter to how far we've scanned. We know we're consistent
		 * up to here.
		 */
		scn->scn_issued_before_pass = scn->scn_phys.scn_examined;

		if (dsl_scan_is_running(scn) &&
		    spa_prev_software_version(dp->dp_spa) < SPA_VERSION_SCAN) {
			/*
			 * A new-type scrub was in progress on an old
			 * pool, and the pool was accessed by old
			 * software.  Restart from the beginning, since
			 * the old software may have changed the pool in
			 * the meantime.
			 */
			scn->scn_restart_txg = txg;
			zfs_dbgmsg("new-style scrub was modified "
			    "by old software; restarting in txg %llu",
			    (longlong_t)scn->scn_restart_txg);
		}
	}

	bcopy(&scn->scn_phys, &scn->scn_phys_cached, sizeof (scn->scn_phys));

	/* reload the queue into the in-core state */
	if (scn->scn_phys.scn_queue_obj != 0) {
		zap_cursor_t zc;
		zap_attribute_t za;

		for (zap_cursor_init(&zc, dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj);
		    zap_cursor_retrieve(&zc, &za) == 0;
		    (void) zap_cursor_advance(&zc)) {
			scan_ds_queue_insert(scn,
			    zfs_strtonum(za.za_name, NULL),
			    za.za_first_integer);
		}
		zap_cursor_fini(&zc);
	}

	spa_scan_stat_init(spa);
	return (0);
}

void
dsl_scan_fini(dsl_pool_t *dp)
{
	if (dp->dp_scan != NULL) {
		dsl_scan_t *scn = dp->dp_scan;

		if (scn->scn_taskq != NULL)
			taskq_destroy(scn->scn_taskq);

		scan_ds_queue_clear(scn);
		avl_destroy(&scn->scn_queue);
		scan_ds_prefetch_queue_clear(scn);
		avl_destroy(&scn->scn_prefetch_queue);

		kmem_free(dp->dp_scan, sizeof (dsl_scan_t));
		dp->dp_scan = NULL;
	}
}

static boolean_t
dsl_scan_restarting(dsl_scan_t *scn, dmu_tx_t *tx)
{
	return (scn->scn_restart_txg != 0 &&
	    scn->scn_restart_txg <= tx->tx_txg);
}

boolean_t
dsl_scan_scrubbing(const dsl_pool_t *dp)
{
	dsl_scan_phys_t *scn_phys = &dp->dp_scan->scn_phys;

	return (scn_phys->scn_state == DSS_SCANNING &&
	    scn_phys->scn_func == POOL_SCAN_SCRUB);
}

boolean_t
dsl_scan_is_paused_scrub(const dsl_scan_t *scn)
{
	return (dsl_scan_scrubbing(scn->scn_dp) &&
	    scn->scn_phys.scn_flags & DSF_SCRUB_PAUSED);
}

/*
 * Writes out a persistent dsl_scan_phys_t record to the pool directory.
 * Because we can be running in the block sorting algorithm, we do not always
 * want to write out the record, only when it is "safe" to do so. This safety
 * condition is achieved by making sure that the sorting queues are empty
 * (scn_bytes_pending == 0). When this condition is not true, the sync'd state
 * is inconsistent with how much actual scanning progress has been made. The
 * kind of sync to be performed is specified by the sync_type argument. If the
 * sync is optional, we only sync if the queues are empty. If the sync is
 * mandatory, we do a hard ASSERT to make sure that the queues are empty. The
 * third possible state is a "cached" sync. This is done in response to:
 * 1) The dataset that was in the last sync'd dsl_scan_phys_t having been
 *	destroyed, so we wouldn't be able to restart scanning from it.
 * 2) The snapshot that was in the last sync'd dsl_scan_phys_t having been
 *	superseded by a newer snapshot.
 * 3) The dataset that was in the last sync'd dsl_scan_phys_t having been
 *	swapped with its clone.
 * In all cases, a cached sync simply rewrites the last record we've written,
 * just slightly modified. For the modifications that are performed to the
 * last written dsl_scan_phys_t, see dsl_scan_ds_destroyed,
 * dsl_scan_ds_snapshotted and dsl_scan_ds_clone_swapped.
 */
static void
dsl_scan_sync_state(dsl_scan_t *scn, dmu_tx_t *tx, state_sync_type_t sync_type)
{
	int i;
	spa_t *spa = scn->scn_dp->dp_spa;

	ASSERT(sync_type != SYNC_MANDATORY || scn->scn_bytes_pending == 0);
	if (scn->scn_bytes_pending == 0) {
		for (i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
			vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
			dsl_scan_io_queue_t *q = vd->vdev_scan_io_queue;

			if (q == NULL)
				continue;

			mutex_enter(&vd->vdev_scan_io_queue_lock);
			ASSERT3P(avl_first(&q->q_sios_by_addr), ==, NULL);
			ASSERT3P(avl_first(&q->q_exts_by_size), ==, NULL);
			ASSERT3P(range_tree_first(q->q_exts_by_addr), ==, NULL);
			mutex_exit(&vd->vdev_scan_io_queue_lock);
		}

		if (scn->scn_phys.scn_queue_obj != 0)
			scan_ds_queue_sync(scn, tx);
		VERIFY0(zap_update(scn->scn_dp->dp_meta_objset,
		    DMU_POOL_DIRECTORY_OBJECT,
		    DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
		    &scn->scn_phys, tx));
		bcopy(&scn->scn_phys, &scn->scn_phys_cached,
		    sizeof (scn->scn_phys));

		if (scn->scn_checkpointing)
			zfs_dbgmsg("finish scan checkpoint");

		scn->scn_checkpointing = B_FALSE;
		scn->scn_last_checkpoint = ddi_get_lbolt();
	} else if (sync_type == SYNC_CACHED) {
		VERIFY0(zap_update(scn->scn_dp->dp_meta_objset,
		    DMU_POOL_DIRECTORY_OBJECT,
		    DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
		    &scn->scn_phys_cached, tx));
	}
}

/* ARGSUSED */
static int
dsl_scan_setup_check(void *arg, dmu_tx_t *tx)
{
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;

	if (dsl_scan_is_running(scn))
		return (SET_ERROR(EBUSY));

	return (0);
}

static void
dsl_scan_setup_sync(void *arg, dmu_tx_t *tx)
{
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
	pool_scan_func_t *funcp = arg;
	dmu_object_type_t ot = 0;
	dsl_pool_t *dp = scn->scn_dp;
	spa_t *spa = dp->dp_spa;

	ASSERT(!dsl_scan_is_running(scn));
	ASSERT(*funcp > POOL_SCAN_NONE && *funcp < POOL_SCAN_FUNCS);
	bzero(&scn->scn_phys, sizeof (scn->scn_phys));
	scn->scn_phys.scn_func = *funcp;
	scn->scn_phys.scn_state = DSS_SCANNING;
	scn->scn_phys.scn_min_txg = 0;
	scn->scn_phys.scn_max_txg = tx->tx_txg;
	scn->scn_phys.scn_ddt_class_max = DDT_CLASSES - 1; /* the entire DDT */
	scn->scn_phys.scn_start_time = gethrestime_sec();
	scn->scn_phys.scn_errors = 0;
	scn->scn_phys.scn_to_examine = spa->spa_root_vdev->vdev_stat.vs_alloc;
	scn->scn_issued_before_pass = 0;
	scn->scn_restart_txg = 0;
	scn->scn_done_txg = 0;
	scn->scn_last_checkpoint = 0;
	scn->scn_checkpointing = B_FALSE;
	spa_scan_stat_init(spa);

	if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
		scn->scn_phys.scn_ddt_class_max = zfs_scrub_ddt_class_max;

		/* rewrite all disk labels */
		vdev_config_dirty(spa->spa_root_vdev);

		if (vdev_resilver_needed(spa->spa_root_vdev,
		    &scn->scn_phys.scn_min_txg, &scn->scn_phys.scn_max_txg)) {
			spa_event_notify(spa, NULL, NULL,
			    ESC_ZFS_RESILVER_START);
		} else {
			spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_START);
		}

		spa->spa_scrub_started = B_TRUE;
		/*
		 * If this is an incremental scrub, limit the DDT scrub phase
		 * to just the auto-ditto class (for correctness); the rest
		 * of the scrub should go faster using top-down pruning.
		 */
		if (scn->scn_phys.scn_min_txg > TXG_INITIAL)
			scn->scn_phys.scn_ddt_class_max = DDT_CLASS_DITTO;

	}

	/* back to the generic stuff */

	if (dp->dp_blkstats == NULL) {
		dp->dp_blkstats =
		    vmem_alloc(sizeof (zfs_all_blkstats_t), KM_SLEEP);
		mutex_init(&dp->dp_blkstats->zab_lock, NULL,
		    MUTEX_DEFAULT, NULL);
	}
	bzero(&dp->dp_blkstats->zab_type, sizeof (dp->dp_blkstats->zab_type));

	if (spa_version(spa) < SPA_VERSION_DSL_SCRUB)
		ot = DMU_OT_ZAP_OTHER;

	scn->scn_phys.scn_queue_obj = zap_create(dp->dp_meta_objset,
	    ot ? ot : DMU_OT_SCAN_QUEUE, DMU_OT_NONE, 0, tx);

	bcopy(&scn->scn_phys, &scn->scn_phys_cached, sizeof (scn->scn_phys));

	dsl_scan_sync_state(scn, tx, SYNC_MANDATORY);

	spa_history_log_internal(spa, "scan setup", tx,
	    "func=%u mintxg=%llu maxtxg=%llu",
	    *funcp, scn->scn_phys.scn_min_txg, scn->scn_phys.scn_max_txg);
}

/*
 * Called by the ZFS_IOC_POOL_SCAN ioctl to start a scrub or resilver.
 * Can also be called to resume a paused scrub.
 */
int
dsl_scan(dsl_pool_t *dp, pool_scan_func_t func)
{
	spa_t *spa = dp->dp_spa;
	dsl_scan_t *scn = dp->dp_scan;

	/*
	 * Purge all vdev caches and probe all devices.  We do this here
	 * rather than in sync context because this requires a writer lock
	 * on the spa_config lock, which we can't do from sync context.  The
	 * spa_scrub_reopen flag indicates that vdev_open() should not
	 * attempt to start another scrub.
	 */
	spa_vdev_state_enter(spa, SCL_NONE);
	spa->spa_scrub_reopen = B_TRUE;
	vdev_reopen(spa->spa_root_vdev);
	spa->spa_scrub_reopen = B_FALSE;
	(void) spa_vdev_state_exit(spa, NULL, 0);

	if (func == POOL_SCAN_RESILVER) {
		dsl_resilver_restart(spa->spa_dsl_pool, 0);
		return (0);
	}

	if (func == POOL_SCAN_SCRUB && dsl_scan_is_paused_scrub(scn)) {
		/* got scrub start cmd, resume paused scrub */
		int err = dsl_scrub_set_pause_resume(scn->scn_dp,
		    POOL_SCRUB_NORMAL);
		if (err == 0) {
			spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_RESUME);
			return (ECANCELED);
		}

		return (SET_ERROR(err));
	}

	return (dsl_sync_task(spa_name(spa), dsl_scan_setup_check,
	    dsl_scan_setup_sync, &func, 0, ZFS_SPACE_CHECK_EXTRA_RESERVED));
}

/*
 * Sets the resilver defer flag to B_FALSE on all leaf devs under vd. Returns
 * B_TRUE if we have devices that need to be resilvered and are available to
 * accept resilver I/Os.
 */
static boolean_t
dsl_scan_clear_deferred(vdev_t *vd, dmu_tx_t *tx)
{
	boolean_t resilver_needed = B_FALSE;
	spa_t *spa = vd->vdev_spa;

	for (int c = 0; c < vd->vdev_children; c++) {
		resilver_needed |=
		    dsl_scan_clear_deferred(vd->vdev_child[c], tx);
	}

	if (vd == spa->spa_root_vdev &&
	    spa_feature_is_active(spa, SPA_FEATURE_RESILVER_DEFER)) {
		spa_feature_decr(spa, SPA_FEATURE_RESILVER_DEFER, tx);
		vdev_config_dirty(vd);
		spa->spa_resilver_deferred = B_FALSE;
		return (resilver_needed);
	}

	if (!vdev_is_concrete(vd) || vd->vdev_aux ||
	    !vd->vdev_ops->vdev_op_leaf)
		return (resilver_needed);

	if (vd->vdev_resilver_deferred)
		vd->vdev_resilver_deferred = B_FALSE;

	return (!vdev_is_dead(vd) && !vd->vdev_offline &&
	    vdev_resilver_needed(vd, NULL, NULL));
}

/* ARGSUSED */
static void
dsl_scan_done(dsl_scan_t *scn, boolean_t complete, dmu_tx_t *tx)
{
	static const char *old_names[] = {
		"scrub_bookmark",
		"scrub_ddt_bookmark",
		"scrub_ddt_class_max",
		"scrub_queue",
		"scrub_min_txg",
		"scrub_max_txg",
		"scrub_func",
		"scrub_errors",
		NULL
	};

	dsl_pool_t *dp = scn->scn_dp;
	spa_t *spa = dp->dp_spa;
	int i;

	/* Remove any remnants of an old-style scrub. */
	for (i = 0; old_names[i]; i++) {
		(void) zap_remove(dp->dp_meta_objset,
		    DMU_POOL_DIRECTORY_OBJECT, old_names[i], tx);
	}

	if (scn->scn_phys.scn_queue_obj != 0) {
		VERIFY0(dmu_object_free(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, tx));
		scn->scn_phys.scn_queue_obj = 0;
	}
	scan_ds_queue_clear(scn);
	scan_ds_prefetch_queue_clear(scn);

	scn->scn_phys.scn_flags &= ~DSF_SCRUB_PAUSED;

	/*
	 * If we were "restarted" from a stopped state, don't bother
	 * with anything else.
	 */
	if (!dsl_scan_is_running(scn)) {
		ASSERT(!scn->scn_is_sorted);
		return;
	}

	if (scn->scn_is_sorted) {
		scan_io_queues_destroy(scn);
		scn->scn_is_sorted = B_FALSE;

		if (scn->scn_taskq != NULL) {
			taskq_destroy(scn->scn_taskq);
			scn->scn_taskq = NULL;
		}
	}

	scn->scn_phys.scn_state = complete ? DSS_FINISHED : DSS_CANCELED;

	if (dsl_scan_restarting(scn, tx))
		spa_history_log_internal(spa, "scan aborted, restarting", tx,
		    "errors=%llu", spa_get_errlog_size(spa));
	else if (!complete)
		spa_history_log_internal(spa, "scan cancelled", tx,
		    "errors=%llu", spa_get_errlog_size(spa));
	else
		spa_history_log_internal(spa, "scan done", tx,
		    "errors=%llu", spa_get_errlog_size(spa));

	if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
		spa->spa_scrub_started = B_FALSE;
		spa->spa_scrub_active = B_FALSE;

		/*
		 * If the scrub/resilver completed, update all DTLs to
		 * reflect this.  Whether it succeeded or not, vacate
		 * all temporary scrub DTLs.
		 *
		 * As the scrub does not currently support traversing
		 * data that have been freed but are part of a checkpoint,
		 * we don't mark the scrub as done in the DTLs as faults
		 * may still exist in those vdevs.
		 */
		if (complete &&
		    !spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
			vdev_dtl_reassess(spa->spa_root_vdev, tx->tx_txg,
			    scn->scn_phys.scn_max_txg, B_TRUE);

			spa_event_notify(spa, NULL, NULL,
			    scn->scn_phys.scn_min_txg ?
			    ESC_ZFS_RESILVER_FINISH : ESC_ZFS_SCRUB_FINISH);
		} else {
			vdev_dtl_reassess(spa->spa_root_vdev, tx->tx_txg,
			    0, B_TRUE);
		}
		spa_errlog_rotate(spa);

		/*
		 * We may have finished replacing a device.
		 * Let the async thread assess this and handle the detach.
		 */
		spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);

		/*
		 * Clear any deferred_resilver flags in the config.
		 * If there are drives that need resilvering, kick
		 * off an asynchronous request to start resilver.
		 * dsl_scan_clear_deferred() may update the config
		 * before the resilver can restart. In the event of
		 * a crash during this period, the spa loading code
		 * will find the drives that need to be resilvered
		 * when the machine reboots and start the resilver then.
		 */
		boolean_t resilver_needed =
		    dsl_scan_clear_deferred(spa->spa_root_vdev, tx);
		if (resilver_needed) {
			spa_history_log_internal(spa,
			    "starting deferred resilver", tx,
			    "errors=%llu", spa_get_errlog_size(spa));
			spa_async_request(spa, SPA_ASYNC_RESILVER);
		}
	}

	scn->scn_phys.scn_end_time = gethrestime_sec();

	if (spa->spa_errata == ZPOOL_ERRATA_ZOL_2094_SCRUB)
		spa->spa_errata = 0;

	ASSERT(!dsl_scan_is_running(scn));
}

/* ARGSUSED */
static int
dsl_scan_cancel_check(void *arg, dmu_tx_t *tx)
{
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;

	if (!dsl_scan_is_running(scn))
		return (SET_ERROR(ENOENT));
	return (0);
}

/* ARGSUSED */
static void
dsl_scan_cancel_sync(void *arg, dmu_tx_t *tx)
{
	dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;

	dsl_scan_done(scn, B_FALSE, tx);
	dsl_scan_sync_state(scn, tx, SYNC_MANDATORY);
	spa_event_notify(scn->scn_dp->dp_spa, NULL, NULL, ESC_ZFS_SCRUB_ABORT);
}

int
dsl_scan_cancel(dsl_pool_t *dp)
{
	return (dsl_sync_task(spa_name(dp->dp_spa), dsl_scan_cancel_check,
	    dsl_scan_cancel_sync, NULL, 3, ZFS_SPACE_CHECK_RESERVED));
}

static int
dsl_scrub_pause_resume_check(void *arg, dmu_tx_t *tx)
{
	pool_scrub_cmd_t *cmd = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_scan_t *scn = dp->dp_scan;

	if (*cmd == POOL_SCRUB_PAUSE) {
		/* can't pause a scrub when there is no in-progress scrub */
		if (!dsl_scan_scrubbing(dp))
			return (SET_ERROR(ENOENT));

		/* can't pause a paused scrub */
		if (dsl_scan_is_paused_scrub(scn))
			return (SET_ERROR(EBUSY));
	} else if (*cmd != POOL_SCRUB_NORMAL) {
		return (SET_ERROR(ENOTSUP));
	}

	return (0);
}

static void
dsl_scrub_pause_resume_sync(void *arg, dmu_tx_t *tx)
{
	pool_scrub_cmd_t *cmd = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	spa_t *spa = dp->dp_spa;
	dsl_scan_t *scn = dp->dp_scan;

	if (*cmd == POOL_SCRUB_PAUSE) {
		/* can't pause a scrub when there is no in-progress scrub */
		spa->spa_scan_pass_scrub_pause = gethrestime_sec();
		scn->scn_phys.scn_flags |= DSF_SCRUB_PAUSED;
		scn->scn_phys_cached.scn_flags |= DSF_SCRUB_PAUSED;
		dsl_scan_sync_state(scn, tx, SYNC_CACHED);
		spa_event_notify(spa, NULL, NULL, ESC_ZFS_SCRUB_PAUSED);
	} else {
		ASSERT3U(*cmd, ==, POOL_SCRUB_NORMAL);
		if (dsl_scan_is_paused_scrub(scn)) {
			/*
			 * We need to keep track of how much time we spend
			 * paused per pass so that we can adjust the scrub rate
			 * shown in the output of 'zpool status'
			 */
			spa->spa_scan_pass_scrub_spent_paused +=
			    gethrestime_sec() - spa->spa_scan_pass_scrub_pause;
			spa->spa_scan_pass_scrub_pause = 0;
			scn->scn_phys.scn_flags &= ~DSF_SCRUB_PAUSED;
			scn->scn_phys_cached.scn_flags &= ~DSF_SCRUB_PAUSED;
			dsl_scan_sync_state(scn, tx, SYNC_CACHED);
		}
	}
}

/*
 * Set scrub pause/resume state if it makes sense to do so
 */
int
dsl_scrub_set_pause_resume(const dsl_pool_t *dp, pool_scrub_cmd_t cmd)
{
	return (dsl_sync_task(spa_name(dp->dp_spa),
	    dsl_scrub_pause_resume_check, dsl_scrub_pause_resume_sync, &cmd, 3,
	    ZFS_SPACE_CHECK_RESERVED));
}


/* start a new scan, or restart an existing one. */
void
dsl_resilver_restart(dsl_pool_t *dp, uint64_t txg)
{
	if (txg == 0) {
		dmu_tx_t *tx;
		tx = dmu_tx_create_dd(dp->dp_mos_dir);
		VERIFY(0 == dmu_tx_assign(tx, TXG_WAIT));

		txg = dmu_tx_get_txg(tx);
		dp->dp_scan->scn_restart_txg = txg;
		dmu_tx_commit(tx);
	} else {
		dp->dp_scan->scn_restart_txg = txg;
	}
	zfs_dbgmsg("restarting resilver txg=%llu", (longlong_t)txg);
}

void
dsl_free(dsl_pool_t *dp, uint64_t txg, const blkptr_t *bp)
{
	zio_free(dp->dp_spa, txg, bp);
}

void
dsl_free_sync(zio_t *pio, dsl_pool_t *dp, uint64_t txg, const blkptr_t *bpp)
{
	ASSERT(dsl_pool_sync_context(dp));
	zio_nowait(zio_free_sync(pio, dp->dp_spa, txg, bpp, pio->io_flags));
}

static int
scan_ds_queue_compare(const void *a, const void *b)
{
	const scan_ds_t *sds_a = a, *sds_b = b;

	if (sds_a->sds_dsobj < sds_b->sds_dsobj)
		return (-1);
	if (sds_a->sds_dsobj == sds_b->sds_dsobj)
		return (0);
	return (1);
}

static void
scan_ds_queue_clear(dsl_scan_t *scn)
{
	void *cookie = NULL;
	scan_ds_t *sds;
	while ((sds = avl_destroy_nodes(&scn->scn_queue, &cookie)) != NULL) {
		kmem_free(sds, sizeof (*sds));
	}
}

static boolean_t
scan_ds_queue_contains(dsl_scan_t *scn, uint64_t dsobj, uint64_t *txg)
{
	scan_ds_t srch, *sds;

	srch.sds_dsobj = dsobj;
	sds = avl_find(&scn->scn_queue, &srch, NULL);
	if (sds != NULL && txg != NULL)
		*txg = sds->sds_txg;
	return (sds != NULL);
}

static void
scan_ds_queue_insert(dsl_scan_t *scn, uint64_t dsobj, uint64_t txg)
{
	scan_ds_t *sds;
	avl_index_t where;

	sds = kmem_zalloc(sizeof (*sds), KM_SLEEP);
	sds->sds_dsobj = dsobj;
	sds->sds_txg = txg;

	VERIFY3P(avl_find(&scn->scn_queue, sds, &where), ==, NULL);
	avl_insert(&scn->scn_queue, sds, where);
}

static void
scan_ds_queue_remove(dsl_scan_t *scn, uint64_t dsobj)
{
	scan_ds_t srch, *sds;

	srch.sds_dsobj = dsobj;

	sds = avl_find(&scn->scn_queue, &srch, NULL);
	VERIFY(sds != NULL);
	avl_remove(&scn->scn_queue, sds);
	kmem_free(sds, sizeof (*sds));
}

static void
scan_ds_queue_sync(dsl_scan_t *scn, dmu_tx_t *tx)
{
	dsl_pool_t *dp = scn->scn_dp;
	spa_t *spa = dp->dp_spa;
	dmu_object_type_t ot = (spa_version(spa) >= SPA_VERSION_DSL_SCRUB) ?
	    DMU_OT_SCAN_QUEUE : DMU_OT_ZAP_OTHER;

	ASSERT0(scn->scn_bytes_pending);
	ASSERT(scn->scn_phys.scn_queue_obj != 0);

	VERIFY0(dmu_object_free(dp->dp_meta_objset,
	    scn->scn_phys.scn_queue_obj, tx));
	scn->scn_phys.scn_queue_obj = zap_create(dp->dp_meta_objset, ot,
	    DMU_OT_NONE, 0, tx);
	for (scan_ds_t *sds = avl_first(&scn->scn_queue);
	    sds != NULL; sds = AVL_NEXT(&scn->scn_queue, sds)) {
		VERIFY0(zap_add_int_key(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, sds->sds_dsobj,
		    sds->sds_txg, tx));
	}
}

/*
 * Computes the memory limit state that we're currently in. A sorted scan
 * needs quite a bit of memory to hold the sorting queue, so we need to
 * reasonably constrain the size so it doesn't impact overall system
 * performance. We compute two limits:
 * 1) Hard memory limit: if the amount of memory used by the sorting
 *	queues on a pool gets above this value, we stop the metadata
 *	scanning portion and start issuing the queued up and sorted
 *	I/Os to reduce memory usage.
 *	This limit is calculated as a fraction of physmem (by default 5%).
 *	We constrain the lower bound of the hard limit to an absolute
 *	minimum of zfs_scan_mem_lim_min (default: 16 MiB). We also constrain
 *	the upper bound to 5% of the total pool size - no chance we'll
 *	ever need that much memory, but just to keep the value in check.
 * 2) Soft memory limit: once we hit the hard memory limit, we start
 *	issuing I/O to reduce queue memory usage, but we don't want to
 *	completely empty out the queues, since we might be able to find I/Os
 *	that will fill in the gaps of our non-sequential IOs at some point
 *	in the future. So we stop the issuing of I/Os once the amount of
 *	memory used drops below the soft limit (at which point we stop issuing
 *	I/O and start scanning metadata again).
 *
 *	This limit is calculated by subtracting a fraction of the hard
 *	limit from the hard limit. By default this fraction is 5%, so
 *	the soft limit is 95% of the hard limit. We cap the size of the
 *	difference between the hard and soft limits at an absolute
 *	maximum of zfs_scan_mem_lim_soft_max (default: 128 MiB) - this is
 *	sufficient to not cause too frequent switching between the
 *	metadata scan and I/O issue (even at 2k recordsize, 128 MiB's
 *	worth of queues is about 1.2 GiB of on-pool data, so scanning
 *	that should take at least a decent fraction of a second).
 */
static boolean_t
dsl_scan_should_clear(dsl_scan_t *scn)
{
	vdev_t *rvd = scn->scn_dp->dp_spa->spa_root_vdev;
	uint64_t mlim_hard, mlim_soft, mused;
	uint64_t alloc = metaslab_class_get_alloc(spa_normal_class(
	    scn->scn_dp->dp_spa));

	mlim_hard = MAX((physmem / zfs_scan_mem_lim_fact) * PAGESIZE,
	    zfs_scan_mem_lim_min);
	mlim_hard = MIN(mlim_hard, alloc / 20);
	mlim_soft = mlim_hard - MIN(mlim_hard / zfs_scan_mem_lim_soft_fact,
	    zfs_scan_mem_lim_soft_max);
	mused = 0;
	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
		vdev_t *tvd = rvd->vdev_child[i];
		dsl_scan_io_queue_t *queue;

		mutex_enter(&tvd->vdev_scan_io_queue_lock);
		queue = tvd->vdev_scan_io_queue;
		if (queue != NULL) {
			/* # extents in exts_by_size = # in exts_by_addr */
			mused += avl_numnodes(&queue->q_exts_by_size) *
			    sizeof (range_seg_t) + queue->q_sio_memused;
		}
		mutex_exit(&tvd->vdev_scan_io_queue_lock);
	}

	dprintf("current scan memory usage: %llu bytes\n", (longlong_t)mused);

	if (mused == 0)
		ASSERT0(scn->scn_bytes_pending);

	/*
	 * If we are above our hard limit, we need to clear out memory.
	 * If we are below our soft limit, we need to accumulate sequential IOs.
	 * Otherwise, we should keep doing whatever we are currently doing.
	 */
	if (mused >= mlim_hard)
		return (B_TRUE);
	else if (mused < mlim_soft)
		return (B_FALSE);
	else
		return (scn->scn_clearing);
}

static boolean_t
dsl_scan_check_suspend(dsl_scan_t *scn, const zbookmark_phys_t *zb)
{
	/* we never skip user/group accounting objects */
	if (zb && (int64_t)zb->zb_object < 0)
		return (B_FALSE);

	if (scn->scn_suspending)
		return (B_TRUE); /* we're already suspending */

	if (!ZB_IS_ZERO(&scn->scn_phys.scn_bookmark))
		return (B_FALSE); /* we're resuming */

	/* We only know how to resume from level-0 blocks. */
	if (zb && zb->zb_level != 0)
		return (B_FALSE);

	/*
	 * We suspend if:
	 *  - we have scanned for at least the minimum time (default 1 sec
	 *    for scrub, 3 sec for resilver), and either we have sufficient
	 *    dirty data that we are starting to write more quickly
	 *    (default 30%), someone is explicitly waiting for this txg
	 *    to complete, or we have used up all of the time in the txg
	 *    timeout (default 5 sec).
	 *  or
	 *  - the spa is shutting down because this pool is being exported
	 *    or the machine is rebooting.
	 *  or
	 *  - the scan queue has reached its memory use limit
	 */
	uint64_t curr_time_ns = gethrtime();
	uint64_t scan_time_ns = curr_time_ns - scn->scn_sync_start_time;
	uint64_t sync_time_ns = curr_time_ns -
	    scn->scn_dp->dp_spa->spa_sync_starttime;
	int dirty_pct = scn->scn_dp->dp_dirty_total * 100 / zfs_dirty_data_max;
	int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
	    zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;

	if ((NSEC2MSEC(scan_time_ns) > mintime &&
	    (dirty_pct >= zfs_vdev_async_write_active_min_dirty_percent ||
	    txg_sync_waiting(scn->scn_dp) ||
	    NSEC2SEC(sync_time_ns) >= zfs_txg_timeout)) ||
	    spa_shutting_down(scn->scn_dp->dp_spa) ||
	    (zfs_scan_strict_mem_lim && dsl_scan_should_clear(scn))) {
		if (zb) {
			dprintf("suspending at bookmark %llx/%llx/%llx/%llx\n",
			    (longlong_t)zb->zb_objset,
			    (longlong_t)zb->zb_object,
			    (longlong_t)zb->zb_level,
			    (longlong_t)zb->zb_blkid);
			scn->scn_phys.scn_bookmark = *zb;
		} else {
#ifdef ZFS_DEBUG
			dsl_scan_phys_t *scnp = &scn->scn_phys;
			dprintf("suspending at at DDT bookmark "
			    "%llx/%llx/%llx/%llx\n",
			    (longlong_t)scnp->scn_ddt_bookmark.ddb_class,
			    (longlong_t)scnp->scn_ddt_bookmark.ddb_type,
			    (longlong_t)scnp->scn_ddt_bookmark.ddb_checksum,
			    (longlong_t)scnp->scn_ddt_bookmark.ddb_cursor);
#endif
		}
		scn->scn_suspending = B_TRUE;
		return (B_TRUE);
	}
	return (B_FALSE);
}

typedef struct zil_scan_arg {
	dsl_pool_t	*zsa_dp;
	zil_header_t	*zsa_zh;
} zil_scan_arg_t;

/* ARGSUSED */
static int
dsl_scan_zil_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
{
	zil_scan_arg_t *zsa = arg;
	dsl_pool_t *dp = zsa->zsa_dp;
	dsl_scan_t *scn = dp->dp_scan;
	zil_header_t *zh = zsa->zsa_zh;
	zbookmark_phys_t zb;

	if (BP_IS_HOLE(bp) || bp->blk_birth <= scn->scn_phys.scn_cur_min_txg)
		return (0);

	/*
	 * One block ("stubby") can be allocated a long time ago; we
	 * want to visit that one because it has been allocated
	 * (on-disk) even if it hasn't been claimed (even though for
	 * scrub there's nothing to do to it).
	 */
	if (claim_txg == 0 && bp->blk_birth >= spa_min_claim_txg(dp->dp_spa))
		return (0);

	SET_BOOKMARK(&zb, zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET],
	    ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);

	VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb));
	return (0);
}

/* ARGSUSED */
static int
dsl_scan_zil_record(zilog_t *zilog, lr_t *lrc, void *arg, uint64_t claim_txg)
{
	if (lrc->lrc_txtype == TX_WRITE) {
		zil_scan_arg_t *zsa = arg;
		dsl_pool_t *dp = zsa->zsa_dp;
		dsl_scan_t *scn = dp->dp_scan;
		zil_header_t *zh = zsa->zsa_zh;
		lr_write_t *lr = (lr_write_t *)lrc;
		blkptr_t *bp = &lr->lr_blkptr;
		zbookmark_phys_t zb;

		if (BP_IS_HOLE(bp) ||
		    bp->blk_birth <= scn->scn_phys.scn_cur_min_txg)
			return (0);

		/*
		 * birth can be < claim_txg if this record's txg is
		 * already txg sync'ed (but this log block contains
		 * other records that are not synced)
		 */
		if (claim_txg == 0 || bp->blk_birth < claim_txg)
			return (0);

		SET_BOOKMARK(&zb, zh->zh_log.blk_cksum.zc_word[ZIL_ZC_OBJSET],
		    lr->lr_foid, ZB_ZIL_LEVEL,
		    lr->lr_offset / BP_GET_LSIZE(bp));

		VERIFY(0 == scan_funcs[scn->scn_phys.scn_func](dp, bp, &zb));
	}
	return (0);
}

static void
dsl_scan_zil(dsl_pool_t *dp, zil_header_t *zh)
{
	uint64_t claim_txg = zh->zh_claim_txg;
	zil_scan_arg_t zsa = { dp, zh };
	zilog_t *zilog;

	ASSERT(spa_writeable(dp->dp_spa));

	/*
	 * We only want to visit blocks that have been claimed but not yet
	 * replayed (or, in read-only mode, blocks that *would* be claimed).
	 */
	if (claim_txg == 0)
		return;

	zilog = zil_alloc(dp->dp_meta_objset, zh);

	(void) zil_parse(zilog, dsl_scan_zil_block, dsl_scan_zil_record, &zsa,
	    claim_txg, B_FALSE);

	zil_free(zilog);
}

/*
 * We compare scan_prefetch_issue_ctx_t's based on their bookmarks. The idea
 * here is to sort the AVL tree by the order each block will be needed.
 */
static int
scan_prefetch_queue_compare(const void *a, const void *b)
{
	const scan_prefetch_issue_ctx_t *spic_a = a, *spic_b = b;
	const scan_prefetch_ctx_t *spc_a = spic_a->spic_spc;
	const scan_prefetch_ctx_t *spc_b = spic_b->spic_spc;

	return (zbookmark_compare(spc_a->spc_datablkszsec,
	    spc_a->spc_indblkshift, spc_b->spc_datablkszsec,
	    spc_b->spc_indblkshift, &spic_a->spic_zb, &spic_b->spic_zb));
}

static void
scan_prefetch_ctx_rele(scan_prefetch_ctx_t *spc, void *tag)
{
	if (zfs_refcount_remove(&spc->spc_refcnt, tag) == 0) {
		zfs_refcount_destroy(&spc->spc_refcnt);
		kmem_free(spc, sizeof (scan_prefetch_ctx_t));
	}
}

static scan_prefetch_ctx_t *
scan_prefetch_ctx_create(dsl_scan_t *scn, dnode_phys_t *dnp, void *tag)
{
	scan_prefetch_ctx_t *spc;

	spc = kmem_alloc(sizeof (scan_prefetch_ctx_t), KM_SLEEP);
	zfs_refcount_create(&spc->spc_refcnt);
	zfs_refcount_add(&spc->spc_refcnt, tag);
	spc->spc_scn = scn;
	if (dnp != NULL) {
		spc->spc_datablkszsec = dnp->dn_datablkszsec;
		spc->spc_indblkshift = dnp->dn_indblkshift;
		spc->spc_root = B_FALSE;
	} else {
		spc->spc_datablkszsec = 0;
		spc->spc_indblkshift = 0;
		spc->spc_root = B_TRUE;
	}

	return (spc);
}

static void
scan_prefetch_ctx_add_ref(scan_prefetch_ctx_t *spc, void *tag)
{
	zfs_refcount_add(&spc->spc_refcnt, tag);
}

static void
scan_ds_prefetch_queue_clear(dsl_scan_t *scn)
{
	spa_t *spa = scn->scn_dp->dp_spa;
	void *cookie = NULL;
	scan_prefetch_issue_ctx_t *spic = NULL;

	mutex_enter(&spa->spa_scrub_lock);
	while ((spic = avl_destroy_nodes(&scn->scn_prefetch_queue,
	    &cookie)) != NULL) {
		scan_prefetch_ctx_rele(spic->spic_spc, scn);
		kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
	}
	mutex_exit(&spa->spa_scrub_lock);
}

static boolean_t
dsl_scan_check_prefetch_resume(scan_prefetch_ctx_t *spc,
    const zbookmark_phys_t *zb)
{
	zbookmark_phys_t *last_zb = &spc->spc_scn->scn_prefetch_bookmark;
	dnode_phys_t tmp_dnp;
	dnode_phys_t *dnp = (spc->spc_root) ? NULL : &tmp_dnp;

	if (zb->zb_objset != last_zb->zb_objset)
		return (B_TRUE);
	if ((int64_t)zb->zb_object < 0)
		return (B_FALSE);

	tmp_dnp.dn_datablkszsec = spc->spc_datablkszsec;
	tmp_dnp.dn_indblkshift = spc->spc_indblkshift;

	if (zbookmark_subtree_completed(dnp, zb, last_zb))
		return (B_TRUE);

	return (B_FALSE);
}

static void
dsl_scan_prefetch(scan_prefetch_ctx_t *spc, blkptr_t *bp, zbookmark_phys_t *zb)
{
	avl_index_t idx;
	dsl_scan_t *scn = spc->spc_scn;
	spa_t *spa = scn->scn_dp->dp_spa;
	scan_prefetch_issue_ctx_t *spic;

	if (zfs_no_scrub_prefetch)
		return;

	if (BP_IS_HOLE(bp) || bp->blk_birth <= scn->scn_phys.scn_cur_min_txg ||
	    (BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE &&
	    BP_GET_TYPE(bp) != DMU_OT_OBJSET))
		return;

	if (dsl_scan_check_prefetch_resume(spc, zb))
		return;

	scan_prefetch_ctx_add_ref(spc, scn);
	spic = kmem_alloc(sizeof (scan_prefetch_issue_ctx_t), KM_SLEEP);
	spic->spic_spc = spc;
	spic->spic_bp = *bp;
	spic->spic_zb = *zb;

	/*
	 * Add the IO to the queue of blocks to prefetch. This allows us to
	 * prioritize blocks that we will need first for the main traversal
	 * thread.
	 */
	mutex_enter(&spa->spa_scrub_lock);
	if (avl_find(&scn->scn_prefetch_queue, spic, &idx) != NULL) {
		/* this block is already queued for prefetch */
		kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
		scan_prefetch_ctx_rele(spc, scn);
		mutex_exit(&spa->spa_scrub_lock);
		return;
	}

	avl_insert(&scn->scn_prefetch_queue, spic, idx);
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&spa->spa_scrub_lock);
}

static void
dsl_scan_prefetch_dnode(dsl_scan_t *scn, dnode_phys_t *dnp,
    uint64_t objset, uint64_t object)
{
	int i;
	zbookmark_phys_t zb;
	scan_prefetch_ctx_t *spc;

	if (dnp->dn_nblkptr == 0 && !(dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR))
		return;

	SET_BOOKMARK(&zb, objset, object, 0, 0);

	spc = scan_prefetch_ctx_create(scn, dnp, FTAG);

	for (i = 0; i < dnp->dn_nblkptr; i++) {
		zb.zb_level = BP_GET_LEVEL(&dnp->dn_blkptr[i]);
		zb.zb_blkid = i;
		dsl_scan_prefetch(spc, &dnp->dn_blkptr[i], &zb);
	}

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
		zb.zb_level = 0;
		zb.zb_blkid = DMU_SPILL_BLKID;
		dsl_scan_prefetch(spc, DN_SPILL_BLKPTR(dnp), &zb);
	}

	scan_prefetch_ctx_rele(spc, FTAG);
}

void
dsl_scan_prefetch_cb(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp,
    arc_buf_t *buf, void *private)
{
	scan_prefetch_ctx_t *spc = private;
	dsl_scan_t *scn = spc->spc_scn;
	spa_t *spa = scn->scn_dp->dp_spa;

	/* broadcast that the IO has completed for rate limiting purposes */
	mutex_enter(&spa->spa_scrub_lock);
	ASSERT3U(spa->spa_scrub_inflight, >=, BP_GET_PSIZE(bp));
	spa->spa_scrub_inflight -= BP_GET_PSIZE(bp);
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&spa->spa_scrub_lock);

	/* if there was an error or we are done prefetching, just cleanup */
	if (buf == NULL || scn->scn_prefetch_stop)
		goto out;

	if (BP_GET_LEVEL(bp) > 0) {
		int i;
		blkptr_t *cbp;
		int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
		zbookmark_phys_t czb;

		for (i = 0, cbp = buf->b_data; i < epb; i++, cbp++) {
			SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
			    zb->zb_level - 1, zb->zb_blkid * epb + i);
			dsl_scan_prefetch(spc, cbp, &czb);
		}
	} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
		dnode_phys_t *cdnp;
		int i;
		int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;

		for (i = 0, cdnp = buf->b_data; i < epb;
		    i += cdnp->dn_extra_slots + 1,
		    cdnp += cdnp->dn_extra_slots + 1) {
			dsl_scan_prefetch_dnode(scn, cdnp,
			    zb->zb_objset, zb->zb_blkid * epb + i);
		}
	} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
		objset_phys_t *osp = buf->b_data;

		dsl_scan_prefetch_dnode(scn, &osp->os_meta_dnode,
		    zb->zb_objset, DMU_META_DNODE_OBJECT);

		if (OBJSET_BUF_HAS_USERUSED(buf)) {
			dsl_scan_prefetch_dnode(scn,
			    &osp->os_groupused_dnode, zb->zb_objset,
			    DMU_GROUPUSED_OBJECT);
			dsl_scan_prefetch_dnode(scn,
			    &osp->os_userused_dnode, zb->zb_objset,
			    DMU_USERUSED_OBJECT);
		}
	}

out:
	if (buf != NULL)
		arc_buf_destroy(buf, private);
	scan_prefetch_ctx_rele(spc, scn);
}

/* ARGSUSED */
static void
dsl_scan_prefetch_thread(void *arg)
{
	dsl_scan_t *scn = arg;
	spa_t *spa = scn->scn_dp->dp_spa;
	scan_prefetch_issue_ctx_t *spic;

	/* loop until we are told to stop */
	while (!scn->scn_prefetch_stop) {
		arc_flags_t flags = ARC_FLAG_NOWAIT |
		    ARC_FLAG_PRESCIENT_PREFETCH | ARC_FLAG_PREFETCH;
		int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SCAN_THREAD;

		mutex_enter(&spa->spa_scrub_lock);

		/*
		 * Wait until we have an IO to issue and are not above our
		 * maximum in flight limit.
		 */
		while (!scn->scn_prefetch_stop &&
		    (avl_numnodes(&scn->scn_prefetch_queue) == 0 ||
		    spa->spa_scrub_inflight >= scn->scn_maxinflight_bytes)) {
			cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
		}

		/* recheck if we should stop since we waited for the cv */
		if (scn->scn_prefetch_stop) {
			mutex_exit(&spa->spa_scrub_lock);
			break;
		}

		/* remove the prefetch IO from the tree */
		spic = avl_first(&scn->scn_prefetch_queue);
		spa->spa_scrub_inflight += BP_GET_PSIZE(&spic->spic_bp);
		avl_remove(&scn->scn_prefetch_queue, spic);

		mutex_exit(&spa->spa_scrub_lock);

		if (BP_IS_PROTECTED(&spic->spic_bp)) {
			ASSERT(BP_GET_TYPE(&spic->spic_bp) == DMU_OT_DNODE ||
			    BP_GET_TYPE(&spic->spic_bp) == DMU_OT_OBJSET);
			ASSERT3U(BP_GET_LEVEL(&spic->spic_bp), ==, 0);
			zio_flags |= ZIO_FLAG_RAW;
		}

		/* issue the prefetch asynchronously */
		(void) arc_read(scn->scn_zio_root, scn->scn_dp->dp_spa,
		    &spic->spic_bp, dsl_scan_prefetch_cb, spic->spic_spc,
		    ZIO_PRIORITY_SCRUB, zio_flags, &flags, &spic->spic_zb);

		kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
	}

	ASSERT(scn->scn_prefetch_stop);

	/* free any prefetches we didn't get to complete */
	mutex_enter(&spa->spa_scrub_lock);
	while ((spic = avl_first(&scn->scn_prefetch_queue)) != NULL) {
		avl_remove(&scn->scn_prefetch_queue, spic);
		scan_prefetch_ctx_rele(spic->spic_spc, scn);
		kmem_free(spic, sizeof (scan_prefetch_issue_ctx_t));
	}
	ASSERT0(avl_numnodes(&scn->scn_prefetch_queue));
	mutex_exit(&spa->spa_scrub_lock);
}

static boolean_t
dsl_scan_check_resume(dsl_scan_t *scn, const dnode_phys_t *dnp,
    const zbookmark_phys_t *zb)
{
	/*
	 * We never skip over user/group accounting objects (obj<0)
	 */
	if (!ZB_IS_ZERO(&scn->scn_phys.scn_bookmark) &&
	    (int64_t)zb->zb_object >= 0) {
		/*
		 * If we already visited this bp & everything below (in
		 * a prior txg sync), don't bother doing it again.
		 */
		if (zbookmark_subtree_completed(dnp, zb,
		    &scn->scn_phys.scn_bookmark))
			return (B_TRUE);

		/*
		 * If we found the block we're trying to resume from, or
		 * we went past it to a different object, zero it out to
		 * indicate that it's OK to start checking for suspending
		 * again.
		 */
		if (bcmp(zb, &scn->scn_phys.scn_bookmark, sizeof (*zb)) == 0 ||
		    zb->zb_object > scn->scn_phys.scn_bookmark.zb_object) {
			dprintf("resuming at %llx/%llx/%llx/%llx\n",
			    (longlong_t)zb->zb_objset,
			    (longlong_t)zb->zb_object,
			    (longlong_t)zb->zb_level,
			    (longlong_t)zb->zb_blkid);
			bzero(&scn->scn_phys.scn_bookmark, sizeof (*zb));
		}
	}
	return (B_FALSE);
}

static void dsl_scan_visitbp(blkptr_t *bp, const zbookmark_phys_t *zb,
    dnode_phys_t *dnp, dsl_dataset_t *ds, dsl_scan_t *scn,
    dmu_objset_type_t ostype, dmu_tx_t *tx);
inline __attribute__((always_inline)) static void dsl_scan_visitdnode(
    dsl_scan_t *, dsl_dataset_t *ds, dmu_objset_type_t ostype,
    dnode_phys_t *dnp, uint64_t object, dmu_tx_t *tx);

/*
 * Return nonzero on i/o error.
 * Return new buf to write out in *bufp.
 */
inline __attribute__((always_inline)) static int
dsl_scan_recurse(dsl_scan_t *scn, dsl_dataset_t *ds, dmu_objset_type_t ostype,
    dnode_phys_t *dnp, const blkptr_t *bp,
    const zbookmark_phys_t *zb, dmu_tx_t *tx)
{
	dsl_pool_t *dp = scn->scn_dp;
	int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SCAN_THREAD;
	int err;

	if (BP_GET_LEVEL(bp) > 0) {
		arc_flags_t flags = ARC_FLAG_WAIT;
		int i;
		blkptr_t *cbp;
		int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
		arc_buf_t *buf;

		err = arc_read(NULL, dp->dp_spa, bp, arc_getbuf_func, &buf,
		    ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
		if (err) {
			scn->scn_phys.scn_errors++;
			return (err);
		}
		for (i = 0, cbp = buf->b_data; i < epb; i++, cbp++) {
			zbookmark_phys_t czb;

			SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
			    zb->zb_level - 1,
			    zb->zb_blkid * epb + i);
			dsl_scan_visitbp(cbp, &czb, dnp,
			    ds, scn, ostype, tx);
		}
		arc_buf_destroy(buf, &buf);
	} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
		arc_flags_t flags = ARC_FLAG_WAIT;
		dnode_phys_t *cdnp;
		int i;
		int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
		arc_buf_t *buf;

		if (BP_IS_PROTECTED(bp)) {
			ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
			zio_flags |= ZIO_FLAG_RAW;
		}

		err = arc_read(NULL, dp->dp_spa, bp, arc_getbuf_func, &buf,
		    ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
		if (err) {
			scn->scn_phys.scn_errors++;
			return (err);
		}
		for (i = 0, cdnp = buf->b_data; i < epb;
		    i += cdnp->dn_extra_slots + 1,
		    cdnp += cdnp->dn_extra_slots + 1) {
			dsl_scan_visitdnode(scn, ds, ostype,
			    cdnp, zb->zb_blkid * epb + i, tx);
		}

		arc_buf_destroy(buf, &buf);
	} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
		arc_flags_t flags = ARC_FLAG_WAIT;
		objset_phys_t *osp;
		arc_buf_t *buf;

		err = arc_read(NULL, dp->dp_spa, bp, arc_getbuf_func, &buf,
		    ZIO_PRIORITY_SCRUB, zio_flags, &flags, zb);
		if (err) {
			scn->scn_phys.scn_errors++;
			return (err);
		}

		osp = buf->b_data;

		dsl_scan_visitdnode(scn, ds, osp->os_type,
		    &osp->os_meta_dnode, DMU_META_DNODE_OBJECT, tx);

		if (OBJSET_BUF_HAS_USERUSED(buf)) {
			/*
			 * We also always visit user/group/project accounting
			 * objects, and never skip them, even if we are
			 * suspending. This is necessary so that the
			 * space deltas from this txg get integrated.
			 */
			if (OBJSET_BUF_HAS_PROJECTUSED(buf))
				dsl_scan_visitdnode(scn, ds, osp->os_type,
				    &osp->os_projectused_dnode,
				    DMU_PROJECTUSED_OBJECT, tx);
			dsl_scan_visitdnode(scn, ds, osp->os_type,
			    &osp->os_groupused_dnode,
			    DMU_GROUPUSED_OBJECT, tx);
			dsl_scan_visitdnode(scn, ds, osp->os_type,
			    &osp->os_userused_dnode,
			    DMU_USERUSED_OBJECT, tx);
		}
		arc_buf_destroy(buf, &buf);
	}

	return (0);
}

inline __attribute__((always_inline)) static void
dsl_scan_visitdnode(dsl_scan_t *scn, dsl_dataset_t *ds,
    dmu_objset_type_t ostype, dnode_phys_t *dnp,
    uint64_t object, dmu_tx_t *tx)
{
	int j;

	for (j = 0; j < dnp->dn_nblkptr; j++) {
		zbookmark_phys_t czb;

		SET_BOOKMARK(&czb, ds ? ds->ds_object : 0, object,
		    dnp->dn_nlevels - 1, j);
		dsl_scan_visitbp(&dnp->dn_blkptr[j],
		    &czb, dnp, ds, scn, ostype, tx);
	}

	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
		zbookmark_phys_t czb;
		SET_BOOKMARK(&czb, ds ? ds->ds_object : 0, object,
		    0, DMU_SPILL_BLKID);
		dsl_scan_visitbp(DN_SPILL_BLKPTR(dnp),
		    &czb, dnp, ds, scn, ostype, tx);
	}
}

/*
 * The arguments are in this order because mdb can only print the
 * first 5; we want them to be useful.
 */
static void
dsl_scan_visitbp(blkptr_t *bp, const zbookmark_phys_t *zb,
    dnode_phys_t *dnp, dsl_dataset_t *ds, dsl_scan_t *scn,
    dmu_objset_type_t ostype, dmu_tx_t *tx)
{
	dsl_pool_t *dp = scn->scn_dp;
	blkptr_t *bp_toread = NULL;

	if (dsl_scan_check_suspend(scn, zb))
		return;

	if (dsl_scan_check_resume(scn, dnp, zb))
		return;

	scn->scn_visited_this_txg++;

	/*
	 * This debugging is commented out to conserve stack space.  This
	 * function is called recursively and the debugging addes several
	 * bytes to the stack for each call.  It can be commented back in
	 * if required to debug an issue in dsl_scan_visitbp().
	 *
	 * dprintf_bp(bp,
	 *     "visiting ds=%p/%llu zb=%llx/%llx/%llx/%llx bp=%p",
	 *     ds, ds ? ds->ds_object : 0,
	 *     zb->zb_objset, zb->zb_object, zb->zb_level, zb->zb_blkid,
	 *     bp);
	 */

	if (BP_IS_HOLE(bp)) {
		scn->scn_holes_this_txg++;
		return;
	}

	if (bp->blk_birth <= scn->scn_phys.scn_cur_min_txg) {
		scn->scn_lt_min_this_txg++;
		return;
	}

	bp_toread = kmem_alloc(sizeof (blkptr_t), KM_SLEEP);
	*bp_toread = *bp;

	if (dsl_scan_recurse(scn, ds, ostype, dnp, bp_toread, zb, tx) != 0)
		goto out;

	/*
	 * If dsl_scan_ddt() has already visited this block, it will have
	 * already done any translations or scrubbing, so don't call the
	 * callback again.
	 */
	if (ddt_class_contains(dp->dp_spa,
	    scn->scn_phys.scn_ddt_class_max, bp)) {
		scn->scn_ddt_contained_this_txg++;
		goto out;
	}

	/*
	 * If this block is from the future (after cur_max_txg), then we
	 * are doing this on behalf of a deleted snapshot, and we will
	 * revisit the future block on the next pass of this dataset.
	 * Don't scan it now unless we need to because something
	 * under it was modified.
	 */
	if (BP_PHYSICAL_BIRTH(bp) > scn->scn_phys.scn_cur_max_txg) {
		scn->scn_gt_max_this_txg++;
		goto out;
	}

	scan_funcs[scn->scn_phys.scn_func](dp, bp, zb);

out:
	kmem_free(bp_toread, sizeof (blkptr_t));
}

static void
dsl_scan_visit_rootbp(dsl_scan_t *scn, dsl_dataset_t *ds, blkptr_t *bp,
    dmu_tx_t *tx)
{
	zbookmark_phys_t zb;
	scan_prefetch_ctx_t *spc;

	SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
	    ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);

	if (ZB_IS_ZERO(&scn->scn_phys.scn_bookmark)) {
		SET_BOOKMARK(&scn->scn_prefetch_bookmark,
		    zb.zb_objset, 0, 0, 0);
	} else {
		scn->scn_prefetch_bookmark = scn->scn_phys.scn_bookmark;
	}

	scn->scn_objsets_visited_this_txg++;

	spc = scan_prefetch_ctx_create(scn, NULL, FTAG);
	dsl_scan_prefetch(spc, bp, &zb);
	scan_prefetch_ctx_rele(spc, FTAG);

	dsl_scan_visitbp(bp, &zb, NULL, ds, scn, DMU_OST_NONE, tx);

	dprintf_ds(ds, "finished scan%s", "");
}

static void
ds_destroyed_scn_phys(dsl_dataset_t *ds, dsl_scan_phys_t *scn_phys)
{
	if (scn_phys->scn_bookmark.zb_objset == ds->ds_object) {
		if (ds->ds_is_snapshot) {
			/*
			 * Note:
			 *  - scn_cur_{min,max}_txg stays the same.
			 *  - Setting the flag is not really necessary if
			 *    scn_cur_max_txg == scn_max_txg, because there
			 *    is nothing after this snapshot that we care
			 *    about.  However, we set it anyway and then
			 *    ignore it when we retraverse it in
			 *    dsl_scan_visitds().
			 */
			scn_phys->scn_bookmark.zb_objset =
			    dsl_dataset_phys(ds)->ds_next_snap_obj;
			zfs_dbgmsg("destroying ds %llu; currently traversing; "
			    "reset zb_objset to %llu",
			    (u_longlong_t)ds->ds_object,
			    (u_longlong_t)dsl_dataset_phys(ds)->
			    ds_next_snap_obj);
			scn_phys->scn_flags |= DSF_VISIT_DS_AGAIN;
		} else {
			SET_BOOKMARK(&scn_phys->scn_bookmark,
			    ZB_DESTROYED_OBJSET, 0, 0, 0);
			zfs_dbgmsg("destroying ds %llu; currently traversing; "
			    "reset bookmark to -1,0,0,0",
			    (u_longlong_t)ds->ds_object);
		}
	}
}

/*
 * Invoked when a dataset is destroyed. We need to make sure that:
 *
 * 1) If it is the dataset that was currently being scanned, we write
 *	a new dsl_scan_phys_t and marking the objset reference in it
 *	as destroyed.
 * 2) Remove it from the work queue, if it was present.
 *
 * If the dataset was actually a snapshot, instead of marking the dataset
 * as destroyed, we instead substitute the next snapshot in line.
 */
void
dsl_scan_ds_destroyed(dsl_dataset_t *ds, dmu_tx_t *tx)
{
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	dsl_scan_t *scn = dp->dp_scan;
	uint64_t mintxg;

	if (!dsl_scan_is_running(scn))
		return;

	ds_destroyed_scn_phys(ds, &scn->scn_phys);
	ds_destroyed_scn_phys(ds, &scn->scn_phys_cached);

	if (scan_ds_queue_contains(scn, ds->ds_object, &mintxg)) {
		scan_ds_queue_remove(scn, ds->ds_object);
		if (ds->ds_is_snapshot)
			scan_ds_queue_insert(scn,
			    dsl_dataset_phys(ds)->ds_next_snap_obj, mintxg);
	}

	if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
	    ds->ds_object, &mintxg) == 0) {
		ASSERT3U(dsl_dataset_phys(ds)->ds_num_children, <=, 1);
		VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, ds->ds_object, tx));
		if (ds->ds_is_snapshot) {
			/*
			 * We keep the same mintxg; it could be >
			 * ds_creation_txg if the previous snapshot was
			 * deleted too.
			 */
			VERIFY(zap_add_int_key(dp->dp_meta_objset,
			    scn->scn_phys.scn_queue_obj,
			    dsl_dataset_phys(ds)->ds_next_snap_obj,
			    mintxg, tx) == 0);
			zfs_dbgmsg("destroying ds %llu; in queue; "
			    "replacing with %llu",
			    (u_longlong_t)ds->ds_object,
			    (u_longlong_t)dsl_dataset_phys(ds)->
			    ds_next_snap_obj);
		} else {
			zfs_dbgmsg("destroying ds %llu; in queue; removing",
			    (u_longlong_t)ds->ds_object);
		}
	}

	/*
	 * dsl_scan_sync() should be called after this, and should sync
	 * out our changed state, but just to be safe, do it here.
	 */
	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
}

static void
ds_snapshotted_bookmark(dsl_dataset_t *ds, zbookmark_phys_t *scn_bookmark)
{
	if (scn_bookmark->zb_objset == ds->ds_object) {
		scn_bookmark->zb_objset =
		    dsl_dataset_phys(ds)->ds_prev_snap_obj;
		zfs_dbgmsg("snapshotting ds %llu; currently traversing; "
		    "reset zb_objset to %llu",
		    (u_longlong_t)ds->ds_object,
		    (u_longlong_t)dsl_dataset_phys(ds)->ds_prev_snap_obj);
	}
}

/*
 * Called when a dataset is snapshotted. If we were currently traversing
 * this snapshot, we reset our bookmark to point at the newly created
 * snapshot. We also modify our work queue to remove the old snapshot and
 * replace with the new one.
 */
void
dsl_scan_ds_snapshotted(dsl_dataset_t *ds, dmu_tx_t *tx)
{
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	dsl_scan_t *scn = dp->dp_scan;
	uint64_t mintxg;

	if (!dsl_scan_is_running(scn))
		return;

	ASSERT(dsl_dataset_phys(ds)->ds_prev_snap_obj != 0);

	ds_snapshotted_bookmark(ds, &scn->scn_phys.scn_bookmark);
	ds_snapshotted_bookmark(ds, &scn->scn_phys_cached.scn_bookmark);

	if (scan_ds_queue_contains(scn, ds->ds_object, &mintxg)) {
		scan_ds_queue_remove(scn, ds->ds_object);
		scan_ds_queue_insert(scn,
		    dsl_dataset_phys(ds)->ds_prev_snap_obj, mintxg);
	}

	if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
	    ds->ds_object, &mintxg) == 0) {
		VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, ds->ds_object, tx));
		VERIFY(zap_add_int_key(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj,
		    dsl_dataset_phys(ds)->ds_prev_snap_obj, mintxg, tx) == 0);
		zfs_dbgmsg("snapshotting ds %llu; in queue; "
		    "replacing with %llu",
		    (u_longlong_t)ds->ds_object,
		    (u_longlong_t)dsl_dataset_phys(ds)->ds_prev_snap_obj);
	}

	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
}

static void
ds_clone_swapped_bookmark(dsl_dataset_t *ds1, dsl_dataset_t *ds2,
    zbookmark_phys_t *scn_bookmark)
{
	if (scn_bookmark->zb_objset == ds1->ds_object) {
		scn_bookmark->zb_objset = ds2->ds_object;
		zfs_dbgmsg("clone_swap ds %llu; currently traversing; "
		    "reset zb_objset to %llu",
		    (u_longlong_t)ds1->ds_object,
		    (u_longlong_t)ds2->ds_object);
	} else if (scn_bookmark->zb_objset == ds2->ds_object) {
		scn_bookmark->zb_objset = ds1->ds_object;
		zfs_dbgmsg("clone_swap ds %llu; currently traversing; "
		    "reset zb_objset to %llu",
		    (u_longlong_t)ds2->ds_object,
		    (u_longlong_t)ds1->ds_object);
	}
}

/*
 * Called when a parent dataset and its clone are swapped. If we were
 * currently traversing the dataset, we need to switch to traversing the
 * newly promoted parent.
 */
void
dsl_scan_ds_clone_swapped(dsl_dataset_t *ds1, dsl_dataset_t *ds2, dmu_tx_t *tx)
{
	dsl_pool_t *dp = ds1->ds_dir->dd_pool;
	dsl_scan_t *scn = dp->dp_scan;
	uint64_t mintxg;

	if (!dsl_scan_is_running(scn))
		return;

	ds_clone_swapped_bookmark(ds1, ds2, &scn->scn_phys.scn_bookmark);
	ds_clone_swapped_bookmark(ds1, ds2, &scn->scn_phys_cached.scn_bookmark);

	if (scan_ds_queue_contains(scn, ds1->ds_object, &mintxg)) {
		scan_ds_queue_remove(scn, ds1->ds_object);
		scan_ds_queue_insert(scn, ds2->ds_object, mintxg);
	}
	if (scan_ds_queue_contains(scn, ds2->ds_object, &mintxg)) {
		scan_ds_queue_remove(scn, ds2->ds_object);
		scan_ds_queue_insert(scn, ds1->ds_object, mintxg);
	}

	if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
	    ds1->ds_object, &mintxg) == 0) {
		int err;
		ASSERT3U(mintxg, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
		ASSERT3U(mintxg, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
		VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, ds1->ds_object, tx));
		err = zap_add_int_key(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, ds2->ds_object, mintxg, tx);
		VERIFY(err == 0 || err == EEXIST);
		if (err == EEXIST) {
			/* Both were there to begin with */
			VERIFY(0 == zap_add_int_key(dp->dp_meta_objset,
			    scn->scn_phys.scn_queue_obj,
			    ds1->ds_object, mintxg, tx));
		}
		zfs_dbgmsg("clone_swap ds %llu; in queue; "
		    "replacing with %llu",
		    (u_longlong_t)ds1->ds_object,
		    (u_longlong_t)ds2->ds_object);
	}
	if (zap_lookup_int_key(dp->dp_meta_objset, scn->scn_phys.scn_queue_obj,
	    ds2->ds_object, &mintxg) == 0) {
		ASSERT3U(mintxg, ==, dsl_dataset_phys(ds1)->ds_prev_snap_txg);
		ASSERT3U(mintxg, ==, dsl_dataset_phys(ds2)->ds_prev_snap_txg);
		VERIFY3U(0, ==, zap_remove_int(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, ds2->ds_object, tx));
		VERIFY(0 == zap_add_int_key(dp->dp_meta_objset,
		    scn->scn_phys.scn_queue_obj, ds1->ds_object, mintxg, tx));
		zfs_dbgmsg("clone_swap ds %llu; in queue; "
		    "replacing with %llu",
		    (u_longlong_t)ds2->ds_object,
		    (u_longlong_t)ds1->ds_object);
	}

	dsl_scan_sync_state(scn, tx, SYNC_CACHED);
}

/* ARGSUSED */
static int
enqueue_clones_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg)
{
	uint64_t originobj = *(uint64_t *)arg;
	dsl_dataset_t *ds;
	int err;
	dsl_scan_t *scn = dp->dp_scan;

	if (dsl_dir_phys(hds->ds_dir)->dd_origin_obj != originobj)
		return (0);

	err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
	if (err)
		return (err);

	while (dsl_dataset_phys(ds)->ds_prev_snap_obj != originobj) {
		dsl_dataset_t *prev;
		err = dsl_dataset_hold_obj(dp,
		    dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);

		dsl_dataset_rele(ds, FTAG);
		if (err)
			return (err);
		ds = prev;
	}
	scan_ds_queue_insert(scn, ds->ds_object,
	    dsl_dataset_phys(ds)->ds_prev_snap_txg);
	dsl_dataset_rele(ds, FTAG);
	return (0);
}

static void
dsl_scan_visitds(dsl_scan_t *scn, uint64_t dsobj, dmu_tx_t *tx)
{
	dsl_pool_t *dp = scn->scn_dp;
	dsl_dataset_t *ds;

	VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));

	if (scn->scn_phys.scn_cur_min_txg >=
	    scn->scn_phys.scn_max_txg) {
		/*
		 * This can happen if this snapshot was created after the
		 * scan started, and we already completed a previous snapshot
		 * that was created after the scan started.  This snapshot
		 * only references blocks with:
		 *
		 *	birth < our ds_creation_txg
		 *	cur_min_txg is no less than ds_creation_txg.
		 *	We have already visited these blocks.
		 * or
		 *	birth > scn_max_txg
		 *	The scan requested not to visit these blocks.
		 *
		 * Subsequent snapshots (and clones) can reference our
		 * blocks, or blocks with even higher birth times.
		 * Therefore we do not need to visit them either,
		 * so we do not add them to the work queue.
		 *
		 * Note that checking for cur_min_txg >= cur_max_txg
		 * is not sufficient, because in that case we may need to
		 * visit subsequent snapshots.  This happens when min_txg > 0,
		 * which raises cur_min_txg.  In this case we will visit
		 * this dataset but skip all of its blocks, because the
		 * rootbp's birth time is < cur_min_txg.  Then we will
		 * add the next snapshots/clones to the work queue.
		 */
		char *dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
		dsl_dataset_name(ds, dsname);
		zfs_dbgmsg("scanning dataset %llu (%s) is unnecessary because "
		    "cur_min_txg (%llu) >= max_txg (%llu)",
		    (longlong_t)dsobj, dsname,
		    (longlong_t)scn->scn_phys.scn_cur_min_txg,
		    (longlong_t)scn->scn_phys.scn_max_txg);
		kmem_free(dsname, MAXNAMELEN);

		goto out;
	}

	/*
	 * Only the ZIL in the head (non-snapshot) is valid. Even though
	 * snapshots can have ZIL block pointers (which may be the same
	 * BP as in the head), they must be ignored. In addition, $ORIGIN
	 * doesn't have a objset (i.e. its ds_bp is a hole) so we don't
	 * need to look for a ZIL in it either. So we traverse the ZIL here,
	 * rather than in scan_recurse(), because the regular snapshot
	 * block-sharing rules don't apply to it.
	 */
	if (!dsl_dataset_is_snapshot(ds) &&
	    (dp->dp_origin_snap == NULL ||
	    ds->ds_dir != dp->dp_origin_snap->ds_dir)) {
		objset_t *os;
		if (dmu_objset_from_ds(ds, &os) != 0) {
			goto out;
		}
		dsl_scan_zil(dp, &os->os_zil_header);
	}

	/*
	 * Iterate over the bps in this ds.
	 */
	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	dsl_scan_visit_rootbp(scn, ds, &dsl_dataset_phys(ds)->ds_bp, tx);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);

	char *dsname = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP);
	dsl_dataset_name(ds, dsname);
	zfs_dbgmsg("scanned dataset %llu (%s) with min=%llu max=%llu; "
	    "suspending=%u",
	    (longlong_t)dsobj, dsname,
	    (longlong_t)scn->scn_phys.scn_cur_min_txg,
	    (longlong_t)scn->scn_phys.scn_cur_max_txg,
	    (int)scn->scn_suspending);
	kmem_free(dsname, ZFS_MAX_DATASET_NAME_LEN);

	if (scn->scn_suspending)
		goto out;

	/*
	 * We've finished this pass over this dataset.
	 */

	/*
	 * If we did not completely visit this dataset, do another pass.
	 */
	if (scn->scn_phys.scn_flags & DSF_VISIT_DS_AGAIN) {
		zfs_dbgmsg("incomplete pass; visiting again");
		scn->scn_phys.scn_flags &= ~DSF_VISIT_DS_AGAIN;
		scan_ds_queue_insert(scn, ds->ds_object,
		    scn->scn_phys.scn_cur_max_txg);
		goto out;
	}

	/*
	 * Add descendant datasets to work queue.
	 */
	if (dsl_dataset_phys(ds)->ds_next_snap_obj != 0) {
		scan_ds_queue_insert(scn,
		    dsl_dataset_phys(ds)->ds_next_snap_obj,
		    dsl_dataset_phys(ds)->ds_creation_txg);
	}
	if (dsl_dataset_phys(ds)->ds_num_children > 1) {
		boolean_t usenext = B_FALSE;
		if (dsl_dataset_phys(ds)->ds_next_clones_obj != 0) {
			uint64_t count;
			/*
			 * A bug in a previous version of the code could
			 * cause upgrade_clones_cb() to not set
			 * ds_next_snap_obj when it should, leading to a
			 * missing entry.  Therefore we can only use the
			 * next_clones_obj when its count is correct.
			 */
			int err = zap_count(dp->dp_meta_objset,
			    dsl_dataset_phys(ds)->ds_next_clones_obj, &count);
			if (err == 0 &&
			    count == dsl_dataset_phys(ds)->ds_num_children - 1)
				usenext = B_TRUE;
		}

		if (usenext) {
			zap_cursor_t zc;
			zap_attribute_t za;
			for (zap_cursor_init(&zc, dp->dp_meta_objset,
			    dsl_dataset_phys(ds)->ds_next_clones_obj);
			    zap_cursor_retrieve(&zc, &za) == 0;
			    (void) zap_cursor_advance(&zc)) {
				scan_ds_queue_insert(scn,
				    zfs_strtonum(za.za_name, NULL),
				    dsl_dataset_phys(ds)->ds_creation_txg);
			}
			zap_cursor_fini(&zc);
		} else {
			VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
			    enqueue_clones_cb, &ds->ds_object,
			    DS_FIND_CHILDREN));
		}
	}

out:
	dsl_dataset_rele(ds, FTAG);
}

/* ARGSUSED */
static int
enqueue_cb(dsl_pool_t *dp, dsl_dataset_t *hds, void *arg)
{
	dsl_dataset_t *ds;
	int err;
	dsl_scan_t *scn = dp->dp_scan;

	err = dsl_dataset_hold_obj(dp, hds->ds_object, FTAG, &ds);
	if (err)
		return (err);

	while (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
		dsl_dataset_t *prev;
		err = dsl_dataset_hold_obj(dp,
		    dsl_dataset_phys(ds)->ds_prev_snap_obj, FTAG, &prev);
		if (err) {
			dsl_dataset_rele(ds, FTAG);
			return (err);
		}

		/*
		 * If this is a clone, we don't need to worry about it for now.
		 */
		if (dsl_dataset_phys(prev)->ds_next_snap_obj != ds->ds_object) {
			dsl_dataset_rele(ds, FTAG);
			dsl_dataset_rele(prev, FTAG);
			return (0);
		}
		dsl_dataset_rele(ds, FTAG);
		ds = prev;
	}

	scan_ds_queue_insert(scn, ds->ds_object,
	    dsl_dataset_phys(ds)->ds_prev_snap_txg);
	dsl_dataset_rele(ds, FTAG);
	return (0);
}

/* ARGSUSED */
void
dsl_scan_ddt_entry(dsl_scan_t *scn, enum zio_checksum checksum,
    ddt_entry_t *dde, dmu_tx_t *tx)
{
	const ddt_key_t *ddk = &dde->dde_key;
	ddt_phys_t *ddp = dde->dde_phys;
	blkptr_t bp;
	zbookmark_phys_t zb = { 0 };
	int p;

	if (!dsl_scan_is_running(scn))
		return;

	/*
	 * This function is special because it is the only thing
	 * that can add scan_io_t's to the vdev scan queues from
	 * outside dsl_scan_sync(). For the most part this is ok
	 * as long as it is called from within syncing context.
	 * However, dsl_scan_sync() expects that no new sio's will
	 * be added between when all the work for a scan is done
	 * and the next txg when the scan is actually marked as
	 * completed. This check ensures we do not issue new sio's
	 * during this period.
	 */
	if (scn->scn_done_txg != 0)
		return;

	for (p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
		if (ddp->ddp_phys_birth == 0 ||
		    ddp->ddp_phys_birth > scn->scn_phys.scn_max_txg)
			continue;
		ddt_bp_create(checksum, ddk, ddp, &bp);

		scn->scn_visited_this_txg++;
		scan_funcs[scn->scn_phys.scn_func](scn->scn_dp, &bp, &zb);
	}
}

/*
 * Scrub/dedup interaction.
 *
 * If there are N references to a deduped block, we don't want to scrub it
 * N times -- ideally, we should scrub it exactly once.
 *
 * We leverage the fact that the dde's replication class (enum ddt_class)
 * is ordered from highest replication class (DDT_CLASS_DITTO) to lowest
 * (DDT_CLASS_UNIQUE) so that we may walk the DDT in that order.
 *
 * To prevent excess scrubbing, the scrub begins by walking the DDT
 * to find all blocks with refcnt > 1, and scrubs each of these once.
 * Since there are two replication classes which contain blocks with
 * refcnt > 1, we scrub the highest replication class (DDT_CLASS_DITTO) first.
 * Finally the top-down scrub begins, only visiting blocks with refcnt == 1.
 *
 * There would be nothing more to say if a block's refcnt couldn't change
 * during a scrub, but of course it can so we must account for changes
 * in a block's replication class.
 *
 * Here's an example of what can occur:
 *
 * If a block has refcnt > 1 during the DDT scrub phase, but has refcnt == 1
 * when visited during the top-down scrub phase, it will be scrubbed twice.
 * This negates our scrub optimization, but is otherwise harmless.
 *
 * If a block has refcnt == 1 during the DDT scrub phase, but has refcnt > 1
 * on each visit during the top-down scrub phase, it will never be scrubbed.
 * To catch this, ddt_sync_entry() notifies the scrub code whenever a block's
 * reference class transitions to a higher level (i.e DDT_CLASS_UNIQUE to
 * DDT_CLASS_DUPLICATE); if it transitions from refcnt == 1 to refcnt > 1
 * while a scrub is in progress, it scrubs the block right then.
 */
static void
dsl_scan_ddt(dsl_scan_t *scn, dmu_tx_t *tx)
{
	ddt_bookmark_t *ddb = &scn->scn_phys.scn_ddt_bookmark;
	ddt_entry_t dde;
	int error;
	uint64_t n = 0;

	bzero(&dde, sizeof (ddt_entry_t));

	while ((error = ddt_walk(scn->scn_dp->dp_spa, ddb, &dde)) == 0) {
		ddt_t *ddt;

		if (ddb->ddb_class > scn->scn_phys.scn_ddt_class_max)
			break;
		dprintf("visiting ddb=%llu/%llu/%llu/%llx\n",
		    (longlong_t)ddb->ddb_class,
		    (longlong_t)ddb->ddb_type,
		    (longlong_t)ddb->ddb_checksum,
		    (longlong_t)ddb->ddb_cursor);

		/* There should be no pending changes to the dedup table */
		ddt = scn->scn_dp->dp_spa->spa_ddt[ddb->ddb_checksum];
		ASSERT(avl_first(&ddt->ddt_tree) == NULL);

		dsl_scan_ddt_entry(scn, ddb->ddb_checksum, &dde, tx);
		n++;

		if (dsl_scan_check_suspend(scn, NULL))
			break;
	}

	zfs_dbgmsg("scanned %llu ddt entries with class_max = %u; "
	    "suspending=%u", (longlong_t)n,
	    (int)scn->scn_phys.scn_ddt_class_max, (int)scn->scn_suspending);

	ASSERT(error == 0 || error == ENOENT);
	ASSERT(error != ENOENT ||
	    ddb->ddb_class > scn->scn_phys.scn_ddt_class_max);
}

static uint64_t
dsl_scan_ds_maxtxg(dsl_dataset_t *ds)
{
	uint64_t smt = ds->ds_dir->dd_pool->dp_scan->scn_phys.scn_max_txg;
	if (ds->ds_is_snapshot)
		return (MIN(smt, dsl_dataset_phys(ds)->ds_creation_txg));
	return (smt);
}

static void
dsl_scan_visit(dsl_scan_t *scn, dmu_tx_t *tx)
{
	scan_ds_t *sds;
	dsl_pool_t *dp = scn->scn_dp;

	if (scn->scn_phys.scn_ddt_bookmark.ddb_class <=
	    scn->scn_phys.scn_ddt_class_max) {
		scn->scn_phys.scn_cur_min_txg = scn->scn_phys.scn_min_txg;
		scn->scn_phys.scn_cur_max_txg = scn->scn_phys.scn_max_txg;
		dsl_scan_ddt(scn, tx);
		if (scn->scn_suspending)
			return;
	}

	if (scn->scn_phys.scn_bookmark.zb_objset == DMU_META_OBJSET) {
		/* First do the MOS & ORIGIN */

		scn->scn_phys.scn_cur_min_txg = scn->scn_phys.scn_min_txg;
		scn->scn_phys.scn_cur_max_txg = scn->scn_phys.scn_max_txg;
		dsl_scan_visit_rootbp(scn, NULL,
		    &dp->dp_meta_rootbp, tx);
		spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
		if (scn->scn_suspending)
			return;

		if (spa_version(dp->dp_spa) < SPA_VERSION_DSL_SCRUB) {
			VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
			    enqueue_cb, NULL, DS_FIND_CHILDREN));
		} else {
			dsl_scan_visitds(scn,
			    dp->dp_origin_snap->ds_object, tx);
		}
		ASSERT(!scn->scn_suspending);
	} else if (scn->scn_phys.scn_bookmark.zb_objset !=
	    ZB_DESTROYED_OBJSET) {
		uint64_t dsobj = scn->scn_phys.scn_bookmark.zb_objset;
		/*
		 * If we were suspended, continue from here. Note if the
		 * ds we were suspended on was deleted, the zb_objset may
		 * be -1, so we will skip this and find a new objset
		 * below.
		 */
		dsl_scan_visitds(scn, dsobj, tx);
		if (scn->scn_suspending)
			return;
	}

	/*
	 * In case we suspended right at the end of the ds, zero the
	 * bookmark so we don't think that we're still trying to resume.
	 */
	bzero(&scn->scn_phys.scn_bookmark, sizeof (zbookmark_phys_t));

	/*
	 * Keep pulling things out of the dataset avl queue. Updates to the
	 * persistent zap-object-as-queue happen only at checkpoints.
	 */
	while ((sds = avl_first(&scn->scn_queue)) != NULL) {
		dsl_dataset_t *ds;
		uint64_t dsobj = sds->sds_dsobj;
		uint64_t txg = sds->sds_txg;

		/* dequeue and free the ds from the queue */
		scan_ds_queue_remove(scn, dsobj);
		sds = NULL;

		/* set up min / max txg */
		VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
		if (txg != 0) {
			scn->scn_phys.scn_cur_min_txg =
			    MAX(scn->scn_phys.scn_min_txg, txg);
		} else {
			scn->scn_phys.scn_cur_min_txg =
			    MAX(scn->scn_phys.scn_min_txg,
			    dsl_dataset_phys(ds)->ds_prev_snap_txg);
		}
		scn->scn_phys.scn_cur_max_txg = dsl_scan_ds_maxtxg(ds);
		dsl_dataset_rele(ds, FTAG);

		dsl_scan_visitds(scn, dsobj, tx);
		if (scn->scn_suspending)
			return;
	}

	/* No more objsets to fetch, we're done */
	scn->scn_phys.scn_bookmark.zb_objset = ZB_DESTROYED_OBJSET;
	ASSERT0(scn->scn_suspending);
}

static uint64_t
dsl_scan_count_leaves(vdev_t *vd)
{
	uint64_t i, leaves = 0;

	/* we only count leaves that belong to the main pool and are readable */
	if (vd->vdev_islog || vd->vdev_isspare ||
	    vd->vdev_isl2cache || !vdev_readable(vd))
		return (0);

	if (vd->vdev_ops->vdev_op_leaf)
		return (1);

	for (i = 0; i < vd->vdev_children; i++) {
		leaves += dsl_scan_count_leaves(vd->vdev_child[i]);
	}

	return (leaves);
}

static void
scan_io_queues_update_zio_stats(dsl_scan_io_queue_t *q, const blkptr_t *bp)
{
	int i;
	uint64_t cur_size = 0;

	for (i = 0; i < BP_GET_NDVAS(bp); i++) {
		cur_size += DVA_GET_ASIZE(&bp->blk_dva[i]);
	}

	q->q_total_zio_size_this_txg += cur_size;
	q->q_zios_this_txg++;
}

static void
scan_io_queues_update_seg_stats(dsl_scan_io_queue_t *q, uint64_t start,
    uint64_t end)
{
	q->q_total_seg_size_this_txg += end - start;
	q->q_segs_this_txg++;
}

static boolean_t
scan_io_queue_check_suspend(dsl_scan_t *scn)
{
	/* See comment in dsl_scan_check_suspend() */
	uint64_t curr_time_ns = gethrtime();
	uint64_t scan_time_ns = curr_time_ns - scn->scn_sync_start_time;
	uint64_t sync_time_ns = curr_time_ns -
	    scn->scn_dp->dp_spa->spa_sync_starttime;
	int dirty_pct = scn->scn_dp->dp_dirty_total * 100 / zfs_dirty_data_max;
	int mintime = (scn->scn_phys.scn_func == POOL_SCAN_RESILVER) ?
	    zfs_resilver_min_time_ms : zfs_scrub_min_time_ms;

	return ((NSEC2MSEC(scan_time_ns) > mintime &&
	    (dirty_pct >= zfs_vdev_async_write_active_min_dirty_percent ||
	    txg_sync_waiting(scn->scn_dp) ||
	    NSEC2SEC(sync_time_ns) >= zfs_txg_timeout)) ||
	    spa_shutting_down(scn->scn_dp->dp_spa));
}

/*
 * Given a list of scan_io_t's in io_list, this issues the I/Os out to
 * disk. This consumes the io_list and frees the scan_io_t's. This is
 * called when emptying queues, either when we're up against the memory
 * limit or when we have finished scanning. Returns B_TRUE if we stopped
 * processing the list before we finished. Any sios that were not issued
 * will remain in the io_list.
 */
static boolean_t
scan_io_queue_issue(dsl_scan_io_queue_t *queue, list_t *io_list)
{
	dsl_scan_t *scn = queue->q_scn;
	scan_io_t *sio;
	int64_t bytes_issued = 0;
	boolean_t suspended = B_FALSE;

	while ((sio = list_head(io_list)) != NULL) {
		blkptr_t bp;

		if (scan_io_queue_check_suspend(scn)) {
			suspended = B_TRUE;
			break;
		}

		sio2bp(sio, &bp);
		bytes_issued += SIO_GET_ASIZE(sio);
		scan_exec_io(scn->scn_dp, &bp, sio->sio_flags,
		    &sio->sio_zb, queue);
		(void) list_remove_head(io_list);
		scan_io_queues_update_zio_stats(queue, &bp);
		sio_free(sio);
	}

	atomic_add_64(&scn->scn_bytes_pending, -bytes_issued);

	return (suspended);
}

/*
 * This function removes sios from an IO queue which reside within a given
 * range_seg_t and inserts them (in offset order) into a list. Note that
 * we only ever return a maximum of 32 sios at once. If there are more sios
 * to process within this segment that did not make it onto the list we
 * return B_TRUE and otherwise B_FALSE.
 */
static boolean_t
scan_io_queue_gather(dsl_scan_io_queue_t *queue, range_seg_t *rs, list_t *list)
{
	scan_io_t *srch_sio, *sio, *next_sio;
	avl_index_t idx;
	uint_t num_sios = 0;
	int64_t bytes_issued = 0;

	ASSERT(rs != NULL);
	ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));

	srch_sio = sio_alloc(1);
	srch_sio->sio_nr_dvas = 1;
	SIO_SET_OFFSET(srch_sio, rs->rs_start);

	/*
	 * The exact start of the extent might not contain any matching zios,
	 * so if that's the case, examine the next one in the tree.
	 */
	sio = avl_find(&queue->q_sios_by_addr, srch_sio, &idx);
	sio_free(srch_sio);

	if (sio == NULL)
		sio = avl_nearest(&queue->q_sios_by_addr, idx, AVL_AFTER);

	while (sio != NULL &&
	    SIO_GET_OFFSET(sio) < rs->rs_end && num_sios <= 32) {
		ASSERT3U(SIO_GET_OFFSET(sio), >=, rs->rs_start);
		ASSERT3U(SIO_GET_END_OFFSET(sio), <=, rs->rs_end);

		next_sio = AVL_NEXT(&queue->q_sios_by_addr, sio);
		avl_remove(&queue->q_sios_by_addr, sio);
		queue->q_sio_memused -= SIO_GET_MUSED(sio);

		bytes_issued += SIO_GET_ASIZE(sio);
		num_sios++;
		list_insert_tail(list, sio);
		sio = next_sio;
	}

	/*
	 * We limit the number of sios we process at once to 32 to avoid
	 * biting off more than we can chew. If we didn't take everything
	 * in the segment we update it to reflect the work we were able to
	 * complete. Otherwise, we remove it from the range tree entirely.
	 */
	if (sio != NULL && SIO_GET_OFFSET(sio) < rs->rs_end) {
		range_tree_adjust_fill(queue->q_exts_by_addr, rs,
		    -bytes_issued);
		range_tree_resize_segment(queue->q_exts_by_addr, rs,
		    SIO_GET_OFFSET(sio), rs->rs_end - SIO_GET_OFFSET(sio));

		return (B_TRUE);
	} else {
		range_tree_remove(queue->q_exts_by_addr, rs->rs_start,
		    rs->rs_end - rs->rs_start);
		return (B_FALSE);
	}
}

/*
 * This is called from the queue emptying thread and selects the next
 * extent from which we are to issue I/Os. The behavior of this function
 * depends on the state of the scan, the current memory consumption and
 * whether or not we are performing a scan shutdown.
 * 1) We select extents in an elevator algorithm (LBA-order) if the scan
 * 	needs to perform a checkpoint
 * 2) We select the largest available extent if we are up against the
 * 	memory limit.
 * 3) Otherwise we don't select any extents.
 */
static range_seg_t *
scan_io_queue_fetch_ext(dsl_scan_io_queue_t *queue)
{
	dsl_scan_t *scn = queue->q_scn;

	ASSERT(MUTEX_HELD(&queue->q_vd->vdev_scan_io_queue_lock));
	ASSERT(scn->scn_is_sorted);

	/* handle tunable overrides */
	if (scn->scn_checkpointing || scn->scn_clearing) {
		if (zfs_scan_issue_strategy == 1) {
			return (range_tree_first(queue->q_exts_by_addr));
		} else if (zfs_scan_issue_strategy == 2) {
			return (avl_first(&queue->q_exts_by_size));
		}
	}

	/*
	 * During normal clearing, we want to issue our largest segments
	 * first, keeping IO as sequential as possible, and leaving the
	 * smaller extents for later with the hope that they might eventually
	 * grow to larger sequential segments. However, when the scan is
	 * checkpointing, no new extents will be added to the sorting queue,
	 * so the way we are sorted now is as good as it will ever get.
	 * In this case, we instead switch to issuing extents in LBA order.
	 */
	if (scn->scn_checkpointing) {
		return (range_tree_first(queue->q_exts_by_addr));
	} else if (scn->scn_clearing) {
		return (avl_first(&queue->q_exts_by_size));
	} else {
		return (NULL);
	}
}

static void
scan_io_queues_run_one(void *arg)
{
	dsl_scan_io_queue_t *queue = arg;
	kmutex_t *q_lock = &queue->q_vd->vdev_scan_io_queue_lock;
	boolean_t suspended = B_FALSE;
	range_seg_t *rs = NULL;
	scan_io_t *sio = NULL;
	list_t sio_list;
	uint64_t bytes_per_leaf = zfs_scan_vdev_limit;
	uint64_t nr_leaves = dsl_scan_count_leaves(queue->q_vd);

	ASSERT(queue->q_scn->scn_is_sorted);

	list_create(&sio_list, sizeof (scan_io_t),
	    offsetof(scan_io_t, sio_nodes.sio_list_node));
	mutex_enter(q_lock);

	/* calculate maximum in-flight bytes for this txg (min 1MB) */
	queue->q_maxinflight_bytes =
	    MAX(nr_leaves * bytes_per_leaf, 1ULL << 20);

	/* reset per-queue scan statistics for this txg */
	queue->q_total_seg_size_this_txg = 0;
	queue->q_segs_this_txg = 0;
	queue->q_total_zio_size_this_txg = 0;
	queue->q_zios_this_txg = 0;

	/* loop until we run out of time or sios */
	while ((rs = scan_io_queue_fetch_ext(queue)) != NULL) {
		uint64_t seg_start = 0, seg_end = 0;
		boolean_t more_left = B_TRUE;

		ASSERT(list_is_empty(&sio_list));

		/* loop while we still have sios left to process in this rs */
		while (more_left) {
			scan_io_t *first_sio, *last_sio;

			/*
			 * We have selected which extent needs to be
			 * processed next. Gather up the corresponding sios.
			 */
			more_left = scan_io_queue_gather(queue, rs, &sio_list);
			ASSERT(!list_is_empty(&sio_list));
			first_sio = list_head(&sio_list);
			last_sio = list_tail(&sio_list);

			seg_end = SIO_GET_END_OFFSET(last_sio);
			if (seg_start == 0)
				seg_start = SIO_GET_OFFSET(first_sio);

			/*
			 * Issuing sios can take a long time so drop the
			 * queue lock. The sio queue won't be updated by
			 * other threads since we're in syncing context so
			 * we can be sure that our trees will remain exactly
			 * as we left them.
			 */
			mutex_exit(q_lock);
			suspended = scan_io_queue_issue(queue, &sio_list);
			mutex_enter(q_lock);

			if (suspended)
				break;
		}

		/* update statistics for debugging purposes */
		scan_io_queues_update_seg_stats(queue, seg_start, seg_end);

		if (suspended)
			break;
	}

	/*
	 * If we were suspended in the middle of processing,
	 * requeue any unfinished sios and exit.
	 */
	while ((sio = list_head(&sio_list)) != NULL) {
		list_remove(&sio_list, sio);
		scan_io_queue_insert_impl(queue, sio);
	}

	mutex_exit(q_lock);
	list_destroy(&sio_list);
}

/*
 * Performs an emptying run on all scan queues in the pool. This just
 * punches out one thread per top-level vdev, each of which processes
 * only that vdev's scan queue. We can parallelize the I/O here because
 * we know that each queue's I/Os only affect its own top-level vdev.
 *
 * This function waits for the queue runs to complete, and must be
 * called from dsl_scan_sync (or in general, syncing context).
 */
static void
scan_io_queues_run(dsl_scan_t *scn)
{
	spa_t *spa = scn->scn_dp->dp_spa;

	ASSERT(scn->scn_is_sorted);
	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	if (scn->scn_bytes_pending == 0)
		return;

	if (scn->scn_taskq == NULL) {
		int nthreads = spa->spa_root_vdev->vdev_children;

		/*
		 * We need to make this taskq *always* execute as many
		 * threads in parallel as we have top-level vdevs and no
		 * less, otherwise strange serialization of the calls to
		 * scan_io_queues_run_one can occur during spa_sync runs
		 * and that significantly impacts performance.
		 */
		scn->scn_taskq = taskq_create("dsl_scan_iss", nthreads,
		    minclsyspri, nthreads, nthreads, TASKQ_PREPOPULATE);
	}

	for (uint64_t i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
		vdev_t *vd = spa->spa_root_vdev->vdev_child[i];

		mutex_enter(&vd->vdev_scan_io_queue_lock);
		if (vd->vdev_scan_io_queue != NULL) {
			VERIFY(taskq_dispatch(scn->scn_taskq,
			    scan_io_queues_run_one, vd->vdev_scan_io_queue,
			    TQ_SLEEP) != TASKQID_INVALID);
		}
		mutex_exit(&vd->vdev_scan_io_queue_lock);
	}

	/*
	 * Wait for the queues to finish issuing their IOs for this run
	 * before we return. There may still be IOs in flight at this
	 * point.
	 */
	taskq_wait(scn->scn_taskq);
}

static boolean_t
dsl_scan_async_block_should_pause(dsl_scan_t *scn)
{
	uint64_t elapsed_nanosecs;

	if (zfs_recover)
		return (B_FALSE);

	if (scn->scn_visited_this_txg >= zfs_async_block_max_blocks)
		return (B_TRUE);

	elapsed_nanosecs = gethrtime() - scn->scn_sync_start_time;
	return (elapsed_nanosecs / NANOSEC > zfs_txg_timeout ||
	    (NSEC2MSEC(elapsed_nanosecs) > scn->scn_async_block_min_time_ms &&
	    txg_sync_waiting(scn->scn_dp)) ||
	    spa_shutting_down(scn->scn_dp->dp_spa));
}

static int
dsl_scan_free_block_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
	dsl_scan_t *scn = arg;

	if (!scn->scn_is_bptree ||
	    (BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_OBJSET)) {
		if (dsl_scan_async_block_should_pause(scn))
			return (SET_ERROR(ERESTART));
	}

	zio_nowait(zio_free_sync(scn->scn_zio_root, scn->scn_dp->dp_spa,
	    dmu_tx_get_txg(tx), bp, 0));
	dsl_dir_diduse_space(tx->tx_pool->dp_free_dir, DD_USED_HEAD,
	    -bp_get_dsize_sync(scn->scn_dp->dp_spa, bp),
	    -BP_GET_PSIZE(bp), -BP_GET_UCSIZE(bp), tx);
	scn->scn_visited_this_txg++;
	return (0);
}

static void
dsl_scan_update_stats(dsl_scan_t *scn)
{
	spa_t *spa = scn->scn_dp->dp_spa;
	uint64_t i;
	uint64_t seg_size_total = 0, zio_size_total = 0;
	uint64_t seg_count_total = 0, zio_count_total = 0;

	for (i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
		vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
		dsl_scan_io_queue_t *queue = vd->vdev_scan_io_queue;

		if (queue == NULL)
			continue;

		seg_size_total += queue->q_total_seg_size_this_txg;
		zio_size_total += queue->q_total_zio_size_this_txg;
		seg_count_total += queue->q_segs_this_txg;
		zio_count_total += queue->q_zios_this_txg;
	}

	if (seg_count_total == 0 || zio_count_total == 0) {
		scn->scn_avg_seg_size_this_txg = 0;