1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 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
|
/****************************************************************
* *
* Copyright (c) 2006-2023 Fidelity National Information *
* Services, Inc. and/or its subsidiaries. All rights reserved. *
* *
* This source code contains the intellectual property *
* of its copyright holder(s), and is made available *
* under a license. If you do not know the terms of *
* the license, please stop and do not read further. *
* *
****************************************************************/
#if defined(__MVS__) && !defined(_ISOC99_SOURCE)
#define _ISOC99_SOURCE
#endif
#include "mdef.h"
#include "gtm_socket.h"
#include "gtm_inet.h"
#include "gtm_netdb.h"
#include "gtm_time.h"
#include "gtm_fcntl.h"
#include "gtm_unistd.h"
#include "gtm_string.h"
#include "gtm_stdio.h" /* for FILE * in repl_comm.h */
#include <sys/time.h>
#include <errno.h>
#include "gdsroot.h"
#include "gdsblk.h"
#include "gtm_facility.h"
#include "fileinfo.h"
#include "gdsbt.h"
#include "gdsfhead.h"
#include "filestruct.h"
#include "gtmrecv.h"
#include "repl_comm.h"
#include "repl_msg.h"
#include "repl_dbg.h"
#include "repl_errno.h"
#include "io.h"
#include "iosp.h"
#include "eintr_wrappers.h"
#include "jnl.h"
#include "repl_sp.h"
#include "repl_filter.h"
#include "repl_log.h"
#include "gtmsource.h"
#include "sgtm_putmsg.h"
#include "gt_timer.h"
#include "min_max.h"
#include "error.h"
#include "copy.h"
#include "repl_instance.h"
#include "ftok_sems.h"
#include "buddy_list.h" /* needed for muprec.h */
#include "hashtab_mname.h" /* needed for muprec.h */
#include "hashtab_int4.h" /* needed for muprec.h */
#include "hashtab_int8.h" /* needed for muprec.h */
#include "muprec.h"
#include "gtmmsg.h"
#include "is_proc_alive.h"
#include "jnl_typedef.h"
#include "memcoherency.h"
#include "have_crit.h" /* needed for ZLIB_UNCOMPRESS */
#include "deferred_signal_handler.h" /* needed for ZLIB_UNCOMPRESS */
#include "gtm_zlib.h"
#include "wbox_test_init.h"
#ifdef GTM_TRIGGER
#include "repl_sort_tr_buff.h"
#endif
#include "replgbl.h"
#include "gtmio.h"
#include "repl_inst_dump.h" /* for "repl_dump_histinfo" prototype */
#include "gv_trigger_common.h"
#include "anticipatory_freeze.h"
#ifdef GTM_TLS
#include "gtm_repl.h"
#endif
#include "util.h" /* for OUT_BUFF_SIZE */
#define GTM_ZLIB_UNCMP_ERR_STR "error from zlib uncompress function "
#define GTM_ZLIB_Z_MEM_ERROR_STR "Out-of-memory " GTM_ZLIB_UNCMP_ERR_STR
#define GTM_ZLIB_Z_BUF_ERROR_STR "Insufficient output buffer " GTM_ZLIB_UNCMP_ERR_STR
#define GTM_ZLIB_Z_DATA_ERROR_STR "Input-data-incomplete-or-corrupt " GTM_ZLIB_UNCMP_ERR_STR
#define GTM_ZLIB_UNCMPLEN_ERROR_STR "Decompressed message data length %d is not equal to precompressed length %d "
#define GTM_ZLIB_UNCMP_ERR_SEQNO_STR "at seqno "INT8_FMT" "INT8_FMTX"\n"
#define GTM_ZLIB_UNCMP_ERR_SOLVE_STR "before sending REPL_CMP_SOLVE message\n"
#define GTM_ZLIB_UNCMPTRANSITION_STR "Defaulting to NO decompression\n"
#define RECVBUFF_REPLMSGLEN_FACTOR 8
#define GTMRECV_WAIT_FOR_STARTJNLSEQNO 100 /* ms */
#define GTMRECV_WAIT_FOR_UPD_PROGRESS 100 /* ms */
#define PROC_RECVOPS_PRINT_MSG_LEN 2048 /* bytes*/
/* By having different high and low watermarks, we can reduce the # of XOFF/XON exchanges */
#define RECVPOOL_HIGH_WATERMARK_PCTG 90 /* Send XOFF when %age of receive pool space occupied goes beyond this */
#define RECVPOOL_LOW_WATERMARK_PCTG 80 /* Send XON when %age of receive pool space occupied falls below this */
#define RECVPOOL_XON_TRIGGER_SIZE (1 * 1024 * 1024) /* Keep the low water mark within this amount of high water mark
* so that we don't wait too long to send XON */
#define GTMRECV_XOFF_LOG_CNT 100
#define GTMRECV_HEARTBEAT_PERIOD 10 /* seconds, timer that goes off every this period is the time keeper for
* receiver server; used to reduce calls to time related systemc calls */
#define ONLN_RLBK_CMD_MAXLEN 1024
#define MUPIP_DIST_STR "$gtm_dist/mupip "
#define ONLN_RLBK_CMD "journal "
#define ONLN_RLBK_VERBOSE "-verbose "
#define ONLN_RLBK_QUALIFIERS "-online -rollback -backward \"*\" -fetchresync=" /* port# will be filled later */
#if defined(__hpux) && !defined(__hppa) || defined(_AIX)
#define KEEPALIVE_PROTO_LEVEL IPPROTO_TCP
#define KEEPALIVE_TIME 5
#define KEEPALIVE_INTVL 5
#define KEEPALIVE_PROBES 5
#elif defined(__linux__)
#define KEEPALIVE_PROTO_LEVEL SOL_TCP
#define KEEPALIVE_TIME 5
#define KEEPALIVE_INTVL 5
#define KEEPALIVE_PROBES 5
#endif
GBLDEF repl_msg_ptr_t gtmrecv_msgp;
GBLDEF int gtmrecv_max_repl_msglen;
GBLDEF int gtmrecv_sock_fd = FD_INVALID;
GBLDEF boolean_t repl_connection_reset = TRUE;
GBLDEF boolean_t gtmrecv_wait_for_jnl_seqno = FALSE;
GBLDEF boolean_t gtmrecv_bad_trans_sent = FALSE;
GBLDEF boolean_t gtmrecv_send_cmp2uncmp = FALSE;
GBLDEF qw_num repl_recv_data_recvd = 0;
GBLDEF qw_num repl_recv_data_processed = 0;
GBLDEF qw_num repl_recv_postfltr_data_procd = 0;
GBLDEF qw_num repl_recv_lastlog_data_recvd = 0;
GBLDEF qw_num repl_recv_lastlog_data_procd = 0;
GBLDEF time_t repl_recv_prev_log_time;
GBLDEF time_t repl_recv_this_log_time;
GBLDEF volatile time_t gtmrecv_now = 0;
STATICDEF uchar_ptr_t gtmrecv_cmpmsgp;
STATICDEF int gtmrecv_cur_cmpmsglen;
STATICDEF int gtmrecv_max_repl_cmpmsglen;
STATICDEF uchar_ptr_t gtmrecv_uncmpmsgp;
STATICDEF int gtmrecv_max_repl_uncmpmsglen;
STATICDEF int gtmrecv_repl_cmpmsglen;
STATICDEF int gtmrecv_repl_uncmpmsglen;
STATICFNDCL void gtmrecv_repl_send_loop_error(int status, char *msgtypestr);
STATICFNDCL int repl_tr_endian_convert(unsigned char remote_jnl_ver, uchar_ptr_t jnl_buff, uint4 jnl_len);
STATICFNDCL void do_flow_control(gtm_uint64_t write_pos);
STATICFNDCL int gtmrecv_est_conn(void);
STATICFNDCL int gtmrecv_start_onln_rlbk(void);
STATICFNDCL void prepare_recvpool_for_write(gtm_uint64_t datalen, gtm_uint64_t pre_filter_write_len);
STATICFNDCL void copy_to_recvpool(uchar_ptr_t databuff, gtm_uint64_t datalen);
STATICFNDCL void wait_for_updproc_to_clear_backlog(void);
STATICFNDCL void process_tr_buff(int msg_type);
STATICFNDCL void gtmrecv_updresync_histinfo_find_seqno(seq_num input_seqno, int4 strm_num, repl_histinfo *histinfo);
STATICFNDCL void gtmrecv_updresync_histinfo_get(int4 index, repl_histinfo *histinfo);
STATICFNDCL void gtmrecv_process_need_strminfo_msg(repl_needstrminfo_msg_ptr_t need_strminfo_msg);
STATICFNDCL void gtmrecv_process_need_histinfo_msg(repl_needhistinfo_msg_ptr_t need_histinfo_msg, repl_histinfo *histinfo);
STATICFNDCL void do_main_loop(boolean_t crash_restart);
STATICFNDCL void gtmrecv_heartbeat_timer(TID tid, int4 interval_len, int *interval_ptr);
STATICFNDCL void gtmrecv_main_loop(boolean_t crash_restart);
#ifdef GTM_TLS
STATICFNDCL boolean_t gtmrecv_exchange_tls_info(uint4 remote_API_ver, uint4 remote_lib_ver);
#endif
GBLREF gtmrecv_options_t gtmrecv_options;
GBLREF int gtmrecv_listen_sock_fd;
GBLREF recvpool_addrs recvpool;
GBLREF boolean_t gtmrecv_logstats;
GBLREF int gtmrecv_filter;
GBLREF int gtmrecv_log_fd;
GBLREF FILE *gtmrecv_log_fp;
GBLREF seq_num seq_num_zero, seq_num_one, seq_num_minus_one;
GBLREF unsigned char *repl_filter_buff;
GBLREF int repl_filter_bufsiz;
GBLREF unsigned int jnl_source_datalen, jnl_dest_maxdatalen;
GBLREF unsigned char jnl_source_rectype, jnl_dest_maxrectype;
GBLREF int repl_max_send_buffsize, repl_max_recv_buffsize;
GBLREF seq_num lastlog_seqno;
GBLREF uint4 log_interval;
GBLREF qw_num trans_recvd_cnt, last_log_tr_recvd_cnt;
GBLREF jnlpool_addrs_ptr_t jnlpool;
GBLREF jnl_gbls_t jgbl;
GBLREF mur_opt_struct mur_options;
GBLREF mur_gbls_t murgbl;
GBLREF repl_conn_info_t *this_side, *remote_side;
GBLREF int4 strm_index;
error_def(ERR_INSNOTJOINED);
error_def(ERR_INSROLECHANGE);
error_def(ERR_INSUNKNOWN);
error_def(ERR_JNLNEWREC);
error_def(ERR_JNLSETDATA2LONG);
error_def(ERR_NOSUPPLSUPPL);
error_def(ERR_PRIMARYNOTROOT);
error_def(ERR_RCVRMANYSTRMS);
error_def(ERR_REPL2OLD);
error_def(ERR_REPLCOMM);
error_def(ERR_REPLINSTNOHIST);
error_def(ERR_REPLINSTREAD);
error_def(ERR_REPLNOTLS);
error_def(ERR_REPLTRANS2BIG);
error_def(ERR_REPLXENDIANFAIL);
error_def(ERR_RESUMESTRMNUM);
error_def(ERR_REUSEINSTNAME);
error_def(ERR_REPLAHEAD);
error_def(ERR_STRMNUMIS);
error_def(ERR_SUPRCVRNEEDSSUPSRC);
error_def(ERR_SYSCALL);
error_def(ERR_TEXT);
error_def(ERR_TLSCONVSOCK);
error_def(ERR_TLSHANDSHAKE);
error_def(ERR_UNIMPLOP);
error_def(ERR_UPDSYNCINSTFILE);
typedef enum
{
GTM_RECV_POOL,
GTM_RECV_CMPBUFF
} gtmrecv_buff_t;
static unsigned char *buffp, *buff_start, *msgbuff;
static gtm_uint64_t buff_unprocessed;
static gtm_uint64_t buffered_data_len;
static gtm_uint64_t max_recv_bufsiz;
static gtm_uint64_t data_len;
static gtm_uint64_t exp_data_len;
static boolean_t xoff_sent;
static repl_msg_t xon_msg, xoff_msg;
static int xoff_msg_log_cnt = 0;
static long recvpool_high_watermark, recvpool_low_watermark;
static gtm_uint64_t write_loc, write_wrap;
static gtm_uint64_t write_off;
static double time_elapsed;
static gtm_uint64_t recvpool_size;
static int heartbeat_period;
#ifdef REPL_CMP_SOLVE_TESTING
static boolean_t repl_cmp_solve_timer_set;
#endif
#define ISSUE_REPLCOMM_ERROR(REASON, SAVE_ERRNO) \
{ \
SEND_SYSMSG_REPLCOMM(LEN_AND_LIT(REASON)); \
if (0 != SAVE_ERRNO) \
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(7) ERR_REPLCOMM, 0, ERR_TEXT, 2, LEN_AND_LIT(REASON), SAVE_ERRNO);\
else \
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_REPLCOMM, 0, ERR_TEXT, 2, LEN_AND_LIT(REASON)); \
}
#define GTMRECV_EXPAND_CMPBUFF_IF_NEEDED(cmpmsglen) \
{ \
int lclcmpmsglen; \
\
lclcmpmsglen = ROUND_UP2(cmpmsglen, REPL_MSG_ALIGN); \
if (lclcmpmsglen > gtmrecv_max_repl_cmpmsglen) \
{ \
if (NULL != gtmrecv_cmpmsgp) \
free(gtmrecv_cmpmsgp); \
gtmrecv_cmpmsgp = (uchar_ptr_t)malloc(lclcmpmsglen); \
gtmrecv_max_repl_cmpmsglen = (lclcmpmsglen); \
} \
assert(0 == (gtmrecv_max_repl_cmpmsglen % REPL_MSG_ALIGN)); \
}
#define GTMRECV_EXPAND_UNCMPBUFF_IF_NEEDED(uncmpmsglen) \
{ \
if (uncmpmsglen > gtmrecv_max_repl_uncmpmsglen) \
{ \
if (NULL != gtmrecv_uncmpmsgp) \
free(gtmrecv_uncmpmsgp); \
gtmrecv_uncmpmsgp = (uchar_ptr_t)malloc(uncmpmsglen); \
gtmrecv_max_repl_uncmpmsglen = (uncmpmsglen); \
} \
assert(0 == (gtmrecv_max_repl_uncmpmsglen % REPL_MSG_ALIGN)); \
}
/* Set an alternate buffer as the target to hold the incoming compressed journal records.
* This will then be uncompressed into yet another buffer from where the records will be
* transferred to the receive pool one transaction at a time.
*/
#define GTMRECV_SET_BUFF_TARGET_CMPBUFF(cmplen, uncmplen, curcmplen) \
{ \
GTMRECV_EXPAND_CMPBUFF_IF_NEEDED(cmplen); \
GTMRECV_EXPAND_UNCMPBUFF_IF_NEEDED(uncmplen); \
curcmplen = 0; \
}
/* Wrap the "prepare_recvpool_for_write", "copy_to_recvpool", "do_flow_control" and "gtmrecv_poll_actions" functions in macros.
* This is needed because every invocation of these functions (except for a few gtmrecv_poll_actions invocations) should check
* a few global variables to determine if there was an error and in that case return from the caller.
* All callers of these functions should use the macro and not invoke the function directly.
* This avoids duplication of the error check logic.
*/
#define PREPARE_RECVPOOL_FOR_WRITE(curdatalen, prefilterdatalen) \
{ \
prepare_recvpool_for_write(curdatalen, prefilterdatalen); \
/* could update "recvpool_ctl->write" and "write_loc" */ \
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno) \
return; \
}
#define COPY_TO_RECVPOOL(ptr, datalen) \
{ \
copy_to_recvpool(ptr, datalen); /* uses and updates "write_loc" */ \
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno) \
return; \
}
#if defined(GTM_REPLIC_FLOW_CONTROL_ENABLED)
#define DO_FLOW_CONTROL(write_loc) \
{ \
do_flow_control(write_loc); \
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno) \
return; \
}
#else
#define DO_FLOW_CONTROL(write_loc) /* nothing */
#endif
/* Wrapper for gtmrecv_poll_actions to handle connection reset. The arguments passed in are typically either the static variables
* `data_len', `buff_unprocessed' and `buffp'. But, these values are relevant only when this is being called from do_main_loop as
* they indicate how much of the received buffer is still to be processed and the pointer to the beginning of the unprocessed
* buffer. Outside do_main_loop (for instance, in gtmrecv_est_conn), gtmrecv_poll_actions is invoked with 0,0, NULL respectively.
*/
#define GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp) \
{ \
gtmrecv_poll_actions(data_len, buff_unprocessed, buffp); \
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno) \
return; \
}
#define WACKY_MESSAGE(MSG, SOCK_FD, TYPE) \
MBSTART { /* If the header is wacky, assume it's a rogue transmission and reset the connection */ \
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received UNKNOWN message (type = %d / %d). " \
"Discarding it and resetting connection.\n", MSG & REPL_TR_CMP_MSG_TYPE_MASK, TYPE); \
repl_connection_reset = TRUE; \
repl_close(SOCK_FD); \
return; \
} MBEND
/* The below macro is used (within this module) to check for errors (and issue appropriate message) after REPL_SEND_LOOP
* returns.
* NOTE: This macro, in its current form, cannot be used in all REPL_SEND_LOOP usages due to specific post-error processing
* done in a few places. There is scope to:
* (a) Improve the macro to adjust to post-error processing
* (b) Provide similar macro to be used for REPL_RECV_LOOP usages
*/
#define CHECK_REPL_SEND_LOOP_ERROR(status, msgstr) \
{ \
if (SS_NORMAL != status) \
{ \
gtmrecv_repl_send_loop_error(status, msgstr); \
if (repl_connection_reset) \
return; \
} \
}
#define GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED \
{ \
sgmnt_addrs *repl_csa; \
\
repl_csa = &FILE_INFO(jnlpool->jnlpool_dummy_reg)->s_addrs; \
assert(repl_csa->now_crit); \
assert(NULL != jnlpool->jnlpool_ctl); \
if (repl_csa->onln_rlbk_cycle != jnlpool->jnlpool_ctl->onln_rlbk_cycle) \
{ \
SYNC_ONLN_RLBK_CYCLES; \
rel_lock(jnlpool->jnlpool_dummy_reg); \
gtmrecv_onln_rlbk_clnup(); \
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno) \
return; \
} \
}
/* For cross-endian conversion to happen on the receiving side, the receiver must understand the layout of the journal
* records. To keep the endian conversion logic on both primary and secondary simple, the following scheme is used:
* (a) If primary < secondary, endian conversion will happen on primary.
* (b) If primary >= secondary, primary will apply internal filters to convert the records to secondary's format. The
* secondary on receiving them will do the necessary endian conversion before letting the update process see them.
*
* However, the above logic will cause the older versions (< V5.4-002) to NOT replicate to V5.4-002 as the endian-conversion
* by-primary is introduced only from V5.4-002 and above. Hence, allow secondary to do endian conversion for this special
* case when the primary is a GT.M version running V5.3-003 (V18_JNL_VER) to V5.4-001 (V20_JNL_VER). The lower limit is
* chosen to be V5.3-003 since that was the first version where cross-endian conversion was supported.
*
* There is one other exception. V5.5 source server (V22_JNL_VER) had a bug wherein a history record is endian-converted
* when the replication is NOT cross-endian and vice versa. In either case, do an endian conversion of the history record.
*
* The below macro takes all the above conditions into consideration to determine if the receiver server needs to do endian
* converison or not.
*/
#define ENDIAN_CONVERSION_NEEDED(IS_NEW_HISTREC, THIS_JNL_VER, REMOTE_JNL_VER, X_ENDIAN) \
((IS_NEW_HISTREC && (V22_JNL_VER == REMOTE_JNL_VER)) \
|| (X_ENDIAN && ((REMOTE_JNL_VER >= THIS_JNL_VER) || (V21_JNL_VER > REMOTE_JNL_VER))))
STATICFNDEF void gtmrecv_repl_send_loop_error(int status, char *msgtypestr)
{
char print_msg[1024];
assert((EREPL_SEND == repl_errno) || (EREPL_SELECT == repl_errno));
if (REPL_CONN_RESET(status) && EREPL_SEND == repl_errno)
{
SNPRINTF(print_msg, SIZEOF(print_msg), "Connection got reset while sending %s message. Status = %d ; %s\n",
msgtypestr, status, STRERROR(status));
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) MAKE_MSG_INFO(ERR_REPLCOMM), 0, ERR_TEXT, 2,
LEN_AND_STR(print_msg));
repl_connection_reset = TRUE;
repl_close(>mrecv_sock_fd);
SNPRINTF(print_msg, SIZEOF(print_msg), "Closing connection on receiver side\n");
repl_log(gtmrecv_log_fp, TRUE, TRUE, print_msg);
return;
} else if (EREPL_SEND == repl_errno)
{
SNPRINTF(print_msg, SIZEOF(print_msg), "Error sending %s message. Error in send : %s",
msgtypestr, STRERROR(status));
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_REPLCOMM, 0, ERR_TEXT, 2, LEN_AND_STR(print_msg));
} else if (EREPL_SELECT == repl_errno)
{
SNPRINTF(print_msg, SIZEOF(print_msg), "Error sending %s message. Error in select : %s",
msgtypestr, STRERROR(status));
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_REPLCOMM, 0, ERR_TEXT, 2, LEN_AND_STR(print_msg));
}
}
/* convert endianness of transaction */
STATICFNDEF int repl_tr_endian_convert(unsigned char remote_jnl_ver, uchar_ptr_t jnl_buff, uint4 jnl_len)
{
unsigned char *jb, *jstart, *ptr;
enum jnl_record_type rectype;
int status, reclen;
uint4 jlen;
jrec_prefix *prefix;
jnl_record *rec;
jnl_string *keystr;
mstr_len_t *vallen_ptr;
mstr_len_t temp_val;
repl_old_triple_jnl_ptr_t oldtriple;
repl_histinfo *histinfo;
jnl_str_len_t nodeflags_keylen;
uint4 update_num, num_participants_4bytes;
unsigned short num_participants_2bytes;
# ifdef DEBUG
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
# endif
assert((remote_jnl_ver >= this_side->jnl_ver) || (V21_JNL_VER > remote_jnl_ver) || (V22_JNL_VER == remote_jnl_ver));
jb = jnl_buff;
status = SS_NORMAL;
jlen = jnl_len;
assert(0 == ((UINTPTR_T)jb % SIZEOF(UINTPTR_T)));
while (JREC_PREFIX_SIZE <= jlen)
{
assert(0 == ((UINTPTR_T)jb % SIZEOF(UINTPTR_T)));
rec = (jnl_record *) jb;
rectype = (enum jnl_record_type)rec->prefix.jrec_type;
reclen = rec->prefix.forwptr = GTM_BYTESWAP_24(rec->prefix.forwptr);
if (!IS_REPLICATED(rectype) || (0 == reclen) || (reclen > jlen))
{ /* Bad OR Incomplete record */
assert(FALSE);
status = -1;
break;
}
assert(!IS_ZTP(rectype));
assert((JRT_HISTREC == rectype) || (JRT_TRIPLE == rectype) || IS_SET_KILL_ZKILL_ZTWORM_LGTRIG_ZTRIG(rectype)
|| (JRT_TCOM == rectype) || (JRT_NULL == rectype));
DEBUG_ONLY(jstart = jb;)
if (JRT_HISTREC == rectype)
{
histinfo = &(((repl_histrec_jnl_ptr_t)rec)->histcontent);
ENDIAN_CONVERT_REPL_HISTINFO(histinfo);
} else if (JRT_TRIPLE == rectype)
{
oldtriple = (repl_old_triple_jnl_ptr_t)rec;
oldtriple->cycle = GTM_BYTESWAP_32(oldtriple->cycle);
oldtriple->start_seqno = GTM_BYTESWAP_64(oldtriple->start_seqno);
} else
{ /* pini_addr, time, checksum and tn field of the journal records created by the source server are
* irrelevant to the receiver server and hence no point doing the endian conversion for them.
*/
((jrec_suffix *)((unsigned char *)rec + reclen - JREC_SUFFIX_SIZE))->backptr = reclen;
rec->jrec_null.jnl_seqno = GTM_BYTESWAP_64(rec->jrec_null.jnl_seqno);
/* Starting jnl ver V22, we have a "strm_seqno" field in the journal record so endian convert that */
if (V22_JNL_VER <= remote_jnl_ver)
{ /* At this point, we could have a TCOM or NULL or SET/KILL/ZKILL/ZTRIG type of record.
* Assert that all of them have "strm_seqno" at the exact same offset so we can avoid
* an if/then/else check on the record types in order to endian convert "strm_seqno".
*/
assert(&rec->jrec_null.strm_seqno == &rec->jrec_set_kill.strm_seqno);
assert(&rec->jrec_null.strm_seqno == &rec->jrec_tcom.strm_seqno);
rec->jrec_null.strm_seqno = GTM_BYTESWAP_64(rec->jrec_null.strm_seqno);
}
if (IS_SET_KILL_ZKILL_ZTWORM_LGTRIG_ZTRIG(rectype))
{ /* This code will need changes in case the jnl-ver changes from V28 to V29 so add an assert to
* alert to that possibility. Once the code is fixed for the new jnl format, change the assert
* to reflect the new latest jnl-ver.
*/
assert(JNL_VER_THIS == V28_JNL_VER);
/* To better understand the logic below (particularly the use of hardcoded offsets), see comment
* in repl_filter.c (search for "struct_jrec_upd layout" for the various jnl versions we support).
*/
if (V22_JNL_VER <= remote_jnl_ver)
{ /* byte-swap update_num */
assert(&rec->jrec_set_kill.update_num == &rec->jrec_ztworm.update_num);
assert(&rec->jrec_set_kill.update_num == &rec->jrec_lgtrig.update_num);
rec->jrec_set_kill.update_num = GTM_BYTESWAP_32(rec->jrec_set_kill.update_num);
/* No need to byte-swap num_participants as it is not used by the update process */
/* Get pointer to mumps_node */
keystr = (jnl_string *)&rec->jrec_set_kill.mumps_node;
assert(keystr == (jnl_string *)&rec->jrec_ztworm.ztworm_str);
assert(keystr == (jnl_string *)&rec->jrec_lgtrig.lgtrig_str);
} else if (V19_JNL_VER <= remote_jnl_ver)
{ /* byte-swap update_num */
ptr = (unsigned char *)rec + 32; /* is offset of update_num in V19 struct_jrec_upd */
update_num = *(uint4 *)ptr;
*(uint4 *)ptr = GTM_BYTESWAP_32(update_num);
/* No need to byte-swap num_participants as it is not used by the update process */
/* Get pointer to mumps_node */
keystr = (jnl_string *)((unsigned char *)rec + 40); /* is offset of mumps_node */
} else
{
assert(V17_JNL_VER <= remote_jnl_ver);
/* Note: In V17, there is no update_num or num_participants like V19 so no endian convert */
/* Get pointer to mumps_node */
keystr = (jnl_string *)((unsigned char *)rec + 32); /* is offset of mumps_node */
}
/* In V18, the jnl_string contained a 32 bit length field followed by mumps_node
* In V19, the "length" field is divided into 8 bit "nodeflags" and 24 bit "length" fields.
* Byteswap the entire 32 bit value
*/
nodeflags_keylen = *(jnl_str_len_t *)keystr;
*(jnl_str_len_t *)keystr = GTM_BYTESWAP_32(nodeflags_keylen);
if (IS_SET(rectype))
{
assert(!IS_ZTWORM(rectype));
assert(!IS_LGTRIG(rectype));
/* SET records have a 'value' part which needs to be endian converted */
vallen_ptr = (mstr_len_t *)&keystr->text[keystr->length];
GET_MSTR_LEN(temp_val, vallen_ptr);
temp_val = GTM_BYTESWAP_32(temp_val);
PUT_MSTR_LEN(vallen_ptr, temp_val);
/* The actual 'value' itself is a character array and hence needs no endian conversion */
}
} else if (JRT_TCOM == rectype)
{ /* byte-swap num_participants as this is relied upon by "repl_sort_tr_buff".
* The offset and size of this field are different for older versions. Endian convert accordingly.
* V22 struct_jrec_tcom layout is as follows.
* offset = 0042 [0x002a] size = 0002 [0x0002] ----> num_participants
* V19/V20/V21 struct_jrec_tcom layout is as follows.
* offset = 0034 [0x0022] size = 0002 [0x0002] ----> num_participants
* V17/V18 struct_jrec_tcom layout is as follows.
* offset = 0040 [0x0028] size = 0004 [0x0004] ----> participants
*/
if (V22_JNL_VER <= remote_jnl_ver)
{
assert(42 == ((INTPTR_T)&rec->jrec_tcom.num_participants - (INTPTR_T)rec));
assert(SIZEOF(num_participants_2bytes) == SIZEOF(rec->jrec_tcom.num_participants));
num_participants_2bytes = rec->jrec_tcom.num_participants;
rec->jrec_tcom.num_participants = GTM_BYTESWAP_16(num_participants_2bytes);
} else if (V19_JNL_VER <= remote_jnl_ver)
{
ptr = (unsigned char *)rec + 34; /* is offset of update_num in V19 struct_jrec_upd */
assert(SIZEOF(num_participants_2bytes) == SIZEOF(unsigned short));
num_participants_2bytes = *(unsigned short *)ptr;
rec->jrec_tcom.num_participants = GTM_BYTESWAP_16(num_participants_2bytes);
} else
{
assert(V17_JNL_VER <= remote_jnl_ver);
ptr = (unsigned char *)rec + 40; /* is offset of update_num in V19 struct_jrec_upd */
assert(SIZEOF(num_participants_4bytes) == SIZEOF(uint4));
num_participants_4bytes = *(uint4 *)ptr;
rec->jrec_tcom.num_participants = GTM_BYTESWAP_32(num_participants_4bytes);
}
assert(rec->jrec_tcom.num_participants);
/* token_seq.jnl_seqno & strm_seqno have already been endian converted. */
}
/* else if (JRT_NULL == rectype)
* token_seq.jnl_seqno & strm_seqno have already been endian converted.
*/
assert(reclen == REC_LEN_FROM_SUFFIX(jb, reclen));
}
jb = jb + reclen;
assert(jb == jstart + reclen);
jlen -= reclen;
}
if ((-1 != status) && (0 != jlen))
{ /* Incomplete record */
assert(FALSE);
status = -1;
}
return status;
}
STATICFNDEF void do_flow_control(gtm_uint64_t write_pos)
{
/* Check for overflow before writing */
recvpool_ctl_ptr_t recvpool_ctl;
upd_proc_local_ptr_t upd_proc_local;
gtmrecv_local_ptr_t gtmrecv_local;
long space_used;
unsigned char *msg_ptr; /* needed for REPL_{SEND,RECV}_LOOP */
int tosend_len, sent_len, sent_this_iter; /* needed for REPL_SEND_LOOP */
int torecv_len, recvd_len, recvd_this_iter; /* needed for REPL_RECV_LOOP */
int status, poll_dir; /* needed for REPL_{SEND,RECV}_LOOP */
gtm_uint64_t read_pos;
seq_num temp_seq_num;
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
recvpool_ctl = recvpool.recvpool_ctl;
upd_proc_local = recvpool.upd_proc_local;
gtmrecv_local = recvpool.gtmrecv_local;
space_used = 0;
if (recvpool_ctl->wrapped)
space_used = write_pos + recvpool_size - (read_pos = upd_proc_local->read);
if (!recvpool_ctl->wrapped || space_used > recvpool_size)
space_used = write_pos - (read_pos = upd_proc_local->read);
assert(remote_side->endianness_known); /* only then is remote_side->cross_endian reliable */
if (space_used >= recvpool_high_watermark && !xoff_sent)
{ /* Send XOFF message */
if (!remote_side->cross_endian)
{
xoff_msg.type = REPL_XOFF;
memcpy((uchar_ptr_t)&xoff_msg.msg[0], (uchar_ptr_t)&upd_proc_local->read_jnl_seqno, SIZEOF(seq_num));
xoff_msg.len = MIN_REPL_MSGLEN;
} else
{
xoff_msg.type = GTM_BYTESWAP_32(REPL_XOFF);
temp_seq_num = GTM_BYTESWAP_64(upd_proc_local->read_jnl_seqno);
memcpy((uchar_ptr_t)&xoff_msg.msg[0], (uchar_ptr_t)&temp_seq_num, SIZEOF(seq_num));
xoff_msg.len = GTM_BYTESWAP_32(MIN_REPL_MSGLEN);
}
REPL_SEND_LOOP(gtmrecv_sock_fd, &xoff_msg, MIN_REPL_MSGLEN, REPL_POLL_NOWAIT)
{
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
CHECK_REPL_SEND_LOOP_ERROR(status, "REPL_XOFF");
if (gtmrecv_logstats)
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Space used = %ld, High water mark = %ld Low water mark = %ld, "
"Updproc Read = %ld, Recv Write = %ld, Sent XOFF\n", space_used, recvpool_high_watermark,
recvpool_low_watermark, read_pos, write_pos);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "REPL_XOFF sent as receive pool has %ld bytes transaction data yet to be "
"processed\n", space_used);
xoff_sent = TRUE;
xoff_msg_log_cnt = 1;
} else if (space_used < recvpool_low_watermark && xoff_sent)
{
if (!remote_side->cross_endian)
{
xon_msg.type = REPL_XON;
memcpy((uchar_ptr_t)&xon_msg.msg[0], (uchar_ptr_t)&upd_proc_local->read_jnl_seqno, SIZEOF(seq_num));
xon_msg.len = MIN_REPL_MSGLEN;
} else
{
xon_msg.type = GTM_BYTESWAP_32(REPL_XON);
temp_seq_num = GTM_BYTESWAP_64(upd_proc_local->read_jnl_seqno);
memcpy((uchar_ptr_t)&xon_msg.msg[0], (uchar_ptr_t)&temp_seq_num, SIZEOF(seq_num));
xon_msg.len = GTM_BYTESWAP_32(MIN_REPL_MSGLEN);
}
REPL_SEND_LOOP(gtmrecv_sock_fd, &xon_msg, MIN_REPL_MSGLEN, REPL_POLL_NOWAIT)
{
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
CHECK_REPL_SEND_LOOP_ERROR(status, "REPL_XON");
if (gtmrecv_logstats)
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Space used now = %ld, High water mark = %ld, "
"Low water mark = %ld, Updproc Read = %ld, Recv Write = %ld, Sent XON\n", space_used,
recvpool_high_watermark, recvpool_low_watermark, read_pos, write_pos);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "REPL_XON sent as receive pool has %ld bytes free space to buffer transaction "
"data\n", recvpool_size - space_used);
xoff_sent = FALSE;
xoff_msg_log_cnt = 0;
}
return;
}
STATICFNDEF int gtmrecv_est_conn(void)
{
recvpool_ctl_ptr_t recvpool_ctl;
upd_proc_local_ptr_t upd_proc_local;
gtmrecv_local_ptr_t gtmrecv_local;
boolean_t keepalive;
GTM_SOCKLEN_TYPE optlen;
int status, keepalive_opt, optval, save_errno;
struct linger disable_linger = {0, 0};
char print_msg[1024];
struct addrinfo primary_ai;
struct sockaddr_storage primary_sas;
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
/* Wait for a connection from a Source Server. The Receiver Server is an iterative server. */
recvpool_ctl = recvpool.recvpool_ctl;
upd_proc_local = recvpool.upd_proc_local;
gtmrecv_local = recvpool.gtmrecv_local;
/* Create a listen socket */
gtmrecv_comm_init((in_port_t)gtmrecv_local->listen_port);
primary_ai.ai_addr = (sockaddr_ptr)&primary_sas;
primary_ai.ai_addrlen = SIZEOF(primary_sas);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Waiting for a connection...\n");
/* Null initialize fields that need to be initialized only after connecting to the primary.
* It is ok not to hold a lock on the journal pool while updating jnlpool_ctl fields since this will be the only
* process updating those fields.
*/
assert(remote_side == >mrecv_local->remote_side);
remote_side->proto_ver = REPL_PROTO_VER_UNINITIALIZED;
jnlpool->jnlpool_ctl->primary_instname[0] = '\0';
while (TRUE)
{
while (TRUE)
{
if (0 < (status = fd_ioready(gtmrecv_listen_sock_fd, REPL_POLLIN, REPL_POLL_WAIT)))
break;
if (-1 == status)
{
save_errno = errno;
assert((EAGAIN != save_errno) && (EINTR != save_errno));
ISSUE_REPLCOMM_ERROR("Error in select on listen socket", save_errno);
} else if (0 == status) /* timeout */
gtmrecv_poll_actions(0, 0, NULL);
}
ACCEPT_SOCKET(gtmrecv_listen_sock_fd, primary_ai.ai_addr,
(GTM_SOCKLEN_TYPE *)&primary_ai.ai_addrlen, gtmrecv_sock_fd);
if (FD_INVALID == gtmrecv_sock_fd)
{
save_errno = errno;
# ifdef __hpux
if (ENOBUFS == save_errno)
{
gtmrecv_poll_actions(0, 0, NULL);
continue;
}
# endif
ISSUE_REPLCOMM_ERROR("Error accepting connection from Source Server", save_errno);
}
break;
}
/* Connection established */
repl_close(>mrecv_listen_sock_fd); /* Close the listener socket */
repl_connection_reset = FALSE;
if (-1 == setsockopt(gtmrecv_sock_fd, SOL_SOCKET, SO_LINGER, (const void *)&disable_linger, SIZEOF(disable_linger)))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket disable linger", errno);
# ifdef REPL_DISABLE_KEEPALIVE
keepalive = FALSE;
# else
keepalive = TRUE;
# endif
if (-1 == setsockopt(gtmrecv_sock_fd, SOL_SOCKET, SO_KEEPALIVE, (const void *)&keepalive,
SIZEOF(keepalive)))
{
ISSUE_REPLCOMM_ERROR("Error with receiver server socket enable keepalive", errno);
}
/* Set up the keepalive parameters
* TCP_KEEPCNT : overrides tcp_keepalive_probes
* TCP_KEEPIDLE: overrides tcp_keepalive_time
* TCP_KEEPINTVL: overrides tcp_keepalive_intvl
*/
# if defined(KEEPALIVE_PROTO_LEVEL)
keepalive_opt = KEEPALIVE_PROBES;
if (-1 == setsockopt(gtmrecv_sock_fd, KEEPALIVE_PROTO_LEVEL, TCP_KEEPCNT, (void*)&keepalive_opt, SIZEOF(keepalive_opt)))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket setting tcp_keepalive_probes", errno);
keepalive_opt = KEEPALIVE_TIME;
if (-1 == setsockopt(gtmrecv_sock_fd, KEEPALIVE_PROTO_LEVEL, TCP_KEEPIDLE, (void*)&keepalive_opt, SIZEOF(keepalive_opt)))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket setting tcp_keepalive_time", errno);
keepalive_opt = KEEPALIVE_INTVL;
if (-1 == setsockopt(gtmrecv_sock_fd, KEEPALIVE_PROTO_LEVEL, TCP_KEEPINTVL, (void*)&keepalive_opt, SIZEOF(keepalive_opt)))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket setting tcp_keepalive_intvl", errno);
# endif
optlen = SIZEOF(optval);
if ( -1 == getsockopt(gtmrecv_sock_fd, SOL_SOCKET, SO_KEEPALIVE, &optval, &optlen))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket checking keepalive enabled or not", errno)
# if !defined(KEEPALIVE_PROTO_LEVEL)
repl_log(gtmrecv_log_fp, TRUE, TRUE, "SO_KEEPALIVE is %s\n", (optval ? "ON" : "OFF"));
# else
repl_log(gtmrecv_log_fp, TRUE, TRUE, "SO_KEEPALIVE is %s. ", (optval ? "ON" : "OFF"));
if (-1 == getsockopt(gtmrecv_sock_fd, KEEPALIVE_PROTO_LEVEL, TCP_KEEPCNT, &optval, &optlen))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket getting tcp_keepalive_probes", errno);
if (optval)
repl_log(gtmrecv_log_fp, FALSE, TRUE, "TCP_KEEPCNT is %d, ", optval);
if (-1 == getsockopt(gtmrecv_sock_fd, KEEPALIVE_PROTO_LEVEL, TCP_KEEPIDLE, &optval, &optlen))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket getting tcp_keepalive_time", errno);
if (optval)
repl_log(gtmrecv_log_fp, FALSE, TRUE, "TCP_KEEPIDLE is %d, ", optval);
if (-1 == getsockopt(gtmrecv_sock_fd, KEEPALIVE_PROTO_LEVEL, TCP_KEEPINTVL, &optval, &optlen))
ISSUE_REPLCOMM_ERROR("Error with receiver server socket getting tcp_keepalive_intvl", errno);
if (optval)
repl_log(gtmrecv_log_fp, FALSE, TRUE, "TCP_KEEPINTVL is %d.\n", optval);
# endif
if (0 != (status = get_send_sock_buff_size(gtmrecv_sock_fd, &repl_max_send_buffsize)))
ISSUE_REPLCOMM_ERROR("Error getting socket send buffsize", errno);
if (0 != (status = get_recv_sock_buff_size(gtmrecv_sock_fd, &repl_max_recv_buffsize)))
ISSUE_REPLCOMM_ERROR("Error getting socket recv buffsize", errno);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Connection established, using TCP send buffer size %d receive buffer size %d\n",
repl_max_send_buffsize, repl_max_recv_buffsize);
repl_log_conn_info(gtmrecv_sock_fd, gtmrecv_log_fp, FALSE);
/* re-determine endianness of other side */
remote_side->endianness_known = FALSE;
/* re-determine journal format of other side */
remote_side->jnl_ver = 0;
/* re-determine compression level on the replication pipe after every connection establishment */
repl_zlib_cmp_level = ZLIB_CMPLVL_NONE;
/* Reset prior connection related state variables (see <C9J02_003091_receiver_server_assert_due_to_lingering_XOFF>) */
xoff_sent = FALSE;
xoff_msg_log_cnt = 0;
# ifdef GTM_TLS
assert(!repl_tls.enabled);
/* We either did not create a TLS/SSL aware socket or the SSL object off of the TLS/SSL aware socket should be NULL since
* we haven't yet connected to the source server. Assert that.
*/
assert((NULL == repl_tls.sock) || (NULL == repl_tls.sock->ssl));
# endif
/* Note that even though we are reopening a fresh connection, we should NOT reset the cached information
* last_rcvd_histinfo, last_valid_histinfo etc. in this case as we might just resume processing from where
* the previous connection left off in which case all the cached information is still valid. If we don't
* resume from where we left off, the receiver will anyways error out asking for a rollback to be done so
* the cached information never gets used if it is invalid.
*/
return (SS_NORMAL);
}
int gtmrecv_alloc_filter_buff(int bufsiz)
{
unsigned char *old_filter_buff;
bufsiz = ROUND_UP2(bufsiz, OS_PAGE_SIZE);
if ((NO_FILTER != gtmrecv_filter) && (repl_filter_bufsiz < bufsiz))
{
REPL_DPRINT3("Expanding filter buff from %d to %d\n", repl_filter_bufsiz, bufsiz);
old_filter_buff = repl_filter_buff;
repl_filter_buff = malloc(bufsiz);
if (NULL != old_filter_buff)
{
memcpy(repl_filter_buff, old_filter_buff, repl_filter_bufsiz);
free(repl_filter_buff);
}
repl_filter_bufsiz = bufsiz;
}
return (SS_NORMAL);
}
void gtmrecv_free_filter_buff(void)
{
if (NULL != repl_filter_buff)
{
free(repl_filter_buff);
repl_filter_buff = NULL;
repl_filter_bufsiz = 0;
}
}
int gtmrecv_alloc_msgbuff(void)
{
gtmrecv_max_repl_msglen = MAX_REPL_MSGLEN;
assert(NULL == gtmrecv_msgp); /* first time initialization. The receiver server doesn't need to re-allocate */
msgbuff = (unsigned char *)malloc(gtmrecv_max_repl_msglen + OS_PAGE_SIZE);
gtmrecv_msgp = (repl_msg_ptr_t)ROUND_UP2((unsigned long)msgbuff, OS_PAGE_SIZE);
gtmrecv_alloc_filter_buff(gtmrecv_max_repl_msglen);
return (SS_NORMAL);
}
void gtmrecv_free_msgbuff(void)
{
if (NULL != msgbuff)
{
assert(NULL != gtmrecv_msgp);
free(msgbuff);
msgbuff = NULL;
gtmrecv_msgp = NULL;
}
}
/* This function can be used to only send fixed-size message types across the replication pipe.
* This in turn uses REPL_SEND* macros but also does error checks and sets the global variables
* "repl_connection_reset" or "gtmrecv_wait_for_jnl_seqno" accordingly.
* It also does the endian conversion of the 'type' and 'len' fields of the repl_msg_t structure being sent.
*
* msg = Pointer to the message buffer to send
* type = One of the various message types listed in repl_msg.h
* len = Length of the message to be sent
* msgtypestr = Message name as a string to display meaningful error messages
* optional_seqno = Optional seqno that needs to be printed along with the message name
*/
void gtmrecv_repl_send(repl_msg_ptr_t msgp, int4 type, int4 len, char *msgtypestr, seq_num optional_seqno)
{
unsigned char *msg_ptr; /* needed for REPL_SEND_LOOP */
int tosend_len, sent_len, sent_this_iter; /* needed for REPL_SEND_LOOP */
int status, poll_dir; /* needed for REPL_{SEND,RECV}_LOOP */
FILE *log_fp;
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
assert(!jgbl.mur_rollback || !jgbl.mur_options_forward); /* ROLLBACK -FORWARD should not call this function */
assert(!mur_options.rollback || (NULL == recvpool.gtmrecv_local));
assert(mur_options.rollback || (NULL != recvpool.gtmrecv_local));
assert((REPL_MULTISITE_MSG_START > type) || (REPL_PROTO_VER_MULTISITE <= remote_side->proto_ver));
log_fp = (NULL == gtmrecv_log_fp) ? stdout : gtmrecv_log_fp;
if (MAX_SEQNO != optional_seqno)
{
repl_log(log_fp, TRUE, TRUE, "Sending %s message with seqno "INT8_FMT" "INT8_FMTX"\n", msgtypestr,
optional_seqno, optional_seqno);
} else
repl_log(log_fp, TRUE, TRUE, "Sending %s message\n", msgtypestr);
/* Assert that if we don't know the endianness of the remote side, we assume it is the same endianness */
assert(remote_side->endianness_known || !remote_side->cross_endian);
if (!remote_side->cross_endian)
{
msgp->type = type;
msgp->len = len;
} else
{
msgp->type = GTM_BYTESWAP_32(type);
msgp->len = GTM_BYTESWAP_32(len);
}
REPL_SEND_LOOP(gtmrecv_sock_fd, msgp, len, REPL_POLL_NOWAIT)
{
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
CHECK_REPL_SEND_LOOP_ERROR(status, msgtypestr);
assert(SS_NORMAL == status);
}
/* This function is invoked on receipt of a REPL_NEED_INSTINFO message.
* After doing some checks, it sends back a REPL_INSTINFO message containing the instance information.
* If any of the checks fail it issues the appropriate error right away.
* There can be TWO callers. One is the receiver server and another fetchresync rollback.
* is_rcvr_srvr is 1 for the former and 0 for the latter.
*/
void gtmrecv_check_and_send_instinfo(repl_needinst_msg_ptr_t need_instinfo_msg, boolean_t is_rcvr_srvr)
{
boolean_t remote_side_is_supplementary, grab_lock_needed, lms_group_different;
repl_inst_hdr_ptr_t inst_hdr;
repl_instinfo_msg_t instinfo_msg;
repl_inst_uuid *strm_start, *strm_top, *strm_info;
FILE *log_fp;
int reuse_slot, first_usable_slot;
uint4 creator_pid;
seq_num strm_jnl_seqno;
sgmnt_addrs *repl_csa;
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
assert(!jgbl.mur_rollback || !jgbl.mur_options_forward); /* ROLLBACK -FORWARD should not call this function */
repl_csa = &FILE_INFO(jnlpool->jnlpool_dummy_reg)->s_addrs;
grab_lock_needed = is_rcvr_srvr || (jnlpool && (NULL != jnlpool->jnlpool_ctl) && !repl_csa->hold_onto_crit);
remote_side_is_supplementary = need_instinfo_msg->is_supplementary;
remote_side->is_supplementary = remote_side_is_supplementary;
assert(remote_side->endianness_known); /* ensure remote_side->cross_endian is reliable */
/* If the remote source server is <= V62000 it would have endian converted "need_instinfo_msg" ONLY IF
* src_jnl_ver < rcv_jnl_ver. If remote source server is > V62000 it would have endian converted unconditionally.
* So we don't need to do the endian conversion in either of those cases. The need_instinfo_msg->proto_ver will let
* us distinguish the > V62000 vs <= V62000 case. But if src version is <= V62000, we cannot easily distinguish what
* the other side's jnl_ver is so we don't know if the source server would have endian converted or not. We work around
* it by assuming that pids are limited to a max of 2**24 and so the most significant byte of the 4-byte pid should be 0.
* This trick might not work always in case an OS supports > 2**24 pid number (I don't know of any now) OR if both the
* most and least significant bytes of the 4-byte pid are 0, but if it does not, the worst we will see is a) an incorrect
* INSNOTJOINED error issued by the receiver server OR b) an incorrect lms_group_info copied over into the instance file
* (which will soon elicit some other related error). We can address (a) by trying out endian converting again below to
* see if that prevents the error and if so use that instead but we cannot address (b) with this trick.
* We can live with since the fix for (b) is to upgrade the receiver side to GTM V62001 or higher.
*/
if (remote_side->cross_endian && (REPL_PROTO_VER_XENDIANFIXES > need_instinfo_msg->proto_ver))
{
creator_pid = need_instinfo_msg->lms_group_info.creator_pid;
if (!(creator_pid & 0xff) && (creator_pid & 0xff000000))
ENDIAN_CONVERT_REPL_INST_UUID(&need_instinfo_msg->lms_group_info);
}
assert((is_rcvr_srvr && (NULL != gtmrecv_log_fp)) || (!is_rcvr_srvr && (NULL == gtmrecv_log_fp)));
assert(!is_rcvr_srvr || !repl_csa->hold_onto_crit);
assert(is_rcvr_srvr || !jgbl.onlnrlbk || repl_csa->hold_onto_crit);
log_fp = !is_rcvr_srvr ? stdout : gtmrecv_log_fp;
repl_log(log_fp, TRUE, TRUE, "Received REPL_NEED_INSTINFO message from primary instance [%s]\n",
need_instinfo_msg->instname);
inst_hdr = jnlpool->repl_inst_filehdr;
assert(NULL != inst_hdr);
/* The fact that we came here (REPL_NEED_INSTINFO message) implies the source server understands the
* supplementary protocol. In that case, make sure -UPDATERESYNC if specified at receiver server startup
* had a value as well. If not, issue error.
*/
if (is_rcvr_srvr && recvpool.gtmrecv_local->updateresync && (FD_INVALID == recvpool.gtmrecv_local->updresync_instfile_fd))
{
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_UPDSYNCINSTFILE, 0, ERR_TEXT, 2,
LEN_AND_LIT("Source side is >= V5.5-000 implies -UPDATERESYNC needs a value specified"));
assert(FALSE); /* we don't expect the rts_error to return control */
}
/* We usually expect the LMS group info to be non-NULL on both primary and secondary. An exception is if
* both of them are being brought up for the first time using a GT.M version that supports supplementary instances.
* In this case, if the primary was brought up as a root primary, then it should still have the group info filled.
* Assert that.
*/
assert(inst_hdr->lms_group_info.created_time
|| need_instinfo_msg->lms_group_info.created_time || !need_instinfo_msg->is_rootprimary);
assert((is_rcvr_srvr && (NULL != jnlpool->jnlpool_ctl)) || (!is_rcvr_srvr && (NULL == jnlpool->jnlpool_ctl))
|| jgbl.onlnrlbk || (jgbl.mur_rollback && INST_FREEZE_ON_ERROR_POLICY));
/* If this instance is supplementary and the journal pool exists (to indicate whether updates are enabled or not
* which in turn helps us know whether this is an originating instance or not) do some additional checks.
*/
if (is_rcvr_srvr && inst_hdr->is_supplementary)
{
assert(NULL != jnlpool->jnlpool_ctl);
if (!jnlpool->jnlpool_ctl->upd_disabled)
{ /* this supplementary instance was started with -UPDOK. Issue error if source is also supplementary */
if (need_instinfo_msg->is_supplementary)
{
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_NOSUPPLSUPPL, 4,
LEN_AND_STR((char *)inst_hdr->inst_info.this_instname),
LEN_AND_STR((char *)need_instinfo_msg->instname));
assert(FALSE); /* we don't expect the rts_error to return control */
}
} else
{ /* this supplementary instance was started with -UPDNOTOK. Issue error if source is not supplementary */
if (!need_instinfo_msg->is_supplementary)
{
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_SUPRCVRNEEDSSUPSRC, 4,
LEN_AND_STR((char *)inst_hdr->inst_info.this_instname),
LEN_AND_STR((char *)need_instinfo_msg->instname));
assert(FALSE); /* we don't expect the rts_error to return control */
}
}
}
/* Assert that if the receiver side is a root primary (i.e. has updates enabled), it better be a
* supplementary instance and have the LMS group info filled in. Exception is if this a FETCHRESYNC
* ROLLBACK run on an instance which was once primary (not necessarily a supplementary one).
*/
assert((NULL == jnlpool->jnlpool_ctl) || jnlpool->jnlpool_ctl->upd_disabled
|| (inst_hdr->is_supplementary && IS_REPL_INST_UUID_NON_NULL(inst_hdr->lms_group_info))
|| jgbl.onlnrlbk || (jgbl.mur_rollback && INST_FREEZE_ON_ERROR_POLICY));
/* Check if primary and secondary are in same LMS group. Otherwise issue error. An exception is if the group info has
* not yet been filled in after instance file creation. In that case, copy the info from primary and skip the error check.
*/
if (IS_REPL_INST_UUID_NON_NULL(inst_hdr->lms_group_info))
{ /* LMS Group info has been initialized. Compare with that of the source side */
if (memcmp(&need_instinfo_msg->lms_group_info, &inst_hdr->lms_group_info, SIZEOF(inst_hdr->lms_group_info)))
{ /* Source and Receiver are part of DIFFERENT LMS Groups. If this instance is supplementary and remote
* side is not supplementary then we expect them to be different. Otherwise issue error.
* Note that because of an issue with endian-conversion (see big comment at start of this function),
* it is possible this lms_group_info would be good (i.e. passed the memcmp) if endian-converted. So
* try that once and if it still fails, go ahead with issuing an error.
*/
lms_group_different = TRUE;
if (remote_side->cross_endian && (REPL_PROTO_VER_XENDIANFIXES > need_instinfo_msg->proto_ver))
{
ENDIAN_CONVERT_REPL_INST_UUID(&need_instinfo_msg->lms_group_info);
if (!memcmp(&need_instinfo_msg->lms_group_info,
&inst_hdr->lms_group_info, SIZEOF(inst_hdr->lms_group_info)))
lms_group_different = FALSE; /* stay with converted lms_group_info as it is good */
else
{ /* neither works so return to the original */
ENDIAN_CONVERT_REPL_INST_UUID(&need_instinfo_msg->lms_group_info);
}
}
} else
lms_group_different = FALSE;
if (lms_group_different)
{
if (!inst_hdr->is_supplementary || remote_side_is_supplementary)
{
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_INSNOTJOINED, 4,
LEN_AND_STR((char *)inst_hdr->inst_info.this_instname),
LEN_AND_STR((char *)need_instinfo_msg->instname));
assert(FALSE); /* we don't expect the rts_error to return control */
}
} else
{ /* Primary and Secondary are part of SAME LMS Group. If this instance is supplementary and remote
* side is not supplementary then we expect them to be different. Issue error in that case.
*/
if (inst_hdr->is_supplementary && !remote_side_is_supplementary)
{
assert(!need_instinfo_msg->is_supplementary); /* else NOSUPPLSUPPL error must have been issued */
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_INSROLECHANGE, 4,
LEN_AND_STR((char *)inst_hdr->inst_info.this_instname),
LEN_AND_STR((char *)need_instinfo_msg->instname));
assert(FALSE); /* we don't expect the rts_error to return control */
}
}
if (is_rcvr_srvr && recvpool.gtmrecv_local->updateresync
&& (FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd))
{ /* Caller is receiver server and -UPDATERESYNC=<INSTFILENAME> was specified. Check if the lms_group_info
* of that file matches that of the source side. If not issue error.
*/
if (memcmp(&recvpool.gtmrecv_local->updresync_lms_group,
&need_instinfo_msg->lms_group_info, SIZEOF(inst_hdr->lms_group_info)))
{
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_UPDSYNCINSTFILE, 0, ERR_TEXT, 2,
LEN_AND_LIT("Specified input instance file does not have same "
"LMS Group information as source server instance"));
}
}
} else if (IS_REPL_INST_UUID_NON_NULL(need_instinfo_msg->lms_group_info))
{ /* LMS Group info is NULL in instance file header. Initialize it in journal pool from the value
* on the primary side AND flush the changes to disk. Get lock before manipulating it.
* If caller is rollback, no other process can be touching the instance file until we are done so
* no need of the lock in that case.
*/
if (grab_lock_needed)
{
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
if (is_rcvr_srvr)
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
}
inst_hdr->lms_group_info = need_instinfo_msg->lms_group_info;
repl_inst_flush_filehdr();
if (grab_lock_needed)
rel_lock(jnlpool->jnlpool_dummy_reg);
}
/* If this instance is supplementary and remote side is not, then find out which stream # the non-supplementary source
* corresponds to. Issue error if the source LMS group is unknown in the instance file. If this is non-supplementary,
* the stream index is 0.
*/
if (inst_hdr->is_supplementary && !remote_side_is_supplementary)
{
assert(ARRAYSIZE(inst_hdr->strm_group_info) == (MAX_SUPPL_STRMS - 1));
assert(SIZEOF(repl_inst_uuid) == SIZEOF(inst_hdr->strm_group_info[0]));
strm_start = &inst_hdr->strm_group_info[0];
strm_top = strm_start + ARRAYSIZE(inst_hdr->strm_group_info);
/* Do this check under a lock as another receiver server started with -UPDATERESYNC=<INSTFILENAME> (not supported
* now but will be in the future) could be changing the "strm_group_info" array concurrently.
* If a matching stream is found, update gtmrecv_local->strm_index to reflect this while still holding lock.
* This field is checked by the -UPDATERESYNC to see if a receiver has a given stream # actively in use.
*/
if (grab_lock_needed)
{
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
if (is_rcvr_srvr)
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
}
reuse_slot = 0;
if (is_rcvr_srvr && recvpool.gtmrecv_local->updateresync && gtmrecv_options.reuse_specified)
{ /* If -REUSE was specified, check if instance name specified matches any existing slot.
* If slot has already been found, don't search any more.
*/
for (strm_info = strm_start; strm_info < strm_top; strm_info++)
{
if (IS_REPL_INST_UUID_NULL(*strm_info))
continue;
if (!STRCMP(gtmrecv_options.reuse_instname, strm_info->this_instname))
{
reuse_slot = (strm_info - strm_start) + 1;
break;
}
}
if (strm_info == strm_top)
{ /* -REUSE specified an instance name that is not present in any of the 15 strm_group slots */
rel_lock(jnlpool->jnlpool_dummy_reg);
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_REUSEINSTNAME, 0, ERR_TEXT, 2,
LEN_AND_LIT("Instance name in REUSE does not match any of 15 slots in instance file"));
}
assert(reuse_slot);
}
first_usable_slot = 0;
strm_index = 0;
for (strm_info = strm_start; strm_info < strm_top; strm_info++)
{
if (IS_REPL_INST_UUID_NULL(*strm_info))
{
if (!first_usable_slot)
first_usable_slot = (strm_info - strm_start) + 1;
continue;
}
if (!memcmp(&need_instinfo_msg->lms_group_info, strm_info, SIZEOF(repl_inst_uuid)))
{ /* Found the stream corresponding to the source side */
strm_index = (strm_info - strm_start) + 1;
break;
}
}
if (strm_info == strm_top)
{ /* Non-supplementary source is unknown to this supplementary instance */
if (reuse_slot) /* -REUSE specified and did locate a reusable slot. Use it. */
{
assert(is_rcvr_srvr);
strm_index = reuse_slot;
/* having reused the slot as specified, the qualifier has no further use & may be disruptive */
gtmrecv_options.reuse_specified = FALSE;
} else if (gtmrecv_options.resume_specified)
{
assert(is_rcvr_srvr);
strm_index = gtmrecv_options.resume_strm_num;
} else if (is_rcvr_srvr && recvpool.gtmrecv_local->updateresync)
{
if (first_usable_slot)
strm_index = first_usable_slot;
else
{
if (grab_lock_needed)
rel_lock(jnlpool->jnlpool_dummy_reg);
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_UPDSYNCINSTFILE, 0, ERR_TEXT, 2,
LEN_AND_LIT("No empty slot found. Specify REUSE to choose one for reuse"));
}
} else
{
if (grab_lock_needed)
rel_lock(jnlpool->jnlpool_dummy_reg);
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_INSUNKNOWN, 4,
LEN_AND_STR((char *)inst_hdr->inst_info.this_instname),
LEN_AND_STR((char *)need_instinfo_msg->instname));
assert(FALSE); /* we don't expect the rts_error to return control */
}
/* Since we did not find the stream in the existing instance file but did find a slot, fill that slot
* while we have the lock on the instance file. This way another -updateresync startup of a receiver
* server (when we have multiple receiver server support) will see this slot taken when it gets the lock.
*/
assert(is_rcvr_srvr && recvpool.gtmrecv_local->updateresync
&& (FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd));
assert(!memcmp(&recvpool.gtmrecv_local->updresync_lms_group,
&need_instinfo_msg->lms_group_info, SIZEOF(inst_hdr->lms_group_info)));
assert(0 < strm_index);
strm_info = &strm_start[strm_index - 1];
assert(strm_info < strm_top);
assert(IS_REPL_INST_UUID_NON_NULL(need_instinfo_msg->lms_group_info));
*strm_info = need_instinfo_msg->lms_group_info;
repl_inst_flush_filehdr();
} else if ((gtmrecv_options.resume_specified) && (gtmrecv_options.resume_strm_num != strm_index))
{ /* If -RESUME was specified, then the slot it matched must be same as slot found without its use */
assert(is_rcvr_srvr);
rel_lock(jnlpool->jnlpool_dummy_reg);
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_RESUMESTRMNUM, 0, ERR_TEXT, 2, LEN_AND_LIT("Source side LMS "
"group is found in instance file but RESUME specifies different stream number"));
} else if (reuse_slot && (reuse_slot != strm_index))
{ /* If -REUSE was specified, then the slot it matched must be same as slot found without its use */
assert(is_rcvr_srvr);
rel_lock(jnlpool->jnlpool_dummy_reg);
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_REUSEINSTNAME, 0, ERR_TEXT, 2, LEN_AND_LIT("Source side LMS "
"group is found in instance file but REUSE specifies different instance name"));
}
assert(INVALID_SUPPL_STRM != strm_index);
assert(0 < strm_index);
assert(MAX_SUPPL_STRMS > strm_index);
/* Maintain stream slot # in shared memory as well so another -UPDATERESYNC=<INSTFILENAME>
* can see which stream# is actively in use by this receiver server. */
if (is_rcvr_srvr)
{
if (recvpool.gtmrecv_local->strm_index && (strm_index != recvpool.gtmrecv_local->strm_index))
{ /* This receiver server has already connected to a source server with a different stream #.
* Since mixing of multiple stream journal records in the same receive pool confuses the
* update process, issue error. Note: This limitation "might" be removed in the future.
*/
rel_lock(jnlpool->jnlpool_dummy_reg);
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(4) ERR_RCVRMANYSTRMS, 2,
strm_index, recvpool.gtmrecv_local->strm_index);
}
recvpool.gtmrecv_local->strm_index = strm_index;
repl_log(gtmrecv_log_fp, TRUE, TRUE,
"Determined non-supplementary source Stream # = %d\n", strm_index);
assert(IS_REPL_INST_UUID_NON_NULL(need_instinfo_msg->lms_group_info));
recvpool.gtmrecv_local->remote_lms_group = need_instinfo_msg->lms_group_info;
rel_lock(jnlpool->jnlpool_dummy_reg);
} else
repl_log(stdout, TRUE, TRUE, "Determined non-supplementary source Stream # = %d\n", strm_index);
/* Compute non-zero strm_jnl_seqno to send across in the REPL_INSTINFO message.
* For receiver server
* If updateresync startup and this is the first connection to a source server
* If -RESUME is not specified :
* --> gtmrecv_local->updresync_jnl_seqno
* If -RESUME is specified :
* --> jnlpool_ctl->strm_seqno[gtmrecv_options.resume_strm_num]
* If updateresync startup and this is not the first connection to a source server
* --> recvpool_ctl->jnl_seqno
* If normal startup and this is the first connection to a source server
* --> inst_hdr->strm_seqno[strm_index] where strm_index > 0
* If this is not the first connection to a source server
* --> recvpool_ctl->jnl_seqno
* In case of instance crash
* --> No need to worry about this case as we would have otherwise issued a
* --> REPLREQROLLBACK error when the source server for this instance started up after the crash.
* For rollback
* If normal startup after a clean shutdown of the instance file
* --> inst_hdr->strm_seqno[strm_index] where strm_index > 0
* In case of crash shutdown of instance file
* --> Take max of cs_data->strm_seqno[strm_index] across all databases and use this (strm_index > 0)
*/
if (is_rcvr_srvr)
{
if (recvpool.upd_proc_local->read_jnl_seqno)
strm_jnl_seqno = recvpool.recvpool_ctl->jnl_seqno;
else
{
if (recvpool.gtmrecv_local->updateresync)
{
assert(FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd);
if (!gtmrecv_options.resume_specified)
strm_jnl_seqno = recvpool.gtmrecv_local->updresync_jnl_seqno;
else
{
strm_jnl_seqno = jnlpool->jnlpool_ctl->strm_seqno[gtmrecv_options.resume_strm_num];
/* It is possible for the strm_seqno to be 0. This implies the stream has
* had no updates. In that case, ideally this value should have been 1.
* But because we want to differentiate a stream that has had updates from
* stream numbers where there is no interest, we follow this convention.
* Therefore, in this case, reset the 0 back to 1 so we never send a zero
* seqno to the other side.
*/
if (!strm_jnl_seqno)
strm_jnl_seqno = 1;
}
} else
strm_jnl_seqno = jnlpool->jnlpool_ctl->strm_seqno[strm_index];
assert(0 == GET_STRM_INDEX(strm_jnl_seqno));
}
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Sending Stream Seqno = "INT8_FMT" "INT8_FMTX"\n",
strm_jnl_seqno, strm_jnl_seqno);
} else
{
if (jnlpool->repl_inst_filehdr->crash)
strm_jnl_seqno = mur_get_max_strm_reg_seqno(strm_index);
else
{
assert((NULL == jnlpool->jnlpool_ctl) || jgbl.onlnrlbk);
assert((NULL == jnlpool->jnlpool_ctl)
|| (jnlpool->jnlpool_ctl->strm_seqno[strm_index] == inst_hdr->strm_seqno[strm_index]));
strm_jnl_seqno = jnlpool->repl_inst_filehdr->strm_seqno[strm_index];
}
repl_log(stdout, TRUE, TRUE, "Sending Stream Seqno = "INT8_FMT" "INT8_FMTX"\n",
strm_jnl_seqno, strm_jnl_seqno);
}
} else
{
strm_jnl_seqno = 0; /* actually no need to initialize this since source server will not look at this
* field in this case but still be safe */
if (inst_hdr->is_supplementary)
{ /* In case caller is receiver server, "strm_index" would have been already set to 0 in jnlpool_init.c.
* But in case caller is rollback, "strm_index" would still be set to -1. In this case, set it to 0.
*/
assert(!is_rcvr_srvr || (0 == strm_index));
strm_index = 0;
}
DEBUG_ONLY(
if (is_rcvr_srvr)
NULL_INITIALIZE_REPL_INST_UUID(recvpool.gtmrecv_local->remote_lms_group);
)
}
assert(!is_rcvr_srvr || (INVALID_SUPPL_STRM == recvpool.gtmrecv_local->strm_index)
|| ((0 <= recvpool.gtmrecv_local->strm_index) && (MAX_SUPPL_STRMS > recvpool.gtmrecv_local->strm_index)));
/* Initialize the remote side protocol version from "proto_ver" field of this msg */
assert(REPL_PROTO_VER_SUPPLEMENTARY <= need_instinfo_msg->proto_ver);
remote_side->proto_ver = need_instinfo_msg->proto_ver;
/*************** Send REPL_INSTINFO message ***************/
memset(&instinfo_msg, 0, SIZEOF(instinfo_msg));
memcpy(instinfo_msg.instname, inst_hdr->inst_info.this_instname, MAX_INSTNAME_LEN - 1);
if (grab_lock_needed)
{
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
if (is_rcvr_srvr)
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
}
instinfo_msg.was_rootprimary = (unsigned char)repl_inst_was_rootprimary();
if (grab_lock_needed)
rel_lock(jnlpool->jnlpool_dummy_reg);
if (!is_rcvr_srvr)
murgbl.was_rootprimary = instinfo_msg.was_rootprimary;
instinfo_msg.strm_jnl_seqno = strm_jnl_seqno;
/* strm_jnl_seqno is not expected to be zero unless this is a non-supplementary instance (like A->B) in which case
* strm_seqno is not maintained OR the remote side is supplementary (like P->Q) in which case the two instances do
* not communicate in-terms of strm_seqno (once the handshake is established)
*/
assert(strm_jnl_seqno || !inst_hdr->is_supplementary || remote_side_is_supplementary);
instinfo_msg.lms_group_info = inst_hdr->lms_group_info;
assert(remote_side->endianness_known); /* only then is remote_side->cross_endian reliable */
if (remote_side->cross_endian)
{
ENDIAN_CONVERT_REPL_INST_UUID(&instinfo_msg.lms_group_info);
instinfo_msg.strm_jnl_seqno = GTM_BYTESWAP_64(instinfo_msg.strm_jnl_seqno);
}
gtmrecv_repl_send((repl_msg_ptr_t)&instinfo_msg, REPL_INSTINFO, SIZEOF(repl_instinfo_msg_t), "REPL_INSTINFO", MAX_SEQNO);
if (repl_connection_reset || (is_rcvr_srvr && gtmrecv_wait_for_jnl_seqno))
return;
/* Do not allow an instance which was formerly a root primary or which still
* has a non-zero value of "zqgblmod_seqno" to start up as a tertiary. The only exception is
* if this is P (supplementary instance) receiving from a non-supplementary instance.
* In that case, P can never be a root primary of the non-supplementary group and therefore
* cannot be affected by lost transactions being applied from the non-supplementary group.
*/
if ((!inst_hdr->is_supplementary || remote_side_is_supplementary)
&& !need_instinfo_msg->is_rootprimary
&& (instinfo_msg.was_rootprimary
|| (is_rcvr_srvr && jnlpool->jnlpool_ctl->max_zqgblmod_seqno)))
{
if (is_rcvr_srvr)
{
gtm_putmsg_csa(CSA_ARG(NULL) VARLSTCNT(4) ERR_PRIMARYNOTROOT, 2,
LEN_AND_STR((char *) need_instinfo_msg->instname));
gtmrecv_autoshutdown(); /* should not return */
} else
rts_error_csa(CSA_ARG(NULL) VARLSTCNT(4) ERR_PRIMARYNOTROOT, 2,
LEN_AND_STR((char *) need_instinfo_msg->instname));
assert(FALSE);
}
if (is_rcvr_srvr)
memcpy(jnlpool->jnlpool_ctl->primary_instname, need_instinfo_msg->instname, MAX_INSTNAME_LEN - 1);
}
STATICFNDEF int gtmrecv_start_onln_rlbk(void)
{
char command[ONLN_RLBK_CMD_MAXLEN + 1], *errptr;
int status, save_errno, cmdlen;
gtmrecv_local_ptr_t gtmrecv_local;
assert(!have_crit(CRIT_HAVE_ANY_REG));
gtmrecv_local = recvpool.gtmrecv_local;
MEMCPY_LIT(command, MUPIP_DIST_STR);
cmdlen = STR_LIT_LEN(MUPIP_DIST_STR);
MEMCPY_LIT(&command[cmdlen], ONLN_RLBK_CMD);
cmdlen += STR_LIT_LEN(ONLN_RLBK_CMD);
if (gtmrecv_options.autorollback_verbose)
{
MEMCPY_LIT(&command[cmdlen], ONLN_RLBK_VERBOSE);
cmdlen += STR_LIT_LEN(ONLN_RLBK_VERBOSE);
}
MEMCPY_LIT(&command[cmdlen], ONLN_RLBK_QUALIFIERS);
cmdlen += STR_LIT_LEN(ONLN_RLBK_QUALIFIERS);
assert(0 < gtmrecv_local->listen_port);
SNPRINTF(&command[cmdlen], ONLN_RLBK_CMD_MAXLEN - cmdlen, "%d", gtmrecv_local->listen_port); /* will add '\0' at the end */
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Executing %s\n", command);
status = SYSTEM(((char *)command));
if (0 != status)
{
if (-1 == status)
{
save_errno = errno;
errptr = (char *)STRERROR(save_errno);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "SYSTEM command failed: %s\n", errptr);
} else
{ /* ONLINE ROLLBACK returned a non-zero status */
repl_log(gtmrecv_log_fp, TRUE, TRUE, "ONLINE FETCHRESYNC ROLLBACK exited with code - %d\n", status);
}
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Could not complete ONLINE FETCHRESYNC ROLLBACK due to the above errors\n");
} else
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "ONLINE FETCHRESYNC ROLLBACK completed successfully\n");
}
return status;
}
/* We are here because a grab_lock or grab_crit saw a concurrent online rollback and we need to break the connection and
* re-establish with the new sequence number (the rolled back one). Wait for the update process to let us know the new sequence
* number from where we should start requesting transactions from.
* Ideally this function should have been a part of onln_rlbk.c. However, the function calls into gtmrecv_poll_actions which
* is strictly in the libmupip archive. Moving the function into onln_rlbk.c means it will try to pull in gtmrecv_poll_actions
* and all the modules that it calls which is something that we don't desire. So, keep this function isolated from onln_rlbk.c
*/
void gtmrecv_onln_rlbk_clnup(void)
{
boolean_t connection_already_reset;
assert(NULL != gtmrecv_log_fp);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "---> ONLINE ROLLBACK. Current Jnlpool Seqno : "INT8_FMT"\n",
jnlpool->jnlpool_ctl->jnl_seqno);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Waiting for update process to set recvpool_ctl->onln_rlbk_flag\n");
connection_already_reset = repl_connection_reset;
assert(!gtmrecv_wait_for_jnl_seqno);
while (TRUE)
{
SHORT_SLEEP(GTMRECV_WAIT_FOR_UPD_PROGRESS);
gtmrecv_poll_actions(data_len, buff_unprocessed, buffp);
/* If connection was already closed before we came here (possible if receiver server closed the connection in
* response to a REPL_ROLLBACK_FIRST message) after spawning off an online rollback, then we should NOT check for
* repl_connection_reset. This way, we avoid breaking prematurely from this loop without waiting for the update
* process to acknowledge the online rollback. In this case, wait for gtmrecv_wait_for_jnl_seqno to be set to TRUE
* by gtmrecv_poll_actions.
*/
if ((!connection_already_reset && repl_connection_reset) || gtmrecv_wait_for_jnl_seqno)
break;
}
return;
}
/* This function is invoked on receipt of a REPL_NEED_HISTINFO message.
* This in turn sends a REPL_HISTINFO message containing the history information.
*/
void gtmrecv_send_histinfo(repl_histinfo *cur_histinfo)
{
repl_histinfo1_msg_t histinfo1_msg;
repl_histinfo2_msg_t histinfo2_msg;
repl_histinfo_msg_t histinfo_msg;
FILE *log_fp;
char remote_proto_ver;
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
/* If sending history from a supplementary to a non-supplementary version, assert that the history record
* particularly the "strm_seqno" is 0 as a non-zero value is not understood by a non-supplementary instance.
*/
assert((NULL == this_side) || (this_side->is_supplementary == jnlpool->repl_inst_filehdr->is_supplementary));
assert(!jnlpool->repl_inst_filehdr->is_supplementary || remote_side->is_supplementary || !cur_histinfo->strm_seqno);
remote_proto_ver = remote_side->proto_ver;
assert(REPL_PROTO_VER_MULTISITE <= remote_proto_ver);
assert(remote_side->endianness_known); /* only then is remote_side->cross_endian reliable */
if (REPL_PROTO_VER_SUPPLEMENTARY > remote_proto_ver)
{ /* Remote side does not support supplementary protocol. Send OLDER histinfo messages */
/*************** Send REPL_OLD_TRIPLEINFO1 message ***************/
histinfo1_msg.start_seqno = !remote_side->cross_endian ?
cur_histinfo->start_seqno : GTM_BYTESWAP_64(cur_histinfo->start_seqno);
memcpy(histinfo1_msg.instname, cur_histinfo->root_primary_instname, MAX_INSTNAME_LEN - 1);
histinfo1_msg.instname[MAX_INSTNAME_LEN - 1] = '\0';
gtmrecv_repl_send((repl_msg_ptr_t)&histinfo1_msg, REPL_OLD_TRIPLEINFO1, MIN_REPL_MSGLEN,
"REPL_OLD_TRIPLEINFO1", cur_histinfo->start_seqno);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
/*************** Send REPL_OLD_TRIPLEINFO2 message ***************/
if (!remote_side->cross_endian)
{
histinfo2_msg.start_seqno = cur_histinfo->start_seqno;
histinfo2_msg.cycle = cur_histinfo->root_primary_cycle;
histinfo2_msg.histinfo_num = cur_histinfo->histinfo_num;
} else
{
histinfo2_msg.start_seqno = GTM_BYTESWAP_64(cur_histinfo->start_seqno);
histinfo2_msg.cycle = GTM_BYTESWAP_32(cur_histinfo->root_primary_cycle);
histinfo2_msg.histinfo_num = GTM_BYTESWAP_32(cur_histinfo->histinfo_num);
}
gtmrecv_repl_send((repl_msg_ptr_t)&histinfo2_msg, REPL_OLD_TRIPLEINFO2, MIN_REPL_MSGLEN,
"REPL_OLD_TRIPLEINFO2", cur_histinfo->start_seqno);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
} else
{ /* Remote side does support supplementary protocol. Send NEWER histinfo message. */
histinfo_msg.history = *cur_histinfo;
histinfo_msg.history.root_primary_instname[MAX_INSTNAME_LEN - 1] = '\0'; /* just in case */
if (remote_side->cross_endian)
ENDIAN_CONVERT_REPL_HISTINFO(&histinfo_msg.history);
gtmrecv_repl_send((repl_msg_ptr_t)&histinfo_msg, REPL_HISTINFO, SIZEOF(repl_histinfo_msg_t),
"REPL_HISTINFO", cur_histinfo->start_seqno);
}
log_fp = (NULL == gtmrecv_log_fp) ? stdout : gtmrecv_log_fp;
repl_dump_histinfo(log_fp, TRUE, FALSE, "History sent", cur_histinfo);
}
STATICFNDEF void prepare_recvpool_for_write(gtm_uint64_t datalen, gtm_uint64_t pre_filter_write_len)
{
recvpool_ctl_ptr_t recvpool_ctl;
recvpool_ctl = recvpool.recvpool_ctl;
if (datalen > recvpool_size)
{ /* Too large a transaction to be accommodated in the Receive Pool */
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(7) ERR_REPLTRANS2BIG, 5, &recvpool_ctl->jnl_seqno,
&datalen, &pre_filter_write_len, LEN_AND_LIT("Receive"));
}
if (write_loc > (recvpool_size - datalen))
{
REPL_DEBUG_ONLY(
if (recvpool_ctl->wrapped)
REPL_DPRINT1("Update Process too slow. Waiting for it to free up space and wrap\n");
)
while (recvpool_ctl->wrapped)
{ /* Wait till the updproc wraps */
SHORT_SLEEP(GTMRECV_WAIT_FOR_UPD_PROGRESS);
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
assert(recvpool_ctl->wrapped == FALSE);
recvpool_ctl->write_wrap = write_wrap = write_loc;
/* The update process reads (a) "recvpool_ctl->write" first. If "write" is not equal to
* "upd_proc_local->read", it then reads (b) "recvpool_ctl->write_wrap" and assumes that
* "write_wrap" holds a non-stale value. This is in turn used to compare "temp_read" and
* "write_wrap" to determine how much of unprocessed data there is in the receive pool. If
* it so happens that the receiver server sets "write_wrap" in the above line to a value
* that is lesser than its previous value (possible if in the previous wrap of the pool,
* transactions used more portions of the pool than in the current wrap), it is important
* that the update process sees the updated value of "write_wrap" as long as it sees the
* corresponding update to "write". This is because it will otherwise end up processing
* the tail section of the receive pool (starting from the uptodate value of "write" to the
* stale value of "write_wrap") that does not contain valid journal data. For this read order
* dependency to hold good, the receiver server needs to do a write memory barrier
* after updating "write_wrap" but before updating "write". The update process
* will do a read memory barrier after reading "wrapped" but before reading "write".
*/
SHM_WRITE_MEMORY_BARRIER;
/* The update process reads (a) "recvpool_ctl->wrapped" first and then reads (b)
* "recvpool_ctl->write". If "wrapped" is TRUE, it assumes that "write" will never hold a stale
* value that reflects a corresponding previous state of "wrapped" (i.e. "write" will point to
* the beginning of the receive pool, either 0 or a small non-zero value instead of pointing
* to the end of the receive pool). For this to hold good, the receiver server needs to do
* a write memory barrier after updating "write" but before updating "wrapped". The update
* process will do a read memory barrier after reading "wrapped" but before reading "write".
*/
recvpool_ctl->write = write_loc = 0;
SHM_WRITE_MEMORY_BARRIER;
recvpool_ctl->wrapped = TRUE;
}
DO_FLOW_CONTROL(write_loc);
}
STATICFNDEF void copy_to_recvpool(uchar_ptr_t databuff, gtm_uint64_t datalen)
{
gtm_uint64_t upd_read;
gtm_uint64_t future_write;
upd_proc_local_ptr_t upd_proc_local;
recvpool_ctl_ptr_t recvpool_ctl;
recvpool_ctl = recvpool.recvpool_ctl;
upd_proc_local = recvpool.upd_proc_local;
future_write = write_loc + datalen;
assertpro(future_write >= write_loc); /* prepare_recvpool_for_write should have handled this */
upd_read = upd_proc_local->read;
REPL_DEBUG_ONLY(
if (recvpool_ctl->wrapped && (upd_read <= future_write))
{
REPL_DPRINT1("Update Process too slow. Waiting for it to free up space\n");
}
)
while (recvpool_ctl->wrapped && (upd_read <= future_write))
{ /* Write will cause overflow. Wait till there is more space available */
SHORT_SLEEP(GTMRECV_WAIT_FOR_UPD_PROGRESS);
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
upd_read = upd_proc_local->read;
}
memcpy(recvpool.recvdata_base + write_loc, databuff, datalen);
write_loc = future_write;
if (write_loc > write_wrap)
write_wrap = write_loc;
}
STATICFNDEF void wait_for_updproc_to_clear_backlog(void)
{
upd_proc_local_ptr_t upd_proc_local;
recvpool_ctl_ptr_t recvpool_ctl;
recvpool_ctl = recvpool.recvpool_ctl;
upd_proc_local = recvpool.upd_proc_local;
while (upd_proc_local->read != recvpool_ctl->write)
{
SHORT_SLEEP(GTMRECV_WAIT_FOR_UPD_PROGRESS);
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
}
STATICFNDEF void process_tr_buff(int msg_type)
{
recvpool_ctl_ptr_t recvpool_ctl;
seq_num log_seqno, recv_jnl_seqno, upd_seqno, diff_seqno;
uint4 in_size, out_size, upd_read, max_strm_histinfo;
boolean_t filter_pass = FALSE, is_new_histrec, is_repl_cmpc;
uchar_ptr_t save_buffp, save_filter_buff, in_buff;
int idx, status, num_strm_histinfo;
qw_num msg_total;
repl_old_triple_jnl_t old_triple_content;
uLongf destlen;
int cmpret, cur_data_len, rc;
repl_msg_ptr_t msgp, msgp_top;
int4 histinfo_strm_num;
uint4 write_len, pre_filter_write_len, pre_filter_write;
boolean_t uncmpfail;
repl_histinfo *cur_histinfo, *pool_histinfo, tmp_histinfo;
repl_histrec_jnl_t *pool_histrec, tmp_histjrec, rcvd_strm_histjrec[MAX_SUPPL_STRMS];
static boolean_t first_histrec = TRUE;
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
recvpool_ctl = recvpool.recvpool_ctl;
is_repl_cmpc = ((REPL_TR_CMP_JNL_RECS == msg_type) || (REPL_TR_CMP_JNL_RECS2 == msg_type));
is_new_histrec = ((REPL_OLD_TRIPLE == msg_type) || (REPL_HISTREC == msg_type));
assert(!is_repl_cmpc || !is_new_histrec); /* HISTINFO records should not be compressed in the pipe */
if (is_repl_cmpc)
{
assert(gtmrecv_max_repl_uncmpmsglen);
destlen = gtmrecv_max_repl_uncmpmsglen;
if (ZLIB_CMPLVL_NONE == gtm_zlib_cmp_level)
{ /* Receiver does not have compression enabled in the first place but yet source server has sent
* compressed records. Stop source server from sending compressed records.
*/
uncmpfail = TRUE;
} else
{
ZLIB_UNCOMPRESS(gtmrecv_uncmpmsgp, destlen, gtmrecv_cmpmsgp, gtmrecv_repl_cmpmsglen, cmpret);
GTM_WHITE_BOX_TEST(WBTEST_REPL_TR_UNCMP_ERROR, cmpret, Z_DATA_ERROR);
recv_jnl_seqno = recvpool_ctl->jnl_seqno;
switch(cmpret)
{
case Z_MEM_ERROR:
assert(FALSE);
repl_log(gtmrecv_log_fp, TRUE, TRUE, GTM_ZLIB_Z_MEM_ERROR_STR
GTM_ZLIB_UNCMP_ERR_SEQNO_STR, recv_jnl_seqno, recv_jnl_seqno);
break;
case Z_BUF_ERROR:
assert(FALSE);
repl_log(gtmrecv_log_fp, TRUE, TRUE, GTM_ZLIB_Z_BUF_ERROR_STR
GTM_ZLIB_UNCMP_ERR_SEQNO_STR, recv_jnl_seqno, recv_jnl_seqno);
break;
case Z_DATA_ERROR:
assert(gtm_white_box_test_case_enabled
&& (WBTEST_REPL_TR_UNCMP_ERROR == gtm_white_box_test_case_number));
repl_log(gtmrecv_log_fp, TRUE, TRUE, GTM_ZLIB_Z_DATA_ERROR_STR
GTM_ZLIB_UNCMP_ERR_SEQNO_STR, recv_jnl_seqno, recv_jnl_seqno);
break;
}
uncmpfail = (Z_OK != cmpret);
if (!uncmpfail)
{
GTM_WHITE_BOX_TEST(WBTEST_REPL_TR_UNCMP_ERROR, destlen, gtmrecv_repl_uncmpmsglen - 1);
if (destlen != gtmrecv_repl_uncmpmsglen)
{ /* decompression did not yield precompressed data length */
assert(gtm_white_box_test_case_enabled
&& (WBTEST_REPL_TR_UNCMP_ERROR == gtm_white_box_test_case_number));
repl_log(gtmrecv_log_fp, TRUE, TRUE, GTM_ZLIB_UNCMPLEN_ERROR_STR
GTM_ZLIB_UNCMP_ERR_SEQNO_STR, destlen, gtmrecv_repl_uncmpmsglen,
recv_jnl_seqno, recv_jnl_seqno);
uncmpfail = TRUE;
}
}
}
if (uncmpfail)
{ /* Since uncompression failed, default to NO compression. Send a REPL_CMP2UNCMP message accordingly */
repl_log(gtmrecv_log_fp, TRUE, TRUE, GTM_ZLIB_UNCMPTRANSITION_STR);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Waiting for update process to clear the backlog first\n");
wait_for_updproc_to_clear_backlog();
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Update process has successfully cleared the backlog\n");
gtmrecv_send_cmp2uncmp = TRUE; /* trigger REPL_CMP2UNCMP message processing */
gtmrecv_poll_actions(data_len, buff_unprocessed, buffp);
assert(repl_connection_reset || (!gtmrecv_send_cmp2uncmp && gtmrecv_wait_for_jnl_seqno));
return;
}
assert(0 == destlen % REPL_MSG_ALIGN);
msgp = (repl_msg_ptr_t)gtmrecv_uncmpmsgp;
msgp_top = (repl_msg_ptr_t)(gtmrecv_uncmpmsgp + destlen);
} else
{
msgp = NULL;
msgp_top = NULL;
}
do
{
assert(remote_side->endianness_known); /* only then is remote_side->cross_endian reliable */
if (is_repl_cmpc)
{
assert(msgp);
assert(msgp_top);
if (msgp >= msgp_top)
{
assert(msgp == msgp_top);
break;
}
/* If primary is of different endianness, endian convert the UNCOMPRESSED message header
* before using the type and len fields (the compressed message header was already endian
* converted as part of receiving the message in do_main_loop())
*/
if (remote_side->cross_endian)
{
msgp->type = GTM_BYTESWAP_32(msgp->type);
msgp->len = GTM_BYTESWAP_32(msgp->len);
}
assert(REPL_TR_JNL_RECS == msgp->type);
cur_data_len = msgp->len - REPL_MSG_HDRLEN;
assert(0 < cur_data_len);
assert(0 == (cur_data_len % REPL_MSG_ALIGN));
PREPARE_RECVPOOL_FOR_WRITE(cur_data_len, 0); /* could update "recvpool_ctl->write" and "write_loc" */
COPY_TO_RECVPOOL((uchar_ptr_t)msgp + REPL_MSG_HDRLEN, cur_data_len);/* uses and updates "write_loc" */
msgp = (repl_msg_ptr_t)((uchar_ptr_t)msgp + cur_data_len + REPL_MSG_HDRLEN);
}
write_off = recvpool_ctl->write;
write_len = (write_loc - write_off);
assert((write_off != write_wrap) || (0 == write_off));
assert(remote_side->jnl_ver);
assert(!remote_side->cross_endian || (V18_JNL_VER <= remote_side->jnl_ver));
if (ENDIAN_CONVERSION_NEEDED(is_new_histrec, this_side->jnl_ver, remote_side->jnl_ver, remote_side->cross_endian))
{
if (SS_NORMAL != (status = repl_tr_endian_convert(remote_side->jnl_ver,
recvpool.recvdata_base + write_off, write_len)))
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(5) ERR_REPLXENDIANFAIL, 3, LEN_AND_LIT("Replicating"),
&recvpool.upd_proc_local->read_jnl_seqno);
}
if (!is_new_histrec)
{
if (NO_FILTER != gtmrecv_filter)
{ /* Need to pass through filter */
pre_filter_write = write_off;
pre_filter_write_len = write_len;
/* If input transaction cannot fit in currently allocated buffer, expand buffer
* directly to that needed length hoping this should be enough (minimizes the # of
* 50% expansions we do). If this is still not enough (because the filter function
* results in bigger sized jnl records) we will then do geometric expansion of this
* buffer in the while loop below as much as needed.
*/
if (write_len > repl_filter_bufsiz)
{
gtmrecv_free_filter_buff();
gtmrecv_alloc_filter_buff(write_len);
}
if (gtmrecv_filter & INTERNAL_FILTER)
{
in_buff = recvpool.recvdata_base + write_off;
in_size = write_len;
while (SS_NORMAL != (status =
repl_filter_old2cur[remote_side->jnl_ver - JNL_VER_EARLIEST_REPL](
in_buff, &in_size, repl_filter_buff, &out_size, repl_filter_bufsiz))
&& (EREPL_INTLFILTER_NOSPC == repl_errno))
{ /* We ran out of space in current buffer. We can try and use the transformed
* records as is and resume transformation from where the space-issue occurred
* but the transformation function might rely on seeing the entire transaction
* context in one shot so better not to take a risk. So free old buffer and
* allocate new buffer and redo transformation from scratch.
*/
gtmrecv_free_filter_buff();
in_size = pre_filter_write_len; /* just in case in_size was modified */
write_len = write_len + (write_len >> 1); /* increase the buffer size by half */
gtmrecv_alloc_filter_buff(write_len);
}
if (SS_NORMAL == status)
write_len = out_size;
else
{
if (EREPL_INTLFILTER_DATA2LONG == repl_errno)
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(4) ERR_JNLSETDATA2LONG, 2,
jnl_source_datalen, jnl_dest_maxdatalen);
else if (EREPL_INTLFILTER_NEWREC == repl_errno)
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(4) ERR_JNLNEWREC, 2,
(unsigned int)jnl_source_rectype,
(unsigned int)jnl_dest_maxrectype);
else
{
INT_FILTER_RTS_ERROR(recvpool_ctl->jnl_seqno, repl_errno); /* no return */
}
}
} else
memcpy(repl_filter_buff, recvpool.recvdata_base + write_off, write_len);
assert(write_len <= repl_filter_bufsiz);
GTMTRIG_ONLY(
if ((unsigned char)V19_JNL_VER <= remote_side->jnl_ver)
{
repl_sort_tr_buff(repl_filter_buff, write_len);
DBG_VERIFY_TR_BUFF_SORTED(repl_filter_buff, write_len);
}
)
if ((gtmrecv_filter & EXTERNAL_FILTER)
&& (SS_NORMAL !=
(status = repl_filter(recvpool_ctl->jnl_seqno, &repl_filter_buff,
(int*)&write_len, &repl_filter_bufsiz))))
repl_filter_error(recvpool_ctl->jnl_seqno, status);
GTMTRIG_ONLY(
/* Ensure that the external filter has not disturbed the sorted sequence of the
* update_num
*/
DEBUG_ONLY(
if ((unsigned char)V19_JNL_VER <= remote_side->jnl_ver)
DBG_VERIFY_TR_BUFF_SORTED(repl_filter_buff, write_len);
)
)
assert(write_len <= repl_filter_bufsiz);
write_loc = write_off; /* reset "write_loc" */
PREPARE_RECVPOOL_FOR_WRITE(write_len, pre_filter_write_len); /* could update "->write"
* and "write_loc" */
COPY_TO_RECVPOOL((uchar_ptr_t)repl_filter_buff, write_len);/* uses and updates "write_loc" */
write_off = recvpool_ctl->write;
repl_recv_postfltr_data_procd += (qw_num)write_len;
filter_pass = TRUE;
} else
{
GTMTRIG_ONLY(
if ((unsigned char)V19_JNL_VER <= remote_side->jnl_ver)
{
repl_sort_tr_buff((uchar_ptr_t)(recvpool.recvdata_base + write_off), write_len);
DBG_VERIFY_TR_BUFF_SORTED((recvpool.recvdata_base + write_off), write_len);
}
)
}
}
if (recvpool_ctl->jnl_seqno - lastlog_seqno >= log_interval)
{
log_seqno = recvpool_ctl->jnl_seqno;
upd_seqno = recvpool.upd_proc_local->read_jnl_seqno;
assert(log_seqno >= upd_seqno);
diff_seqno = (log_seqno - upd_seqno);
trans_recvd_cnt += (log_seqno - lastlog_seqno);
msg_total = repl_recv_data_recvd - buff_unprocessed;
/* Don't include data not yet processed, we'll include that count in a later log */
if (NO_FILTER == gtmrecv_filter)
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "REPL INFO - Seqno : "INT8_FMT" "INT8_FMTX
" Jnl Total : "INT8_FMT" "INT8_FMTX" Msg Total : "INT8_FMT" "INT8_FMTX
" Current backlog : "INT8_FMT" "INT8_FMTX"\n",
log_seqno, log_seqno, repl_recv_data_processed, repl_recv_data_processed,
msg_total, msg_total, diff_seqno, diff_seqno);
} else
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "REPL INFO - Seqno : "INT8_FMT" "INT8_FMTX
" Pre filter total : "INT8_FMT" "INT8_FMTX" Post filter total : "
INT8_FMT" "INT8_FMTX" Msg Total : "INT8_FMT" "INT8_FMTX
" Current backlog : "INT8_FMT" "INT8_FMTX"\n",
log_seqno, log_seqno,
repl_recv_data_processed, repl_recv_data_processed,
repl_recv_postfltr_data_procd, repl_recv_postfltr_data_procd,
msg_total, msg_total, diff_seqno, diff_seqno);
}
/* Approximate time with an error not more than GTMRECV_HEARTBEAT_PERIOD. We use this
* instead of calling time(), and expensive system call, especially on VMS. The
* consequence of this choice is that we may defer logging when we may have logged. We
* can live with that. Currently, the logging interval is not changeable by users.
* When/if we provide means of choosing log interval, this code may have to be re-examined.
* - Vinaya 2003/09/08.
*/
assert(0 != gtmrecv_now);
repl_recv_this_log_time = gtmrecv_now;
assert(repl_recv_this_log_time >= repl_recv_prev_log_time);
time_elapsed = difftime(repl_recv_this_log_time, repl_recv_prev_log_time);
if ((double)GTMRECV_LOGSTATS_INTERVAL <= time_elapsed)
{
repl_log(gtmrecv_log_fp, TRUE, FALSE, "REPL INFO since last log : Time elapsed : "
"%00.f Tr recvd : "INT8_FMT" Tr bytes : "INT8_FMT" Msg bytes : "INT8_FMT"\n",
time_elapsed, trans_recvd_cnt - last_log_tr_recvd_cnt,
repl_recv_data_processed - repl_recv_lastlog_data_procd,
msg_total - repl_recv_lastlog_data_recvd);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "REPL INFO since last log : Time elapsed : "
"%00.f Tr recvd/s : %f Tr bytes/s : %f Msg bytes/s : %f\n", time_elapsed,
(float)(trans_recvd_cnt - last_log_tr_recvd_cnt)/time_elapsed,
(float)(repl_recv_data_processed - repl_recv_lastlog_data_procd)/time_elapsed,
(float)(msg_total - repl_recv_lastlog_data_recvd)/time_elapsed);
repl_recv_lastlog_data_procd = repl_recv_data_processed;
repl_recv_lastlog_data_recvd = msg_total;
last_log_tr_recvd_cnt = trans_recvd_cnt;
repl_recv_prev_log_time = repl_recv_this_log_time;
}
lastlog_seqno = log_seqno;
}
if (gtmrecv_logstats)
{
if (!filter_pass)
{
repl_log(gtmrecv_log_fp, FALSE, FALSE, "Tr : "INT8_FMT" Size : %d Write : %ld "
"Total : "INT8_FMT"\n", recvpool_ctl->jnl_seqno, write_len,
write_off, repl_recv_data_processed);
} else
{
assert(!is_new_histrec);
repl_log(gtmrecv_log_fp, FALSE, FALSE, "Tr : "INT8_FMT" Pre filter Size : %d "
"Post filter Size : %d Pre filter Write : %d Post filter Write : %ld "
"Pre filter Total : "INT8_FMT" Post filter Total : "INT8_FMT"\n",
recvpool_ctl->jnl_seqno, pre_filter_write_len, write_len,
pre_filter_write, write_off, repl_recv_data_processed,
repl_recv_postfltr_data_procd);
}
}
recvpool_ctl->write_wrap = write_wrap;
if (recvpool.gtmrecv_local->noresync)
{ /* With -NORESYNC, we could take the strm_seqno further back with multiple connects of the same receiver
* server with different source servers at different seqnos. So avoid any confusion with asserts
* that check for increasing seqnos using last_valid_histinfo and last_rcvd_histinfo.
*/
memset(&recvpool.recvpool_ctl->last_valid_histinfo, 0, SIZEOF(recvpool.recvpool_ctl->last_valid_histinfo));
memset(&recvpool.recvpool_ctl->last_rcvd_histinfo, 0, SIZEOF(recvpool.recvpool_ctl->last_rcvd_histinfo));
/* In the case of -NORESYNC, we are a root primary supplementary instance (not a propagating primary)
* and therefore recvpool.recvpool_ctl->last_valid_strm_histinfo[] and
* recvpool.recvpool_ctl->last_rcvd_strm_histinfo[] are unused and hence no need to reset them.
*/
assert(!remote_side->is_supplementary); /* assert that we are not a propagating primary supplementary */
assert(jnlpool->repl_inst_filehdr->is_supplementary && !jnlpool->jnlpool_ctl->upd_disabled);
}
if (is_new_histrec)
{
/* Note: The REPL_OLD_TRIPLE or REPL_HISTREC messages are endian converted by the source server
* in case the receiver is running with a this_side->jnl_ver higher than the source. If not, the call to
* function "repl_tr_endian_convert" (done already in the current function) will take care of
* the endian conversion. So no more endian conversion needed at this point.
*/
if (REPL_OLD_TRIPLE == msg_type)
{
assert(REPL_PROTO_VER_MULTISITE <= remote_side->proto_ver);
assert(REPL_PROTO_VER_SUPPLEMENTARY > remote_side->proto_ver);
assert(SIZEOF(old_triple_content) == write_len);
memcpy((sm_uc_ptr_t)&old_triple_content, (recvpool.recvdata_base + write_off),
SIZEOF(old_triple_content));
assert(JRT_TRIPLE == old_triple_content.jrec_type);
assert(old_triple_content.forwptr == SIZEOF(old_triple_content));
assert(old_triple_content.start_seqno == recvpool_ctl->jnl_seqno);
assert(old_triple_content.start_seqno >= recvpool.upd_proc_local->read_jnl_seqno);
assert((old_triple_content.start_seqno > recvpool_ctl->last_valid_histinfo.start_seqno)
|| ((old_triple_content.start_seqno == recvpool_ctl->last_valid_histinfo.start_seqno)
&& gtm_white_box_test_case_enabled
&& (WBTEST_UPD_PROCESS_ERROR == gtm_white_box_test_case_number)));
cur_histinfo = &tmp_histjrec.histcontent;
memcpy(cur_histinfo->root_primary_instname, old_triple_content.instname, MAX_INSTNAME_LEN - 1);
cur_histinfo->root_primary_instname[MAX_INSTNAME_LEN - 1] = '\0';
cur_histinfo->start_seqno = old_triple_content.start_seqno;
cur_histinfo->strm_seqno = 0;
cur_histinfo->root_primary_cycle = old_triple_content.cycle;
cur_histinfo->creator_pid = 0;
cur_histinfo->created_time = 0;
/* No need to initialize the following fields as they are reinitialized later
* when the history record gets added to the instance file (in "repl_inst_histinfo_add").
* cur_histinfo->histinfo_num = INVALID_HISTINFO_NUM;
* cur_histinfo->prev_histinfo_num = INVALID_HISTINFO_NUM;
* cur_histinfo->last_histinfo_num[] = INVALID_HISTINFO_NUM;
*/
cur_histinfo->strm_index = 0;
cur_histinfo->history_type = HISTINFO_TYPE_NORMAL;
NULL_INITIALIZE_REPL_INST_UUID(cur_histinfo->lms_group);
tmp_histjrec.jrec_type = JRT_HISTREC;
tmp_histjrec.forwptr = SIZEOF(repl_histrec_jnl_t);
/* We now have to upgrade a REPL_OLD_TRIPLE format history record to a REPL_HISTREC record.
* The latter is a much bigger record so make room for this in the receive pool.
*/
write_loc = write_off; /* reset "write_loc" */
write_len = tmp_histjrec.forwptr;
PREPARE_RECVPOOL_FOR_WRITE(write_len, SIZEOF(old_triple_content));
/* above macro could update "recvpool_ctl->write" and "write_loc" */
COPY_TO_RECVPOOL((uchar_ptr_t)&tmp_histjrec, write_len);/* uses and updates "write_loc" */
write_off = recvpool_ctl->write;
/* Copy relevant fields from received histinfo message to "last_rcvd_histinfo" if applicable */
cur_histinfo = &recvpool_ctl->last_rcvd_histinfo;
assert(old_triple_content.start_seqno >= cur_histinfo->start_seqno);
*cur_histinfo = tmp_histjrec.histcontent;
assert(!remote_side->is_supplementary); /* so no need to maintain last_rcvd_strm_histinfo[] */
} else
{
assert(REPL_HISTREC == msg_type);
assert(REPL_PROTO_VER_SUPPLEMENTARY <= remote_side->proto_ver);
assert(SIZEOF(repl_histrec_jnl_t) == write_len);
pool_histrec = (repl_histrec_jnl_t *)((sm_uc_ptr_t)recvpool.recvdata_base + write_off);
assert(JRT_HISTREC == pool_histrec->jrec_type);
assert(pool_histrec->forwptr == SIZEOF(repl_histrec_jnl_t));
pool_histinfo = &pool_histrec->histcontent;
assert(pool_histinfo->start_seqno == recvpool_ctl->jnl_seqno);
assert(pool_histinfo->start_seqno >= recvpool.upd_proc_local->read_jnl_seqno);
if (jnlpool->repl_inst_filehdr->is_supplementary && !jnlpool->jnlpool_ctl->upd_disabled)
{ /* Modify the history record to reflect the stream # */
assert((0 <= recvpool.gtmrecv_local->strm_index)
&& (MAX_SUPPL_STRMS > recvpool.gtmrecv_local->strm_index));
assert(strm_index == recvpool.gtmrecv_local->strm_index);
pool_histinfo->strm_index = recvpool.gtmrecv_local->strm_index;
assert(0 == pool_histinfo->strm_seqno);
assert(IS_REPL_INST_UUID_NON_NULL(recvpool.gtmrecv_local->remote_lms_group));
pool_histinfo->lms_group = recvpool.gtmrecv_local->remote_lms_group;
if (recvpool.gtmrecv_local->updateresync)
{
assert(FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd);
/* Make it known that this is an updateresync type history record */
pool_histinfo->history_type = HISTINFO_TYPE_UPDRESYNC;
}
/* We want to do a similar check for -noresync but we cannot use gtmrecv_local->noresync
* since that is reset the moment a REPL_WILL_RESTART_WITH_INFO message is seen (much
* before coming here). So use a static variable to determine if this is the first
* history record that is being received by this receiver server and if -noresync was
* specified in the command line (indicated by gtmrecv_options.noresync), set the
* history_type of the incoming history record to reflect the noresync type.
*/
assert(!recvpool.gtmrecv_local->noresync);
if (first_histrec)
{
if (gtmrecv_options.noresync)
pool_histinfo->history_type = HISTINFO_TYPE_NORESYNC;
first_histrec = FALSE;
}
}
assert((INVALID_SUPPL_STRM != strm_index) || (0 == pool_histinfo->strm_index));
cur_histinfo = &recvpool_ctl->last_rcvd_histinfo;
*cur_histinfo = *pool_histinfo;
if ((INVALID_SUPPL_STRM == strm_index) || (strm_index == pool_histinfo->strm_index))
{
assert((pool_histinfo->start_seqno > recvpool_ctl->last_valid_histinfo.start_seqno)
|| ((pool_histinfo->start_seqno == recvpool_ctl->last_valid_histinfo.start_seqno)
&& gtm_white_box_test_case_enabled
&& (WBTEST_UPD_PROCESS_ERROR == gtm_white_box_test_case_number)));
assert(pool_histinfo->start_seqno >= cur_histinfo->start_seqno);
} else
{
assert(!recvpool_ctl->insert_strm_histinfo);
cur_histinfo = pool_histinfo;
}
}
/* If supplementary instance with updates disabled, initialize last_rcvd_strm_histinfo as well */
if (remote_side->is_supplementary)
{ /* Check if "insert_strm_histinfo" is set. If so and if REPL_STRMINFO messages were
* exchanged between the source and receiver, we need to potentially insert history
* records for valid stream #s > 0. This is to ensure that even in case of a
* -updateresync startup, we record knowledge of all streams that exist on the source side
* (which we previously received as part of the same connection) on the receiver side as well.
* Examples of when REPL_STRMINFO messages are not exchanged are if receiver side is at
* jnl_seqno = 1, or if the receiver side is at the exact same jnl_seqno as the source side
* and no prior history is available.
*/
if (recvpool_ctl->insert_strm_histinfo
&& (max_strm_histinfo = recvpool_ctl->max_strm_histinfo)) /* caution: assignment */
{ /* Make room for the new records in the receive pool first.
* At this point "cur_histinfo" holds the first received history record.
* Add the other stream history records onto it.
*/
assert(0 == cur_histinfo->strm_index);
/* Determine how many history records need to be added */
num_strm_histinfo = 0;
/* Copy 0th stream history record into receive pool first */
rcvd_strm_histjrec[num_strm_histinfo].jrec_type = JRT_HISTREC;
rcvd_strm_histjrec[num_strm_histinfo].forwptr = SIZEOF(repl_histrec_jnl_t);
cur_histinfo->strm_seqno = jnlpool->jnlpool_ctl->strm_seqno[0];
rcvd_strm_histjrec[num_strm_histinfo++].histcontent = *cur_histinfo;
/* Copy non-zero stream history records if they exist */
for (idx = 1; idx < max_strm_histinfo; idx++)
{
if (recvpool_ctl->is_valid_strm_histinfo[idx])
{
DEBUG_ONLY(tmp_histinfo = recvpool_ctl->last_rcvd_strm_histinfo[idx]);
assert(IS_REPL_INST_UUID_NON_NULL(tmp_histinfo.lms_group));
rcvd_strm_histjrec[num_strm_histinfo].jrec_type = JRT_HISTREC;
rcvd_strm_histjrec[num_strm_histinfo].forwptr = SIZEOF(repl_histrec_jnl_t);
rcvd_strm_histjrec[num_strm_histinfo].histcontent
= recvpool_ctl->last_rcvd_strm_histinfo[idx];
/* Fix the "start_seqno" & "strm_seqno" fields of the history record */
rcvd_strm_histjrec[num_strm_histinfo].histcontent.start_seqno
= cur_histinfo->start_seqno;
/* Note that jnlpool_ctl->strm_seqno[idx] could be 0 if we have not yet
* seen any updates in this stream. But since we do have a history record
* for this stream and its strm_seqno cannot be 0 (has to be non-zero),
* set it to 1 in that case.
*/
rcvd_strm_histjrec[num_strm_histinfo].histcontent.strm_seqno
= jnlpool->jnlpool_ctl->strm_seqno[idx]
? jnlpool->jnlpool_ctl->strm_seqno[idx] : 1;
num_strm_histinfo++;
/* Do not reset recvpool_ctl->is_valid_strm_histinfo[idx] to FALSE
* as this reflects reality and staying this way helps later
* communication with the source in cases of reconnects/pipe-drains.
*/
}
}
write_loc = write_off; /* reset "write_loc" */
write_len = num_strm_histinfo * SIZEOF(repl_histrec_jnl_t);
PREPARE_RECVPOOL_FOR_WRITE(write_len, 0);
/* could update "recvpool_ctl->write" * and "write_loc" */
COPY_TO_RECVPOOL((uchar_ptr_t)rcvd_strm_histjrec, write_len);
/* uses and updates "write_loc" */
write_off = recvpool_ctl->write;
/* Note: "last_rcvd_strm_histinfo" is automatically maintained in this case */
/* Now that we have inserted the stream specific history records, reset flag */
recvpool_ctl->insert_strm_histinfo = FALSE;
} else
{ /* Maintain last_rcvd_strm_histinfo[] as well.
* In addition, if insert_strm_histinfo is TRUE, it means no REPL_STRMINFO messages
* were exchanged at startup. In that case, no need of inserting stream specific
* history information in the receive pool so reset variable to FALSE.
*/
if (recvpool_ctl->insert_strm_histinfo)
recvpool_ctl->insert_strm_histinfo = FALSE;
histinfo_strm_num = cur_histinfo->strm_index;
assert((0 <= histinfo_strm_num) && (MAX_SUPPL_STRMS > histinfo_strm_num));
memcpy(&recvpool_ctl->last_rcvd_strm_histinfo[histinfo_strm_num],
cur_histinfo, SIZEOF(repl_histinfo));
recvpool_ctl->is_valid_strm_histinfo[histinfo_strm_num] = TRUE;
if (recvpool_ctl->max_strm_histinfo <= histinfo_strm_num)
recvpool_ctl->max_strm_histinfo = histinfo_strm_num + 1;
}
}
repl_dump_histinfo(gtmrecv_log_fp, TRUE, TRUE, "New History Content", cur_histinfo);
} else
{ /* Note: In case of a propagating primary supplementary instance, the below if implies that we will
* change last_rcvd_histinfo to last_valid_histinfo even if the logical update corresponded to a
* different stream than 0 (the stream corresponding to the history of interest on this receiver).
* But there should not be any issues due to this as it is an update nevertheless and is going to
* bump up the recvpool_ctl->jnl_seqno to one more than last_rcvd_histinfo.start_seqno.
*/
if (recvpool_ctl->jnl_seqno == recvpool_ctl->last_rcvd_histinfo.start_seqno)
{ /* Move over stuff from "last_rcvd_histinfo" to "last_valid_histinfo" */
memcpy(&recvpool_ctl->last_valid_histinfo,
&recvpool_ctl->last_rcvd_histinfo, SIZEOF(repl_histinfo));
if (remote_side->is_supplementary)
{ /* Propagating primary supplementary instance. Maintain last_valid_strm_histinfo too. */
max_strm_histinfo = recvpool_ctl->max_strm_histinfo;
assert(max_strm_histinfo);
for (idx = 0; idx < max_strm_histinfo; idx++)
{
if (recvpool_ctl->is_valid_strm_histinfo[idx])
{
DEBUG_ONLY(tmp_histinfo = recvpool_ctl->last_rcvd_strm_histinfo[idx]);
assert((0 == idx) || IS_REPL_INST_UUID_NON_NULL(tmp_histinfo.lms_group));
assert((0 != idx) || IS_REPL_INST_UUID_NULL(tmp_histinfo.lms_group));
memcpy(&recvpool_ctl->last_valid_strm_histinfo[idx],
&recvpool_ctl->last_rcvd_strm_histinfo[idx], SIZEOF(repl_histinfo));
recvpool_ctl->is_valid_strm_histinfo[idx] = FALSE;
}
}
recvpool_ctl->max_strm_histinfo = 0;
/* Now that last_valid_strm_histinfo[0] is initialized, it is safe to reset the below */
if (recvpool_ctl->insert_strm_histinfo)
recvpool_ctl->insert_strm_histinfo = FALSE;
}
/* Now that at least one history record has been written into the receive pool and is guaranteed
* to be written to the instance file (when this gets processed by the update process), don't use
* -updateresync or -noresync for future handshakes in case the current connection gets reset.
*/
if (recvpool.gtmrecv_local->updateresync)
{
recvpool.gtmrecv_local->updateresync = FALSE;
/* Close fd of the input instance file name used for the -updateresync */
if (FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd)
CLOSEFILE_RESET(recvpool.gtmrecv_local->updresync_instfile_fd, rc);
}
assert(!recvpool.gtmrecv_local->noresync); /* should have been reset already */
}
QWINCRBYDW(recvpool_ctl->jnl_seqno, 1);
assert(recvpool_ctl->last_valid_histinfo.start_seqno < recvpool_ctl->jnl_seqno);
}
/* The update process looks at "recvpool_ctl->write" first and then reads (a) "recvpool_ctl->write_wrap"
* AND (b) all journal data in the receive pool upto this offset. It assumes that (a) and (b) will never
* hold stale values corresponding to a previous state of "recvpool_ctl->write". In order for this
* assumption to hold good, the receiver server needs to do a write memory barrier after updating the
* receive pool data and "write_wrap" but before updating "write". The update process will do a read
* memory barrier after reading "write" but before reading "write_wrap" or the receive pool data. Not
* enforcing the read order will result in the update process attempting to read/process invalid data
* from the receive pool (which could end up in db out of sync situation between primary and secondary).
*/
SHM_WRITE_MEMORY_BARRIER;
recvpool_ctl->write = write_loc;
/* Signal the update process to check for the update. */
PTHREAD_COND_SIGNAL(&recvpool_ctl->write_updated, status);
if (0 != status)
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(8) ERR_SYSCALL, 5,
LEN_AND_LIT("pthread_cond_signal"), CALLFROM, status, 0);
} while (is_repl_cmpc);
return;
}
/* Retrieve the history record corresponding to "input_seqno" in the instance file used with the -UPDATERESYNC qualifier.
* Do special processing in case of P->Q type of connection (where primary and secondary are both supplementary instances)
* which is to search only in the 0th stream.
*
* Note: This function is similar to "repl_inst_histinfo_find_seqno" except that this operates on the input instance file
* provided with the -updateresync qualifier. Two reasons why we need this code duplication is
* a) The "repl_inst_histinfo_find_seqno" function currently is coded to use only the replication instance file.
* b) The input instance file for -updateresync could be cross endian.
* If "repl_inst_histinfo_find_seqno" is enhanced to fix these limitations, then we can avoid this code duplication.
*/
STATICFNDEF void gtmrecv_updresync_histinfo_find_seqno(seq_num input_seqno, int4 strm_num, repl_histinfo *histinfo)
{
char print_msg[1024];
int fd, status;
int4 histinfo_num;
off_t offset;
fd = recvpool.gtmrecv_local->updresync_instfile_fd;
assert(FD_INVALID != fd);
/* If remote side is non-supplementary then its instance file (which is given as input to the -updateresync
* command) knows only strem 0, so reset strm_num to 0 unconditionally.
*/
if (!remote_side->is_supplementary)
strm_num = 0;
if (INVALID_SUPPL_STRM == strm_num)
histinfo_num = recvpool.gtmrecv_local->updresync_num_histinfo;
else
histinfo_num = recvpool.gtmrecv_local->updresync_num_histinfo_strm[strm_num];
if (INVALID_HISTINFO_NUM == histinfo_num)
{ /* The instance file cannot be used for updateresync if it has NO history records. */
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(9) ERR_UPDSYNCINSTFILE, 0, ERR_STRMNUMIS, 1, strm_num, ERR_TEXT, 2,
LEN_AND_LIT("Input instance file has NO history records"));
}
assert(0 <= histinfo_num);
if (!recvpool.gtmrecv_local->updresync_jnl_seqno)
{ /* The instance file cannot be used for updateresync if it has a ZERO seqno */
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(9) ERR_UPDSYNCINSTFILE, 0, ERR_STRMNUMIS, 1, strm_num,
ERR_TEXT, 2, LEN_AND_LIT("Input instance file has jnl_seqno of 0"));
}
if (input_seqno > recvpool.gtmrecv_local->updresync_jnl_seqno)
{ /* Input seqno is greater than the max seqno in the updateresync input instance file. So can never be found. */
SNPRINTF(print_msg, SIZEOF(print_msg), "Seqno "INT8_FMT" "INT8_FMTX" cannot be found in input instance file "
" which has a max seqno of "INT8_FMT" "INT8_FMTX"\n", input_seqno, input_seqno,
recvpool.gtmrecv_local->updresync_jnl_seqno, recvpool.gtmrecv_local->updresync_jnl_seqno);
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(9) ERR_UPDSYNCINSTFILE, 0,
ERR_STRMNUMIS, 1, strm_num, ERR_TEXT, 2, LEN_AND_STR(print_msg));
}
histinfo->start_seqno = 0;
do
{
offset = REPL_INST_HISTINFO_START + ((histinfo_num) * SIZEOF(repl_histinfo));
LSEEKREAD(fd, offset, histinfo, SIZEOF(repl_histinfo), status);
if (0 != status)
{ /* At this point, we don't have the name of the input instance file used in the -updateresync qualifier.
* So we use a value of "" instead. The fact that the REPLINSTREAD message is preceded by a UPDSYNCINSTFILE
* error indicates to the user it is the -updateresync qualifier where the issue is so it is not a big loss.
*/
if (-1 == status)
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(15) ERR_UPDSYNCINSTFILE, 0,
ERR_STRMNUMIS, 1, strm_num,
ERR_TEXT, 2, LEN_AND_LIT("Error reading history record"),
ERR_REPLINSTREAD, 4, SIZEOF(repl_histinfo), (qw_off_t *)&offset, LEN_AND_LIT(""));
else
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(16) ERR_UPDSYNCINSTFILE, 0,
ERR_STRMNUMIS, 1, strm_num,
ERR_TEXT, 2, LEN_AND_LIT("Error reading history record"),
ERR_REPLINSTREAD, 4, SIZEOF(repl_histinfo), (qw_off_t *)&offset, LEN_AND_LIT(""), status);
}
if (recvpool.gtmrecv_local->updresync_cross_endian)
ENDIAN_CONVERT_REPL_HISTINFO(histinfo);
if (input_seqno > histinfo->start_seqno)
return; /* found history record corresponding to input_seqno. return right away */
histinfo_num = (INVALID_SUPPL_STRM == strm_num) ? (histinfo_num - 1) : histinfo->prev_histinfo_num;
} while (INVALID_HISTINFO_NUM != histinfo_num);
/* Could not find history record in -updateresync= input instance file */
SNPRINTF(print_msg, SIZEOF(print_msg),
"Receiver side instance seqno "INT8_FMT" "INT8_FMTX" is less than"
" any history record found in instance file", input_seqno, input_seqno);
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(9) ERR_UPDSYNCINSTFILE, 0,
ERR_STRMNUMIS, 1, strm_num, ERR_TEXT, 2, LEN_AND_STR(print_msg));
}
/* Retrieve the "index"'th history record in the instance file specified using the -UPDATERESYNC qualifier.
*
* This function is similar to "repl_inst_histinfo_get" except that this operates on the input instance file provided
* with the -updateresync qualifier. Two reasons why we need this code duplication is
* a) The "repl_inst_histinfo_get" function currently is coded to use only the replication instance file.
* b) The input instance file for -updateresync could be cross endian.
* If "repl_inst_histinfo_get" is enhanced to fix these limitations, then we can avoid this code duplication.
*/
STATICFNDEF void gtmrecv_updresync_histinfo_get(int4 index, repl_histinfo *histinfo)
{
int fd, status;
off_t offset;
fd = recvpool.gtmrecv_local->updresync_instfile_fd;
assert(FD_INVALID != fd);
assert(0 <= index);
offset = REPL_INST_HISTINFO_START + ((index) * SIZEOF(repl_histinfo));
LSEEKREAD(fd, offset, histinfo, SIZEOF(repl_histinfo), status);
if (0 != status)
{ /* At this point, we don't have the name of the input instance file used in the -updateresync qualifier.
* So we use a value of "" instead. The fact that the REPLINSTREAD message is preceded by a UPDSYNCINSTFILE
* error indicates to the user it is the -updateresync qualifier where the issue is so it is not a big loss.
*/
if (-1 == status)
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(12) ERR_UPDSYNCINSTFILE, 0,
ERR_TEXT, 2, LEN_AND_LIT("Error reading history record"),
ERR_REPLINSTREAD, 4, SIZEOF(repl_histinfo), (qw_off_t *)&offset, LEN_AND_LIT(""));
else
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(13) ERR_UPDSYNCINSTFILE, 0,
ERR_TEXT, 2, LEN_AND_LIT("Error reading history record"),
ERR_REPLINSTREAD, 4, SIZEOF(repl_histinfo), (qw_off_t *)&offset, LEN_AND_LIT(""), status);
}
if (recvpool.gtmrecv_local->updresync_cross_endian)
ENDIAN_CONVERT_REPL_HISTINFO(histinfo);
}
/* This function is invoked on receipt of a REPL_NEED_STRMINFO message. In case of an error, it returns if either
* repl_connection_reset OR gtmrecv_wait_for_jnl_seqno are set so the caller should check for this and return as well.
*/
STATICFNDEF void gtmrecv_process_need_strminfo_msg(repl_needstrminfo_msg_ptr_t need_strminfo_msg)
{
int4 status;
int idx;
repl_histinfo histinfo, *previously_rcvd_histinfo;
repl_strminfo_msg_t strminfo_msg;
seq_num need_strminfo_seqno, last_valid_histinfo_seqno;
assert(remote_side->is_supplementary); /* STRMINFO messages are sent only in case source and receiver are supplementary */
assert(0 == strm_index);
assert(remote_side->endianness_known); /* ensure remote_side->cross_endian is reliable */
if (!remote_side->cross_endian)
need_strminfo_seqno = need_strminfo_msg->seqno;
else
need_strminfo_seqno = GTM_BYTESWAP_64(need_strminfo_msg->seqno);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_NEED_STRMINFO message for seqno "INT8_FMT" "INT8_FMTX"\n",
need_strminfo_seqno, need_strminfo_seqno);
if (recvpool.gtmrecv_local->updateresync && (FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd))
{ /* The stream information that is being requested needs to be found in the -updateresync input instance file
* (not the receiver side instance file).
*/
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Searching for desired stream info in -updateresync input instance file\n");
gtmrecv_updresync_histinfo_find_seqno(need_strminfo_seqno, INVALID_SUPPL_STRM, &histinfo);
for (idx = 0; idx < MAX_SUPPL_STRMS; idx++)
strminfo_msg.last_histinfo_num[idx] = histinfo.last_histinfo_num[idx];
assert(recvpool.recvpool_ctl->insert_strm_histinfo);
} else
{ /* The history record needs to be found in the receiver side instance file or in the receive pool.
* If last_valid_strm_histinfo[0] has non-default content, then because this is a supplementary instance
* and a propagating primary, we can rest assured that last_valid_strm_histinfo[1] thru [15] reflect
* the latest history records for each stream if the stream exists on this instance. And so no need
* to go to the instance file at all. If [0] does not have any content, then it means the cached history
* is empty for not just the 0th stream but for every other stream as well i.e. the receive pool is empty
* and the receiver is connecting with a source for the first time. So go to the instance file in that case.
*/
last_valid_histinfo_seqno = recvpool.recvpool_ctl->last_valid_strm_histinfo[0].start_seqno;
if (last_valid_histinfo_seqno)
{
assert(need_strminfo_seqno > last_valid_histinfo_seqno);
assert(!recvpool.gtmrecv_local->updateresync);
assert(!recvpool.gtmrecv_local->noresync);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Searching for the desired history in the receive pool\n");
for (idx = 0; idx < MAX_SUPPL_STRMS; idx++)
{
previously_rcvd_histinfo = &recvpool.recvpool_ctl->last_valid_strm_histinfo[idx];
assert((0 != idx) || IS_REPL_INST_UUID_NULL(previously_rcvd_histinfo->lms_group));
if ((0 == idx) || IS_REPL_INST_UUID_NON_NULL(previously_rcvd_histinfo->lms_group))
strminfo_msg.last_histinfo_num[idx] = UNKNOWN_HISTINFO_NUM;
else
{
assert(0 != idx);
strminfo_msg.last_histinfo_num[idx] = INVALID_HISTINFO_NUM;
}
}
assert(!recvpool.recvpool_ctl->insert_strm_histinfo);
histinfo.strm_index = 0; /* so "0 < histinfo.strm_index" if block is skipped below for this case */
} else
{
repl_log(gtmrecv_log_fp, TRUE, TRUE,
"Searching for the desired history in the replication instance file\n");
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
/* above macro will "return" if repl_connection_reset OR gtmrecv_wait_for_jnl_seqno is set */
status = repl_inst_wrapper_histinfo_find_seqno(need_strminfo_seqno, INVALID_SUPPL_STRM, &histinfo);
rel_lock(jnlpool->jnlpool_dummy_reg);
if (0 != status)
{ /* Close the connection */
assert(ERR_REPLINSTNOHIST == status);
assert(FALSE);
gtmrecv_autoshutdown(); /* should not return */
}
for (idx = 0; idx < MAX_SUPPL_STRMS; idx++)
strminfo_msg.last_histinfo_num[idx] = histinfo.last_histinfo_num[idx];
}
}
if (0 < histinfo.strm_index)
{
assert(histinfo.last_histinfo_num[histinfo.strm_index] < histinfo.histinfo_num);
strminfo_msg.last_histinfo_num[histinfo.strm_index] = histinfo.histinfo_num;
}
if (remote_side->cross_endian)
{
for (idx = 0; idx < MAX_SUPPL_STRMS; idx++)
{
assert(4 == SIZEOF(strminfo_msg.last_histinfo_num[idx])); /* so GTM_BYTESWAP_32 can be used below */
strminfo_msg.last_histinfo_num[idx] = GTM_BYTESWAP_32(strminfo_msg.last_histinfo_num[idx]);
}
}
gtmrecv_repl_send((repl_msg_ptr_t)&strminfo_msg, REPL_STRMINFO, SIZEOF(repl_strminfo_msg_t), "REPL_STRMINFO", MAX_SEQNO);
return;
}
/* This function is invoked on receipt of a REPL_NEED_HISTINFO message. In case of an error, it returns if either
* repl_connection_reset OR gtmrecv_wait_for_jnl_seqno are set so the caller should check for this and return as well.
*/
STATICFNDEF void gtmrecv_process_need_histinfo_msg(repl_needhistinfo_msg_ptr_t need_histinfo_msg, repl_histinfo *histinfo)
{
boolean_t maintain_rcvd_strm_histinfo, suppl_propagate_primary;
int4 need_histinfo_num, need_histinfo_strm_num, status;
recvpool_ctl_ptr_t recvpool_ctl;
seq_num first_unprocessed_seqno, last_unprocessed_histinfo_seqno;
seq_num need_histinfo_seqno, last_valid_histinfo_seqno;
assert(remote_side->endianness_known); /* ensure remote_side->cross_endian is reliable */
if (!remote_side->cross_endian)
{
need_histinfo_seqno = need_histinfo_msg->seqno;
need_histinfo_num = need_histinfo_msg->histinfo_num;
need_histinfo_strm_num = need_histinfo_msg->strm_num;
} else
{
need_histinfo_seqno = GTM_BYTESWAP_64(need_histinfo_msg->seqno);
need_histinfo_num = GTM_BYTESWAP_32(need_histinfo_msg->histinfo_num);
need_histinfo_strm_num = GTM_BYTESWAP_32(need_histinfo_msg->strm_num);
}
assert((INVALID_SUPPL_STRM == strm_index) || ((0 <= strm_index) && (MAX_SUPPL_STRMS > strm_index)));
if (INVALID_SUPPL_STRM == strm_index)
{ /* Both receiver and source sides are non-supplementary instances */
if (REPL_PROTO_VER_SUPPLEMENTARY > remote_side->proto_ver)
{ /* needhistinfo_msg.strm_num & histinfo_num are uninitialized in this case. Fix it */
need_histinfo_strm_num = INVALID_SUPPL_STRM;
need_histinfo_num = INVALID_HISTINFO_NUM;
} else
{
assert(INVALID_SUPPL_STRM == need_histinfo_strm_num);
assert(INVALID_HISTINFO_NUM == need_histinfo_num);
}
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_NEED_HISTINFO message for seqno "INT8_FMT" "INT8_FMTX"\n",
need_histinfo_seqno, need_histinfo_seqno);
} else if (0 < strm_index)
{ /* Receiver side is supplementary but Source side is a non-supplementary instance */
assert(INVALID_SUPPL_STRM == need_histinfo_strm_num);
assert(INVALID_HISTINFO_NUM == need_histinfo_num);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_NEED_HISTINFO message for Stream %d : Seqno "
INT8_FMT" "INT8_FMTX"\n", strm_index, need_histinfo_seqno, need_histinfo_seqno);
need_histinfo_strm_num = strm_index;
} else
{ /* Both receiver and source sides are supplementary instances */
/* strm_index is 0 at this point (already asserted above) */
assert(INVALID_SUPPL_STRM != need_histinfo_strm_num);
if ((INVALID_HISTINFO_NUM == need_histinfo_num) || (UNKNOWN_HISTINFO_NUM == need_histinfo_num))
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_NEED_HISTINFO message for Stream %d : "
"Seqno "INT8_FMT" "INT8_FMTX"\n", need_histinfo_strm_num, need_histinfo_seqno, need_histinfo_seqno);
else
{
assert(0 <= need_histinfo_num);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_NEED_HISTINFO message for History Number %d\n",
need_histinfo_num);
}
}
/* The only two histinfo_num values that have special meaning are negative. So we can check for a valid value
* by checking for positive. Assert that below before doing the positive check.
*/
assert((0 > INVALID_HISTINFO_NUM) && (0 > UNKNOWN_HISTINFO_NUM));
recvpool_ctl = recvpool.recvpool_ctl;
if (0 <= need_histinfo_num)
{ /* Handle simplest case first. Get the "need_histinfo_num"'th history record directly from the instance file */
if (recvpool.gtmrecv_local->updateresync && (FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd))
{ /* The history record that is being requested needs to be found in the -updateresync input instance file
* (not the receiver side instance file).
*/
repl_log(gtmrecv_log_fp, TRUE, TRUE,
"Searching for desired history in the -updateresync input instance file\n");
gtmrecv_updresync_histinfo_get(need_histinfo_num, histinfo);
} else
{ /* The history record needs to be found in the receiver side instance file */
repl_log(gtmrecv_log_fp, TRUE, TRUE,
"Searching for the desired history in the replication instance file\n");
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
/* above macro will "return" if repl_connection_reset OR gtmrecv_wait_for_jnl_seqno is set */
status = repl_inst_histinfo_get(need_histinfo_num, histinfo);
rel_lock(jnlpool->jnlpool_dummy_reg);
if (0 != status)
{ /* Close the connection */
assert(ERR_REPLINSTNOHIST == status);
assert(FALSE);
gtmrecv_autoshutdown(); /* should not return */
}
}
maintain_rcvd_strm_histinfo = TRUE;
} else if (UNKNOWN_HISTINFO_NUM == need_histinfo_num)
{ /* This value was sent in a previous REPL_STRMINFO message for a particular non-zero stream # because
* we had not yet played this history record in the instance file. Return history record directly from
* where the receiver server saved a copy of the last unprocessed history record for this stream number.
*/
assert(need_histinfo_strm_num);
assert(remote_side->is_supplementary);
assert((0 < need_histinfo_strm_num) && (MAX_SUPPL_STRMS > need_histinfo_strm_num));
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Searching for desired history in the receive pool\n");
*histinfo = recvpool_ctl->last_valid_strm_histinfo[need_histinfo_strm_num];
maintain_rcvd_strm_histinfo = FALSE;
} else
{
first_unprocessed_seqno = recvpool.upd_proc_local->read_jnl_seqno;
repl_log(gtmrecv_log_fp, TRUE, FALSE, "Update process has processed upto seqno "INT8_FMT" "INT8_FMTX"\n",
first_unprocessed_seqno, first_unprocessed_seqno);
suppl_propagate_primary = remote_side->is_supplementary;
if (!suppl_propagate_primary)
last_valid_histinfo_seqno = recvpool_ctl->last_valid_histinfo.start_seqno;
else
last_valid_histinfo_seqno = recvpool_ctl->last_valid_strm_histinfo[0].start_seqno;
repl_log(gtmrecv_log_fp, TRUE, TRUE,
"Starting seqno of the last valid history in the receive pool is "INT8_FMT" "INT8_FMTX"\n",
last_valid_histinfo_seqno, last_valid_histinfo_seqno);
if (last_valid_histinfo_seqno >= first_unprocessed_seqno)
last_unprocessed_histinfo_seqno = last_valid_histinfo_seqno;
else
last_unprocessed_histinfo_seqno = MAX_SEQNO;
if (last_unprocessed_histinfo_seqno && (need_histinfo_seqno > last_unprocessed_histinfo_seqno))
{
/* NOTE0: The source server is requesting histinfo information for a seqno whose corresponding
* histinfo has also not yet been processed by the update process (and hence not present in the
* instance file). Find latest histinfo information that is stored in receive pool.
* NOTE1: Even though there could be more than one unprocessed history record in the receive pool,
* the source should request only the last one for comparison. If the last history record on the
* receiver matches on the source side too, replication can resume from there. If not, the receiver
* side should do a rollback (REPL_ROLLBACK_FIRST message). That is why it is enough to maintain
* recvpool_ctl->last_valid_histinfo and not pointers to all the unprocessed histories.
* NOTE2: There is one exception to this and that is if the receiver server had already connected
* to a source server and placed records in the receiver pool and then lost the connection and
* reestablished connection (with the same or a different source server) AND had been started with
* the -noresync option. In this case, it will not see a REPL_ROLLBACK_FIRST message but instead
* will go back in history to find the first matching history. But in this case, the -noresync is
* valid only for the FIRST connection and is cleared for future connections. So reconnections would
* assume as if -noresync was not specified and hence will fall into the same REPL_ROLLBACK_FIRST
* category as described in NOTE0. Assert it below.
* NOTE3: In the case of a propagating primary supplementary instance, we need to not just store the
* last received history record but also one for each stream possible.
*/
assert(!recvpool.gtmrecv_local->updateresync);
assert(!recvpool.gtmrecv_local->noresync);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Searching for the desired history in the receive pool\n");
if (!suppl_propagate_primary)
memcpy(histinfo, &recvpool_ctl->last_valid_histinfo, SIZEOF(repl_histinfo));
else
{
assert(!need_histinfo_strm_num);
memcpy(histinfo, &recvpool_ctl->last_valid_strm_histinfo[0], SIZEOF(repl_histinfo));
}
assert(!recvpool_ctl->insert_strm_histinfo);
} else if (recvpool.gtmrecv_local->updateresync && (FD_INVALID != recvpool.gtmrecv_local->updresync_instfile_fd))
{ /* The receiver was started with -UPDATERESYNC=<INSTFILENAME>. The history record that is being requested
* needs to be found in the input instance file (not the receiver side instance file).
* Read the history record corresponding to need_histinfo_seqno.
*/
repl_log(gtmrecv_log_fp, TRUE, TRUE,
"Searching for desired history in the -updateresync input instance file\n");
gtmrecv_updresync_histinfo_find_seqno(need_histinfo_seqno, need_histinfo_strm_num, histinfo);
assert(histinfo->start_seqno);
} else
{ /* The seqno has been processed by the update process. Hence the histinfo
* for this will be found in the instance file. Search there.
*/
assert(NULL != jnlpool->jnlpool_dummy_reg);
repl_log(gtmrecv_log_fp, TRUE, TRUE,
"Searching for the desired history in the replication instance file\n");
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
status = repl_inst_wrapper_histinfo_find_seqno(need_histinfo_seqno, need_histinfo_strm_num, histinfo);
rel_lock(jnlpool->jnlpool_dummy_reg);
if (0 != status)
{ /* Close the connection */
assert(ERR_REPLINSTNOHIST == status);
gtmrecv_autoshutdown(); /* should not return */
}
assert((histinfo->histinfo_num != (jnlpool->repl_inst_filehdr->num_histinfo - 1))
|| (histinfo->start_seqno == jnlpool->jnlpool_ctl->last_histinfo_seqno));
if (0 < need_histinfo_strm_num)
{ /* About to send to a non-supplementary instance. It does not understand strm_seqnos.
* So convert it back to a format it understands.
*/
CONVERT_SUPPL2NONSUPPL_HISTINFO(*histinfo);
}
assert(histinfo->start_seqno);
assert(histinfo->start_seqno < need_histinfo_seqno);
}
maintain_rcvd_strm_histinfo = (0 == histinfo->strm_index);
}
if (maintain_rcvd_strm_histinfo && recvpool_ctl->insert_strm_histinfo)
{
assert(remote_side->is_supplementary);
need_histinfo_strm_num = histinfo->strm_index;
assert((0 <= need_histinfo_strm_num) && (MAX_SUPPL_STRMS > need_histinfo_strm_num));
assert(IS_REPL_INST_UUID_NULL(recvpool_ctl->last_rcvd_strm_histinfo[need_histinfo_strm_num].lms_group));
assert((FALSE == recvpool_ctl->is_valid_strm_histinfo[need_histinfo_strm_num])
|| !memcmp(&recvpool_ctl->last_rcvd_strm_histinfo[need_histinfo_strm_num],
histinfo, SIZEOF(repl_histinfo)));
memcpy(&recvpool_ctl->last_rcvd_strm_histinfo[need_histinfo_strm_num], histinfo, SIZEOF(repl_histinfo));
recvpool_ctl->is_valid_strm_histinfo[need_histinfo_strm_num] = TRUE;
if (recvpool_ctl->max_strm_histinfo <= need_histinfo_strm_num)
recvpool_ctl->max_strm_histinfo = need_histinfo_strm_num + 1;
}
return;
}
#ifdef GTM_TLS
/* The below logic is very similar to `gtmsource_exchange_tls_info' but is kept separate because of the following reasons:
* (a) `gtmrecv_poll_actions' needs to be invoked as opposed to `gtmsource_poll_actions'.
* (b) The arguments passed to `gtmrecv_poll_actions' are defined as static.
* (c) The action taken after `gtmsource_poll_actions' and `gtmrecv_poll_actions' are different.
*
* The Receiver and Source servers exchanged TLS plugin library and 3rd party TLS library version information as part of
* negotiating the startup of TLS. In the event of a handshake error, print that information as it may contribute to the
* reason for the TLS handshake error. NOTE: Since gtmsource_exchange_tls_info() initiates the TLS negotiation, it has
* the client TLS information. The Receiver side calls this function after it receives the same details and so those bits
* of information must be passed to this function.
*/
STATICFNDEF boolean_t gtmrecv_exchange_tls_info(uint4 remote_API_ver, uint4 remote_lib_ver)
{
int errlen, poll_dir, status;
static int flags = GTMTLS_OP_FORCE_VERIFY_PEER | GTMTLS_OP_RENEGOTIATE_REQUESTED;
char *errp;
boolean_t pha_fallback;
repl_tlsinfo_msg_t reply;
reply.type = REPL_TLS_INFO;
reply.len = MIN_REPL_MSGLEN;
reply.API_version = GTM_TLS_API_VERSION;
reply.library_version = (uint4)tls_ctx->runtime_version;
if (remote_side->cross_endian)
{
reply.API_version = GTM_BYTESWAP_32(reply.API_version);
reply.library_version = GTM_BYTESWAP_32(reply.library_version);
}
gtmrecv_repl_send((repl_msg_ptr_t)&reply, REPL_TLS_INFO, SIZEOF(repl_tlsinfo_msg_t), "REPL_TLS_INFO", MAX_SEQNO);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return FALSE;
/* At this point, the both sides are ready for a TLS/SSL handshake. Create a TLS/SSL aware socket. */
if (NULL == (repl_tls.sock = gtm_tls_socket(tls_ctx, repl_tls.sock, gtmrecv_sock_fd, repl_tls.id, flags)))
{
if (PLAINTEXT_FALLBACK)
gtm_putmsg_csa(CSA_ARG(NULL) VARLSTCNT(6) MAKE_MSG_WARNING(ERR_TLSCONVSOCK), 0,
ERR_TEXT, 2, LEN_AND_STR(gtm_tls_get_error(NULL)));
else
{ /* Close the connection and issue an error message */
repl_close(>mrecv_sock_fd);
repl_connection_reset = TRUE;
gtm_putmsg_csa(CSA_ARG(NULL) VARLSTCNT(6) ERR_TLSCONVSOCK, 0,
ERR_TEXT, 2, LEN_AND_STR(gtm_tls_get_error(NULL)));
return FALSE;
}
} else
{
/* Do the actual handshake. */
poll_dir = REPL_INVALID_POLL_DIRECTION;
do
{
status = repl_do_tls_handshake(gtmrecv_log_fp, gtmrecv_sock_fd, TRUE, &poll_dir);
assert(0 == data_len);
gtmrecv_poll_actions(data_len, buff_unprocessed, buffp);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return FALSE;
} while ((GTMTLS_WANT_READ == status) || (GTMTLS_WANT_WRITE == status));
if (SS_NORMAL == status) /* Where enabled, do post handshake auth */
status = repl_do_tls_post_handshake(gtmrecv_log_fp, gtmrecv_sock_fd);
if (SS_NORMAL == status)
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Secure communication enabled using TLS/SSL protocol\n");
if (GTMTLS_OP_PHA_EXT_NOT_RECEIVED & flags) /* Re-enable PHA for subsequent connections */
flags &= ~GTMTLS_OP_PHA_EXT_NOT_RECEIVED;
return TRUE;
} else if (REPL_CONN_RESET(status))
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Attempt to connect() with TLS/SSL protocol failed. "
"Status = %d; %s\n", status, STRERROR(status));
repl_close(>mrecv_sock_fd);
repl_connection_reset = TRUE;
return FALSE;
}
errp = (0 > status) ? (char *)gtm_tls_get_error(repl_tls.sock) : STRERROR(status);
if (OUT_BUFF_SIZE > (errlen = strlen(errp))) /* Append version information in one message */
snprintf(errp + errlen, OUT_BUFF_SIZE - errlen, "; Local API:0x%08x, LIB:0x%08x; Remote API:0x%08x, LIB:0x%08x",
GTM_TLS_API_VERSION, (uint4)tls_ctx->runtime_version, remote_API_ver, remote_lib_ver);
pha_fallback = ((GTMTLS_OP_PHA_EXT_FALLBACK & repl_tls.sock->flags) && (-2 == status)
&& !(GTMTLS_OP_PHA_EXT_NOT_RECEIVED & flags));
gtm_putmsg_csa(CSA_ARG(NULL) VARLSTCNT(6) /* Plaintext/PHA fallback issue a WARNING */
((PLAINTEXT_FALLBACK || pha_fallback) ? MAKE_MSG_WARNING(ERR_TLSHANDSHAKE) : ERR_TLSHANDSHAKE),
0, ERR_TEXT, 2, LEN_AND_STR(errp));
/* Handshake error. If PHA related, disable PHA and issue an error. Otherwise, follow plaintext fallback behavior */
if ((!PLAINTEXT_FALLBACK) || ((GTMTLS_OP_PHA_EXT_FALLBACK & repl_tls.sock->flags) && (-2 == status)))
{ /* Close the connection and issue an error message */
if (pha_fallback) /* PHA fallback enabled and extension not received; disable PHA */
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Post-handshake fallback enabled. Disconnecting. "
"Post Handshake Authentication temporarily disabled for the next TLS connection "
"attempt.\n");
flags |= GTMTLS_OP_PHA_EXT_NOT_RECEIVED;
}
repl_close(>mrecv_sock_fd);
repl_connection_reset = TRUE;
return FALSE;
}
}
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Plaintext fallback enabled. Closing and reconnecting without TLS/SSL.\n");
repl_close(>mrecv_sock_fd);
repl_connection_reset = TRUE;
CLEAR_REPL_TLS_REQUESTED; /* As if -tlsid qualifier was never specified. */
return FALSE;
}
#endif
STATICFNDEF void do_main_loop(boolean_t crash_restart)
{
/* The work-horse of the Receiver Server */
boolean_t dont_reply_to_heartbeat = FALSE, is_repl_cmpc, pha_peer_cert_check = FALSE;
boolean_t uncmpfail, send_cross_endian, recvpool_prepared = FALSE, copied_to_recvpool = FALSE;
gtmrecv_local_ptr_t gtmrecv_local;
gtm_time4_t ack_time;
char print_msg[PROC_RECVOPS_PRINT_MSG_LEN];
char print_msg_t[PROC_RECVOPS_PRINT_MSG_LEN];
int4 msghdrlen, strm_num, processed_hdrlen = 0;
int4 need_histinfo_num;
int cmpret;
int msg_type = REPL_MSGTYPE_LAST, msg_len, hdr_msg_type, hdr_msg_len;
int torecv_len, recvd_len, recvd_this_iter; /* needed for REPL_RECV_LOOP */
int tosend_len, sent_len, sent_this_iter; /* needed for REPL_SEND_LOOP */
int status, poll_dir; /* needed for REPL_{SEND,RECV}_LOOP */
recvpool_ctl_ptr_t recvpool_ctl;
repl_cmpinfo_msg_ptr_t cmptest_msg;
repl_cmpinfo_msg_t cmpsolve_msg;
repl_cmpmsg_ptr_t cmpmsgp;
repl_heartbeat_msg_t heartbeat;
repl_histinfo histinfo;
repl_needhistinfo_msg_ptr_t need_histinfo_msg;
repl_needinst_msg_ptr_t need_instinfo_msg;
repl_needstrminfo_msg_ptr_t need_strminfo_msg;
repl_old_instinfo_msg_t old_instinfo_msg;
repl_old_needinst_msg_ptr_t old_need_instinfo_msg;
repl_start_msg_ptr_t msgp;
repl_start_reply_msg_t *start_msg;
seq_num ack_seqno, temp_ack_seqno;
seq_num request_from, recvd_jnl_seqno;
sgmnt_addrs *repl_csa;
uchar_ptr_t old_buffp, buffp_start = NULL;
uint4 recvd_start_flags, len;
uLong cmplen;
uLongf destlen;
unsigned char *msg_ptr; /* needed for REPL_{SEND,RECV}_LOOP */
unsigned char remote_jnl_ver;
upd_proc_local_ptr_t upd_proc_local;
repl_logfile_info_msg_t *logfile_msgp, logfile_msg;
# ifdef GTM_TLS
repl_msg_t renegotiate_msg;
boolean_t reneg_ack_sent;
repl_tlsinfo_msg_t *need_tlsinfo_msgp;
uint4 remote_lib_ver, remote_API_ver;
# endif
DCL_THREADGBL_ACCESS;
SETUP_THREADGBL_ACCESS;
recvpool_ctl = recvpool.recvpool_ctl;
upd_proc_local = recvpool.upd_proc_local;
gtmrecv_local = recvpool.gtmrecv_local;
gtmrecv_wait_for_jnl_seqno = FALSE;
assert((NULL != jnlpool->jnlpool_dummy_reg) && jnlpool->jnlpool_dummy_reg->open);
repl_csa = &FILE_INFO(jnlpool->jnlpool_dummy_reg)->s_addrs;
# ifdef DEBUG
assert(!repl_csa->hold_onto_crit);
ASSERT_VALID_JNLPOOL(repl_csa);
# endif
/* If BAD_TRANS was written by the update process, it would have updated recvpool_ctl->jnl_seqno accordingly.
* Only otherwise, do we need to wait for it to write "recvpool_ctl->jnl_seqno".
*/
if (!gtmrecv_bad_trans_sent)
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Waiting for Update Process to write jnl_seqno\n");
while (QWEQ(recvpool_ctl->jnl_seqno, seq_num_zero))
{
SHORT_SLEEP(GTMRECV_WAIT_FOR_STARTJNLSEQNO);
GTMRECV_POLL_ACTIONS(0, 0, NULL);
}
/* The call to "gtmrecv_poll_actions" above might have set the variable "gtmrecv_wait_for_jnl_seqno" to TRUE.
* In that case, we need to reset it to FALSE here as we are now going to wait for the jnl_seqno below.
* Not doing so will cause us to wait for jnl_seqno TWICE (once now and once when we later enter this function).
*/
gtmrecv_wait_for_jnl_seqno = FALSE;
/* Since remote primary is multisite capable (otherwise we would have issued an error), we need to send the
* journal seqno of this instance for comparison. If the receiver has received more seqnos than have been
* processed by the update process, we should be sending the last received seqno across to avoid receiving
* duplicate and out-of-order seqnos. This is maintained in "recvpool_ctl->jnl_seqno" and is guaranteed to
* be greater than or equal to the journal seqno of this instance.
*/
request_from = recvpool_ctl->jnl_seqno;
/* If this is the first time the update process initialized "recvpool_ctl->jnl_seqno", it should be
* equal to "jnlpool_ctl->jnl_seqno". But if the receiver had already connected and received a bunch
* of seqnos and if the update process did not process all of them and if the receiver disconnects
* and re-establishes the connection, the value of "recvpool_ctl->jnl_seqno" could be greater than
* "jnlpool_ctl->jnl_seqno" if there is non-zero backlog on the secondary. Assert accordingly.
* There is one exception to this and that is if this is a root primary supplementary instance.
* In that case, the receive pool talks about the non-supplementary instance stream jnl_seqnos whereas
* the jnlpool talks about the merged stream of jnl_seqnos (including any local updates). Therefore we
* cannot compare the two at all.
*/
assert((!jnlpool->jnlpool_ctl->upd_disabled && jnlpool->repl_inst_filehdr->is_supplementary)
|| (recvpool_ctl->jnl_seqno >= jnlpool->jnlpool_ctl->jnl_seqno));
assert(request_from);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Requesting transactions from JNL_SEQNO "INT8_FMT" "INT8_FMTX"\n",
request_from, request_from);
/* Send (re)start JNL_SEQNO to Source Server.
* Note that even though we might know the endianness of the source side at this time, we still send this
* message in the native endian format. This keeps the logic on the source server side simple. This is the only
* exception to the general rule that the receiver server does all endian conversion for messages except the
* first one in the connection handshake. The source side knows to endian convert this (instead of expecting it
* to be in the source endian format) if this EPL_START_JNL_SEQNO message is not the first one in the connection.
* The only exception is if the source side is pre-V55000 AND we have already determined the endianness of the
* source side (i.e. the REPL_START_JNL_SEQNO message about to be sent is not the first one for this connection)
* AND the source is cross-endian. In that case, the pre-V55000 source does not know to handle a
* receiver-side-native-endian format message. So endian convert the message only in this case.
*/
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Sending REPL_START_JNL_SEQNO message with seqno "INT8_FMT" "INT8_FMTX"\n",
request_from, request_from);
send_cross_endian = (remote_side->endianness_known && remote_side->cross_endian
&& (REPL_PROTO_VER_SUPPLEMENTARY > remote_side->proto_ver));
msgp = (repl_start_msg_ptr_t)gtmrecv_msgp;
memset(msgp, 0, SIZEOF(*msgp));
/* Since REPL_START_JNL_SEQNO is 0, there is no endian conversion necessary but for completeness we do it */
msgp->type = send_cross_endian ? GTM_BYTESWAP_32(REPL_START_JNL_SEQNO) : REPL_START_JNL_SEQNO;
if (send_cross_endian)
*((seq_num *)msgp->start_seqno) = GTM_BYTESWAP_64(request_from);
else
*((seq_num *)msgp->start_seqno) = request_from;
msgp->start_flags = START_FLAG_NONE;
msgp->start_flags |= (gtmrecv_options.stopsourcefilter ? START_FLAG_STOPSRCFILTER : 0);
/* If -UPDATERESYNC is specified then let pre-V55000 source server know so it does not ask for history record
* exchange. In case of post-V55000 source server, we expect a value to the -updateresync qualifier (the instance
* file name) and that is used for the history record exchange. In that case the source server ignores the
* START_FLAG_UPDATERESYNC bit and requests a history exchange anyways.
*/
msgp->start_flags |= (gtmrecv_local->updateresync ? START_FLAG_UPDATERESYNC : 0);
/* Let source server know if -NORESYNC was specified in receiver server startup */
msgp->start_flags |= (gtmrecv_local->noresync ? START_FLAG_NORESYNC : 0);
msgp->start_flags |= START_FLAG_HASINFO;
if (this_side->is_std_null_coll)
msgp->start_flags |= START_FLAG_COLL_M;
msgp->start_flags |= START_FLAG_VERSION_INFO;
GTMTRIG_ONLY(msgp->start_flags |= START_FLAG_TRIGGER_SUPPORT;)
# ifdef GTM_TLS
if (REPL_TLS_REQUESTED)
{
assert(NULL != tls_ctx); /* gtm_tls_init() must have already happened. */
msgp->start_flags |= START_FLAG_ENABLE_TLS;
}
# endif
if (send_cross_endian)
msgp->start_flags = GTM_BYTESWAP_32(msgp->start_flags);
msgp->jnl_ver = this_side->jnl_ver;
msgp->proto_ver = REPL_PROTO_VER_THIS;
msgp->node_endianness = NODE_ENDIANNESS;
msgp->is_supplementary = jnlpool->repl_inst_filehdr->is_supplementary;
msgp->len = send_cross_endian ? GTM_BYTESWAP_32(MIN_REPL_MSGLEN) : MIN_REPL_MSGLEN;
msg_len = MIN_REPL_MSGLEN;
REPL_SEND_LOOP(gtmrecv_sock_fd, msgp, msg_len, REPL_POLL_NOWAIT)
{
GTMRECV_POLL_ACTIONS(0, 0, NULL);
}
CHECK_REPL_SEND_LOOP_ERROR(status, "REPL_START_JNL_SEQNO");
}
gtmrecv_bad_trans_sent = FALSE;
request_from = recvpool_ctl->jnl_seqno;
assert(request_from >= seq_num_one);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Waiting for REPL_WILL_RESTART_WITH_INFO or REPL_ROLLBACK_FIRST message\n");
/* Receive journal data and place it in the Receive Pool */
buff_start = (unsigned char *)gtmrecv_msgp;
buffp = buff_start;
buff_unprocessed = 0;
data_len = 0;
write_loc = recvpool_ctl->write;
write_wrap = recvpool_ctl->write_wrap;
repl_recv_data_recvd = 0;
repl_recv_data_processed = 0;
repl_recv_postfltr_data_procd = 0;
repl_recv_lastlog_data_recvd = 0;
repl_recv_lastlog_data_procd = 0;
msghdrlen = REPL_MSG_HDRLEN;
GTMTLS_ONLY(DEBUG_ONLY(reneg_ack_sent = FALSE));
GTMTLS_ONLY(repl_tls.renegotiate_state = REPLTLS_RENEG_STATE_NONE);
while (TRUE)
{
recvd_len = gtmrecv_max_repl_msglen - buff_unprocessed;
GTMTLS_ONLY(poll_dir = REPL_INVALID_POLL_DIRECTION);
while ((SS_NORMAL == (status = repl_recv(gtmrecv_sock_fd,
(buffp + buff_unprocessed), &recvd_len, REPL_POLL_WAIT, &poll_dir)))
&& (0 == recvd_len))
{
recvd_len = gtmrecv_max_repl_msglen - buff_unprocessed;
if (xoff_sent)
{
DO_FLOW_CONTROL(write_loc);
}
if (xoff_sent && GTMRECV_XOFF_LOG_CNT <= xoff_msg_log_cnt)
{ /* Update process is still running slow; force wait before logging any message. */
SHORT_SLEEP(MILLISECS_IN_SEC - 1); /* Sleep for one second. */
REPL_DPRINT1("Waiting for Update Process to clear recvpool space\n");
xoff_msg_log_cnt = 0;
} else if (xoff_sent)
xoff_msg_log_cnt++;
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
if (SS_NORMAL != status)
{
if (EREPL_RECV == repl_errno)
{
if (REPL_CONN_RESET(status))
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Connection reset. Status = %d ; %s\n",
status, STRERROR(status));
repl_log_conn_info(gtmrecv_sock_fd, gtmrecv_log_fp, TRUE);
SEND_SYSMSG_REPLCOMM(LEN_AND_LIT("Error in receiving from source. Error in recv"));
}
} else if (EREPL_SELECT == repl_errno)
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Error in receiving from source."
" Error in select: %d ; %s\n", status, STRERROR(status));
ISSUE_REPLCOMM_ERROR("Error in receiving from source. Error in select", status);
}
repl_connection_reset = TRUE;
repl_close(>mrecv_sock_fd);
return;
}
if (repl_connection_reset)
return;
if (pha_peer_cert_check && gtm_tls_did_post_hand_shake(repl_tls.sock))
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Post-handshake authentication completed successfully\n");
repl_log_tls_info(gtmrecv_log_fp, repl_tls.sock);
pha_peer_cert_check = FALSE;
}
# ifdef REPL_CMP_SOLVE_TESTING
/* Received communication from the source server, so we can cancel the timer */
if (TREF(gtm_environment_init) && repl_cmp_solve_timer_set)
{
cancel_timer((TID)repl_cmp_solve_rcv_timeout);
repl_cmp_solve_timer_set = FALSE;
}
# endif
/* Something on the replication pipe - read it */
REPL_DPRINT3("Pending data len : %d Prev buff unprocessed : %ld\n", data_len, buff_unprocessed);
buff_unprocessed += recvd_len;
repl_recv_data_recvd += (qw_num)recvd_len;
if (gtmrecv_logstats)
repl_log(gtmrecv_log_fp, FALSE, FALSE, "Recvd : %d Total : %ld\n", recvd_len, repl_recv_data_recvd);
while (msghdrlen <= buff_unprocessed)
{
buffp_start = buffp;
if (0 == data_len)
{
assert(0 == ((unsigned long)buffp % REPL_MSG_ALIGN));
DEBUG_ONLY(recvpool_prepared = FALSE);
if (!remote_side->endianness_known)
{
remote_side->endianness_known = TRUE;
msg_type = ((repl_msg_ptr_t)buffp)->type;
if (!((REPL_MSGTYPE_LAST > msg_type) || (REPL_MSGTYPE_LAST > GTM_BYTESWAP_32(msg_type))))
WACKY_MESSAGE(msg_type, >mrecv_sock_fd, 1); /* impossible message type */
if ((REPL_MSGTYPE_LAST < msg_type) && (REPL_MSGTYPE_LAST > GTM_BYTESWAP_32(msg_type)))
{
remote_side->cross_endian = TRUE;
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Source and Receiver sides have opposite "
"endianness\n");
} else
{
remote_side->cross_endian = FALSE;
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Source and Receiver sides have same "
"endianness\n");
}
}
if (remote_side->cross_endian)
{
hdr_msg_type = GTM_BYTESWAP_32(((repl_msg_ptr_t)buffp)->type);
hdr_msg_len = GTM_BYTESWAP_32(((repl_msg_ptr_t)buffp)->len);
} else
{
hdr_msg_type = ((repl_msg_ptr_t)buffp)->type;
hdr_msg_len = ((repl_msg_ptr_t)buffp)->len;
}
if (0 >= hdr_msg_len)
WACKY_MESSAGE(hdr_msg_type, >mrecv_sock_fd, 2); /* invalid length - overflow */
msg_type = hdr_msg_type & REPL_TR_CMP_MSG_TYPE_MASK;
if (REPL_TR_CMP_JNL_RECS == msg_type)
{
processed_hdrlen = REPL_MSG_HDRLEN;
msg_len = hdr_msg_len - processed_hdrlen;
gtmrecv_repl_cmpmsglen = msg_len;
gtmrecv_repl_uncmpmsglen = (hdr_msg_type >> REPL_TR_CMP_MSG_TYPE_BITS);
assert(0 < gtmrecv_repl_uncmpmsglen);
assert(REPL_TR_CMP_THRESHOLD > gtmrecv_repl_uncmpmsglen);
/* Since msg_len is compressed length, it need not be 8-byte aligned. Make it so
* since 8-byte aligned length would have been sent by the source server anyways.
*/
msg_len = ROUND_UP(msg_len, REPL_MSG_ALIGN);
buffp += processed_hdrlen;
exp_data_len = gtmrecv_repl_uncmpmsglen;
buff_unprocessed -= processed_hdrlen;
GTMRECV_SET_BUFF_TARGET_CMPBUFF(gtmrecv_repl_cmpmsglen, gtmrecv_repl_uncmpmsglen,
gtmrecv_cur_cmpmsglen);
} else if (REPL_TR_CMP_JNL_RECS2 == msg_type)
{ /* A REPL_TR_CMP_JNL_RECS2 message is special in that it has a bigger message header.
* So check if unprocessed length is greater than the header. If not need to read more.
*/
msghdrlen = REPL_MSG_HDRLEN2;
if (msghdrlen > buff_unprocessed) /* Did not receive the full-header. */
break; /* Break out of here and read more data first. */
msghdrlen = REPL_MSG_HDRLEN; /* reset to regular msg hdr length for future messages */
processed_hdrlen = REPL_MSG_HDRLEN2;
cmpmsgp = (repl_cmpmsg_ptr_t)buffp;
if (remote_side->cross_endian)
{
cmpmsgp->cmplen = GTM_BYTESWAP_32(cmpmsgp->cmplen);
cmpmsgp->uncmplen = GTM_BYTESWAP_32(cmpmsgp->uncmplen);
}
gtmrecv_repl_cmpmsglen = cmpmsgp->cmplen;
gtmrecv_repl_uncmpmsglen = cmpmsgp->uncmplen;
assert(0 < gtmrecv_repl_uncmpmsglen);
assert(REPL_TR_CMP_THRESHOLD <= gtmrecv_repl_uncmpmsglen);
msg_len = hdr_msg_len - processed_hdrlen;
/* Unlike REPL_TR_CMP_JNL_RECS message, msg_len is guaranteed to be 8-byte aligned here */
buffp += processed_hdrlen;
exp_data_len = gtmrecv_repl_uncmpmsglen;
buff_unprocessed -= processed_hdrlen;
GTMRECV_SET_BUFF_TARGET_CMPBUFF(gtmrecv_repl_cmpmsglen, gtmrecv_repl_uncmpmsglen,
gtmrecv_cur_cmpmsglen);
} else
{
processed_hdrlen = REPL_MSG_HDRLEN;
msg_len = hdr_msg_len - processed_hdrlen;
exp_data_len = msg_len;
if (REPL_TR_JNL_RECS == msg_type || REPL_OLD_TRIPLE == msg_type || REPL_HISTREC == msg_type)
{
/* Target buffer is the receive pool. Prepare the receive pool for write (also
* checks if the transaction will fit in).
* Note: this is a special case where PREPARE_RECVPOOL_FOR_WRITE is not invoked
* right before COPY_TO_RECVPOOL because we want to prepare the receive pool just
* once even though the actual data (coming from the other end) might come in
* different pieces.
*/
PREPARE_RECVPOOL_FOR_WRITE(exp_data_len, 0);
DEBUG_ONLY(recvpool_prepared = TRUE);
} /* for REPL_TR_CMP_JNL_RECS{2}, receive pool is prepared after uncompression */
buffp += processed_hdrlen;
buff_unprocessed -= processed_hdrlen;
}
assert(0 <= buff_unprocessed);
assert(0 == (msg_len % REPL_MSG_ALIGN));
data_len = msg_len;
assert(0 == (exp_data_len % REPL_MSG_ALIGN));
}
assert(0 == (data_len % REPL_MSG_ALIGN));
buffered_data_len = ((data_len <= buff_unprocessed) ? data_len : buff_unprocessed);
buffered_data_len = ROUND_DOWN2(buffered_data_len, REPL_MSG_ALIGN);
old_buffp = buffp;
buffp += buffered_data_len;
buff_unprocessed -= buffered_data_len;
assert(0 <= buff_unprocessed);
data_len -= buffered_data_len;
/* Once we have sent a REPL_RENEG_ACK, the only message we should get is the REPL_RENEG_COMPLETE. */
assert(buffp > buff_start);
assert(buffp <= (buff_start + gtmrecv_max_repl_msglen));
assert(!reneg_ack_sent || (REPL_RENEG_COMPLETE == msg_type));
copied_to_recvpool = FALSE;
switch(msg_type)
{
case REPL_TR_JNL_RECS:
case REPL_TR_CMP_JNL_RECS:
case REPL_TR_CMP_JNL_RECS2:
case REPL_OLD_TRIPLE:
case REPL_HISTREC:
is_repl_cmpc = ((REPL_TR_CMP_JNL_RECS == msg_type) || (REPL_TR_CMP_JNL_RECS2 == msg_type));
if (!is_repl_cmpc)
{
assert(recvpool_prepared);
COPY_TO_RECVPOOL(old_buffp, buffered_data_len); /* uses and updates "write_loc" */
} else
{
memcpy(gtmrecv_cmpmsgp + gtmrecv_cur_cmpmsglen, old_buffp, buffered_data_len);
gtmrecv_cur_cmpmsglen += buffered_data_len;
assert(gtmrecv_cur_cmpmsglen <= gtmrecv_max_repl_cmpmsglen);
}
/* The partial/complete message is copied either to the receive pool or the private
* compression buffer space. Set copied_to_recvpool to TRUE so that we can safely
* set buffp = buff_start if this is the last message in the current received buffer.
*/
copied_to_recvpool = TRUE;
repl_recv_data_processed += (qw_num)buffered_data_len;
if (0 == data_len)
{
process_tr_buff(msg_type);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
}
break;
case REPL_LOSTTNCOMPLETE:
if (0 == data_len)
{
assert(REPL_PROTO_VER_MULTISITE <= remote_side->proto_ver);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_LOSTTNCOMPLETE message\n");
status = repl_inst_reset_zqgblmod_seqno_and_tn();
assert(-1 == status || EXIT_ERR == status || SS_NORMAL == status);
if (-1 == status)
{ /* only reason we know currently for the above function to return -1 is due
* to a concurrent online rollback. In this case, we cannot continue
* and need to start afresh.
*/
gtmrecv_onln_rlbk_clnup();
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
}
}
break;
case REPL_HEARTBEAT:
if (0 == data_len)
{ /* Heartbeat msg contents start from buffp - msg_len */
GTM_WHITE_BOX_TEST(WBTEST_REPL_HEARTBEAT_NO_ACK, dont_reply_to_heartbeat, TRUE);
if (dont_reply_to_heartbeat)
{
dont_reply_to_heartbeat = FALSE;
break;
}
memcpy(heartbeat.ack_seqno, buffp - msg_len, msg_len);
assert(remote_side->endianness_known); /* only then is remote_side->cross_endian
* reliable */
if (!remote_side->cross_endian)
{
ack_time = *(gtm_time4_t *)&heartbeat.ack_time[0];
memcpy((uchar_ptr_t)&ack_seqno,
(uchar_ptr_t)&heartbeat.ack_seqno[0], SIZEOF(seq_num));
} else
{
ack_time = GTM_BYTESWAP_32(*(gtm_time4_t *)&heartbeat.ack_time[0]);
memcpy((uchar_ptr_t)&temp_ack_seqno,
(uchar_ptr_t)&heartbeat.ack_seqno[0], SIZEOF(seq_num));
ack_seqno = GTM_BYTESWAP_64(temp_ack_seqno);
}
REPL_DPRINT4("HEARTBEAT received with time %ld SEQNO "INT8_FMT" at %ld\n",
ack_time, ack_seqno, time(NULL));
ack_seqno = upd_proc_local->read_jnl_seqno;
if (!remote_side->cross_endian)
{
heartbeat.type = REPL_HEARTBEAT;
heartbeat.len = MIN_REPL_MSGLEN;
memcpy((uchar_ptr_t)&heartbeat.ack_seqno[0],
(uchar_ptr_t)&ack_seqno, SIZEOF(seq_num));
} else
{
heartbeat.type = GTM_BYTESWAP_32(REPL_HEARTBEAT);
heartbeat.len = GTM_BYTESWAP_32(MIN_REPL_MSGLEN);
temp_ack_seqno = GTM_BYTESWAP_64(ack_seqno);
memcpy((uchar_ptr_t)&heartbeat.ack_seqno[0],
(uchar_ptr_t)&temp_ack_seqno, SIZEOF(seq_num));
}
REPL_SEND_LOOP(gtmrecv_sock_fd, &heartbeat, MIN_REPL_MSGLEN, REPL_POLL_NOWAIT)
{
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
CHECK_REPL_SEND_LOOP_ERROR(status, "REPL_HEARTBEAT");
REPL_DPRINT4("HEARTBEAT sent with time %ld SEQNO "INT8_FMT" at %ld\n",
ack_time, ack_seqno, time(NULL));
}
break;
case REPL_OLD_NEED_INSTANCE_INFO:
if (0 == data_len)
{
assert(msg_len == MIN_REPL_MSGLEN - REPL_MSG_HDRLEN);
old_need_instinfo_msg = (repl_old_needinst_msg_ptr_t)(buffp -
msg_len - REPL_MSG_HDRLEN);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_OLD_NEED_INSTANCE_INFO message"
" from primary instance [%s]\n", old_need_instinfo_msg->instname);
if (jnlpool->repl_inst_filehdr->is_supplementary)
{ /* Issue REPL2OLD error because this is a supplementary instance and remote
* side runs a GT.M version that does not know the supplementary protocol */
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_REPL2OLD, 4,
LEN_AND_STR(old_need_instinfo_msg->instname),
LEN_AND_STR(jnlpool->repl_inst_filehdr->inst_info.this_instname));
}
/* The source server does not understand the supplementary protocol.
* So make sure -UPDATERESYNC if specified at receiver server startup
* had no value. If not, issue error.
*/
if (gtmrecv_local->updateresync
&& (FD_INVALID != gtmrecv_local->updresync_instfile_fd))
{
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_UPDSYNCINSTFILE, 0,
ERR_TEXT, 2,
LEN_AND_LIT("Source side is non-supplementary implies "
"-UPDATERESYNC needs no value specified"));
}
/* Initialize the remote side protocol version from "proto_ver"
* field of this msg
*/
remote_side->proto_ver = old_need_instinfo_msg->proto_ver;
assert(REPL_PROTO_VER_MULTISITE <= remote_side->proto_ver);
assert(REPL_PROTO_VER_SUPPLEMENTARY > remote_side->proto_ver);
/*************** Send REPL_OLD_INSTANCE_INFO message ***************/
memset(&old_instinfo_msg, 0, SIZEOF(old_instinfo_msg));
memcpy(old_instinfo_msg.instname,
jnlpool->repl_inst_filehdr->inst_info.this_instname, MAX_INSTNAME_LEN - 1);
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
old_instinfo_msg.was_rootprimary = (unsigned char)repl_inst_was_rootprimary();
rel_lock(jnlpool->jnlpool_dummy_reg);
gtmrecv_repl_send((repl_msg_ptr_t)&old_instinfo_msg, REPL_OLD_INSTANCE_INFO,
MIN_REPL_MSGLEN, "REPL_OLD_INSTANCE_INFO", MAX_SEQNO);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
/* Do not allow an instance which was formerly a root primary or which still
* has a non-zero value of "zqgblmod_seqno" to start up as a tertiary.
*/
if ((old_instinfo_msg.was_rootprimary || jnlpool->jnlpool_ctl->max_zqgblmod_seqno)
&& !old_need_instinfo_msg->is_rootprimary)
{
gtm_putmsg_csa(CSA_ARG(NULL) VARLSTCNT(4) ERR_PRIMARYNOTROOT, 2,
LEN_AND_STR((char *) old_need_instinfo_msg->instname));
gtmrecv_autoshutdown(); /* should not return */
assert(FALSE);
}
memcpy(jnlpool->jnlpool_ctl->primary_instname, old_need_instinfo_msg->instname,
MAX_INSTNAME_LEN - 1);
}
break;
case REPL_NEED_INSTINFO:
if (0 == data_len)
{
assert(msg_len == SIZEOF(repl_needinst_msg_t) - REPL_MSG_HDRLEN);
need_instinfo_msg = (repl_needinst_msg_ptr_t)(buffp - msg_len - REPL_MSG_HDRLEN);
gtmrecv_check_and_send_instinfo(need_instinfo_msg, IS_RCVR_SRVR_TRUE);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
}
break;
case REPL_CMP_TEST:
if (0 == data_len)
{
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_CMP_TEST message\n");
uncmpfail = FALSE;
if (ZLIB_CMPLVL_NONE == gtm_zlib_cmp_level)
{ /* Receiver does not have compression enabled in the first place.
* Send dummy REPL_CMP_SOLVE response message.
*/
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Environment variable "
"gtm_zlib_cmp_level specifies NO decompression (set to %d)\n",
gtm_zlib_cmp_level);
uncmpfail = TRUE;
}
assert(remote_side->endianness_known); /* ensure remote_side->cross_endian
* is reliable */
if (!uncmpfail)
{
assert(msg_len == REPL_MSG_CMPINFOLEN - REPL_MSG_HDRLEN);
cmptest_msg = (repl_cmpinfo_msg_ptr_t)(buffp - msg_len - REPL_MSG_HDRLEN);
if (!remote_side->cross_endian)
cmplen = cmptest_msg->datalen;
else
cmplen = GTM_BYTESWAP_32(cmptest_msg->datalen);
if (REPL_MSG_CMPEXPDATALEN < cmplen)
{
assert(FALSE); /* since src srvr should not have sent such a msg */
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Compression test message "
"has compressed data length (%d) greater than receiver "
"allocated length (%d)\n", (int)cmplen,
REPL_MSG_CMPEXPDATALEN);
uncmpfail = TRUE;
}
}
if (!uncmpfail)
{
destlen = REPL_MSG_CMPEXPDATALEN; /* initialize available
* decompressed buffer space */
ZLIB_UNCOMPRESS(&cmpsolve_msg.data[0], destlen, &cmptest_msg->data[0],
cmplen, cmpret);
GTM_WHITE_BOX_TEST(WBTEST_REPL_TEST_UNCMP_ERROR, cmpret, Z_DATA_ERROR);
switch(cmpret)
{
case Z_MEM_ERROR:
assert(FALSE);
repl_log(gtmrecv_log_fp, TRUE, TRUE,
GTM_ZLIB_Z_MEM_ERROR_STR
GTM_ZLIB_UNCMP_ERR_SOLVE_STR);
break;
case Z_BUF_ERROR:
assert(FALSE);
repl_log(gtmrecv_log_fp, TRUE, TRUE,
GTM_ZLIB_Z_BUF_ERROR_STR
GTM_ZLIB_UNCMP_ERR_SOLVE_STR);
break;
case Z_DATA_ERROR:
assert(gtm_white_box_test_case_enabled
&& (WBTEST_REPL_TEST_UNCMP_ERROR
== gtm_white_box_test_case_number));
repl_log(gtmrecv_log_fp, TRUE, TRUE,
GTM_ZLIB_Z_DATA_ERROR_STR
GTM_ZLIB_UNCMP_ERR_SOLVE_STR);
break;
}
if (Z_OK != cmpret)
uncmpfail = TRUE;
}
if (!uncmpfail)
{
cmpsolve_msg.datalen = (int4)destlen;
GTM_WHITE_BOX_TEST(WBTEST_REPL_TEST_UNCMP_ERROR, cmpsolve_msg.datalen,
REPL_MSG_CMPDATALEN - 1);
if (REPL_MSG_CMPDATALEN != cmpsolve_msg.datalen)
{ /* decompression did not yield precompressed data length */
assert(gtm_white_box_test_case_enabled
&& (WBTEST_REPL_TEST_UNCMP_ERROR
== gtm_white_box_test_case_number));
repl_log(gtmrecv_log_fp, TRUE, TRUE, GTM_ZLIB_UNCMPLEN_ERROR_STR
"\n", cmpsolve_msg.datalen, REPL_MSG_CMPDATALEN);
uncmpfail = TRUE;
}
}
if (uncmpfail)
{
cmpsolve_msg.datalen = REPL_RCVR_CMP_TEST_FAIL;
repl_log(gtmrecv_log_fp, TRUE, TRUE, GTM_ZLIB_UNCMPTRANSITION_STR);
}
if (remote_side->cross_endian)
cmpsolve_msg.datalen = GTM_BYTESWAP_32(cmpsolve_msg.datalen);
cmpsolve_msg.proto_ver = REPL_PROTO_VER_THIS;
# ifdef REPL_CMP_SOLVE_TESTING
if (TREF(gtm_environment_init))
{
start_timer((TID)repl_cmp_solve_rcv_timeout, 15 * 60 * 1000,
repl_cmp_solve_rcv_timeout, 0, NULL);
repl_cmp_solve_timer_set = TRUE;
}
# endif
gtmrecv_repl_send((repl_msg_ptr_t)&cmpsolve_msg, REPL_CMP_SOLVE,
REPL_MSG_CMPINFOLEN, "REPL_CMP_SOLVE", MAX_SEQNO);
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
if (!uncmpfail)
repl_zlib_cmp_level = gtm_zlib_cmp_level;
}
break;
case REPL_NEED_STRMINFO:
if (0 == data_len)
{
assert(REPL_PROTO_VER_SUPPLEMENTARY <= remote_side->proto_ver);
assert(remote_side->is_supplementary);
assert(msg_len == MIN_REPL_MSGLEN - REPL_MSG_HDRLEN);
need_strminfo_msg = (repl_needstrminfo_msg_ptr_t)(buffp - msg_len
- REPL_MSG_HDRLEN);
gtmrecv_process_need_strminfo_msg(need_strminfo_msg);
/* Check for error return from above function call */
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
}
break;
case REPL_NEED_HISTINFO:
/* case REPL_NEED_TRIPLE_INFO: too but that message has been renamed to REPL_NEED_HISTINFO */
if (0 == data_len)
{
assert(REPL_PROTO_VER_MULTISITE <= remote_side->proto_ver);
assert(msg_len == MIN_REPL_MSGLEN - REPL_MSG_HDRLEN);
need_histinfo_msg = (repl_needhistinfo_msg_ptr_t)(buffp - msg_len
- REPL_MSG_HDRLEN);
gtmrecv_process_need_histinfo_msg(need_histinfo_msg, &histinfo);
/* Check for error return from above function call */
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
/* Send the histinfo */
gtmrecv_send_histinfo(&histinfo);
/* Check for error return from above function call as well */
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
}
break;
case REPL_WILL_RESTART_WITH_INFO:
case REPL_ROLLBACK_FIRST:
if (0 != data_len)
break;
/* Have received a REPL_WILL_RESTART_WITH_INFO or REPL_ROLLBACK_FIRST message. If
* have not yet received a REPL_OLD_NEED_INSTANCE_INFO or REPL_NEED_INSTINFO message
* (which would have initialized "remote_side->proto_ver"), it
* means the remote side does not understand multi-site replication communication
* protocol. Note that down.
*/
if (REPL_PROTO_VER_UNINITIALIZED == remote_side->proto_ver)
{ /* Issue REPL2OLD error because primary is dual-site */
RTS_ERROR_CSA_ABT(NULL, VARLSTCNT(6) ERR_REPL2OLD, 4,
LEN_AND_STR(UNKNOWN_INSTNAME),
LEN_AND_STR(jnlpool->repl_inst_filehdr->inst_info.this_instname));
}
/* Assert that endianness_known and cross_endian have already been initialized.
* This ensures that remote_side->cross_endian is reliable */
assert(remote_side->endianness_known);
assert(REPL_PROTO_VER_MULTISITE <= remote_side->proto_ver);
assert(msg_len == MIN_REPL_MSGLEN - REPL_MSG_HDRLEN);
start_msg = (repl_start_reply_msg_ptr_t)(buffp - msg_len - REPL_MSG_HDRLEN);
assert((unsigned long)start_msg % SIZEOF(seq_num) == 0); /* alignment check */
memcpy((uchar_ptr_t)&recvd_jnl_seqno,
(uchar_ptr_t)start_msg->start_seqno, SIZEOF(seq_num));
/* Assert that "node_endianness" field reflects our cross-endian understanding */
BIGENDIAN_ONLY(assert((REPL_PROTO_VER_SUPPLEMENTARY > remote_side->proto_ver)
|| (remote_side->cross_endian
&& (LITTLE_ENDIAN_MARKER == start_msg->node_endianness))
|| (!remote_side->cross_endian
&& (BIG_ENDIAN_MARKER == start_msg->node_endianness))));
LITTLEENDIAN_ONLY(assert((REPL_PROTO_VER_SUPPLEMENTARY > remote_side->proto_ver)
|| (remote_side->cross_endian
&& (BIG_ENDIAN_MARKER == start_msg->node_endianness))
|| (!remote_side->cross_endian
&& (LITTLE_ENDIAN_MARKER == start_msg->node_endianness))));
if (remote_side->cross_endian)
recvd_jnl_seqno = GTM_BYTESWAP_64(recvd_jnl_seqno);
/* Handle REPL_ROLLBACK_FIRST case (easy one) first */
if (REPL_ROLLBACK_FIRST == msg_type)
{
REPL_DPRINT1("Received REPL_ROLLBACK_FIRST message");
sgtm_putmsg(print_msg, PROC_RECVOPS_PRINT_MSG_LEN, VARLSTCNT(4) ERR_REPLAHEAD, 2,
LEN_AND_LIT(""));
repl_log(gtmrecv_log_fp, TRUE, TRUE, print_msg);
SNPRINTF(print_msg_t, SIZEOF(print_msg_t), "Replicating instance : "INT8_FMT" "
"Originating instance : "INT8_FMT". ", request_from, recvd_jnl_seqno);
sgtm_putmsg(print_msg, PROC_RECVOPS_PRINT_MSG_LEN, VARLSTCNT(4) ERR_TEXT, 2,
LEN_AND_STR(print_msg_t));
repl_log(gtmrecv_log_fp, TRUE, TRUE, print_msg);
if (gtmrecv_options.autorollback)
{
sgtm_putmsg(print_msg, PROC_RECVOPS_PRINT_MSG_LEN, VARLSTCNT(4) ERR_TEXT, 2,
LEN_AND_LIT("Initiating automatic rollback to match the replicating"
" instance with a prior in-sync originating instance state."));
repl_log(gtmrecv_log_fp, TRUE, TRUE, print_msg);
repl_connection_reset = TRUE;
repl_close(>mrecv_sock_fd);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Closing connection before starting "
"ROLLBACK\n");
if (SS_NORMAL != gtmrecv_start_onln_rlbk())
{ /* gtmrecv_start_onln_rlbk() would have issued the appropriate
* error message.
*/
assert(FALSE);
gtmrecv_autoshutdown(); /* should not return */
assert(FALSE);
}
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
/* The ONLINE ROLLBACK did not change the physical or the logical state.
* Otherwise the above macro would have returned to the caller. Since we
* have already disconnected the connection by now, we cannot resume the
* flow from this point on. Return to the calling function to continue
* reconnecting.
*/
rel_lock(jnlpool->jnlpool_dummy_reg);
return;
} else
{
sgtm_putmsg(print_msg, PROC_RECVOPS_PRINT_MSG_LEN, VARLSTCNT(4) ERR_TEXT, 2,
LEN_AND_LIT(" The receiver server was not started with "
"-AUTOROLLBACK. Perform MUPIP JOURNAL -ROLLBACK on the replicating"
" instance to match the originating instance. Check the Source "
"Server log to determine the last in-sync sequence number."));
repl_log(gtmrecv_log_fp, TRUE, TRUE, print_msg);
}
gtmrecv_autoshutdown(); /* should not return */
assert(FALSE);
break;
}
/* Handle REPL_WILL_RESTART_WITH_INFO case now */
if (jnlpool->repl_inst_filehdr->was_rootprimary && jnlpool->jnlpool_ctl->upd_disabled)
{ /* This is the first time an instance that was formerly a root primary
* is brought up as an immediate secondary of the new root primary. Once
* fetchresync rollback has happened and the receiver and source server
* have communicated successfully, the instance file header field that
* indicates this was a root primary can be reset to FALSE as the zero
* or non-zeroness of the "zqgblmod_seqno" field in the respective
* database file headers henceforth controls whether this instance can
* be brought up as a tertiary or not. Flush changes to file on disk.
*/
grab_lock(jnlpool->jnlpool_dummy_reg, TRUE, GRAB_LOCK_ONLY);
GTMRECV_ONLN_RLBK_CLNUP_IF_NEEDED;
jnlpool->repl_inst_filehdr->was_rootprimary = FALSE;
repl_inst_flush_filehdr();
rel_lock(jnlpool->jnlpool_dummy_reg);
}
assert(REPL_WILL_RESTART_WITH_INFO == msg_type);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_WILL_RESTART_WITH_INFO message"
" with seqno "INT8_FMT" "INT8_FMTX"\n", recvd_jnl_seqno, recvd_jnl_seqno);
remote_side->jnl_ver = start_msg->jnl_ver;
remote_jnl_ver = remote_side->jnl_ver;
REPL_DPRINT3("Local jnl ver is octal %o, remote jnl ver is octal %o\n",
this_side->jnl_ver, remote_jnl_ver);
repl_check_jnlver_compat(!remote_side->cross_endian);
/* older versions zero filler that was in place of start_msg->start_flags,
* so we are okay fetching start_msg->start_flags unconditionally.
*/
GET_ULONG(recvd_start_flags, start_msg->start_flags);
if (remote_side->cross_endian)
recvd_start_flags = GTM_BYTESWAP_32(recvd_start_flags);
assert(remote_jnl_ver > V15_JNL_VER || 0 == recvd_start_flags);
if (remote_jnl_ver <= V15_JNL_VER) /* safety in pro */
recvd_start_flags = 0;
remote_side->is_std_null_coll = (recvd_start_flags & START_FLAG_COLL_M) ? TRUE : FALSE;
if (remote_side->is_std_null_coll != this_side->is_std_null_coll)
remote_side->null_subs_xform = (remote_side->is_std_null_coll
? STDNULL_TO_GTMNULL_COLL : GTMNULL_TO_STDNULL_COLL);
else
remote_side->null_subs_xform = FALSE;
/* this sets null_subs_xform regardless of remote_jnl_ver */
remote_side->is_supplementary = start_msg->is_supplementary;
remote_side->trigger_supported = (recvd_start_flags & START_FLAG_TRIGGER_SUPPORT) ? TRUE
: FALSE;
# ifdef GTM_TLS
remote_side->tls_requested = (recvd_start_flags & START_FLAG_ENABLE_TLS) ? TRUE : FALSE;
if (REPL_TLS_REQUESTED && !remote_side->tls_requested)
{
gtm_putmsg_csa(CSA_ARG(NULL) VARLSTCNT(6) ((PLAINTEXT_FALLBACK) ?
MAKE_MSG_WARNING(ERR_REPLNOTLS) : ERR_REPLNOTLS),
4, LEN_AND_LIT(RCVR_SIDE_STR), LEN_AND_LIT(SRC_SIDE_STR));
if (!PLAINTEXT_FALLBACK)
{ /* Close the connection and return to waiting for a new one */
repl_close(>mrecv_sock_fd);
repl_connection_reset = TRUE;
break;
}
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Plaintext fallback enabled. "
"Continuing without TLS/SSL.\n");
CLEAR_REPL_TLS_REQUESTED; /* As if -tlsid qualifier was never specified. */
}
/* If we have not requested TLS/SSL and the originating side requested TLS/SSL, the latter
* reports the REPLNOTLS error and so we need not do anything. Hence, the below assert.
*/
assert(REPL_TLS_REQUESTED || !remote_side->tls_requested);
# endif
if (this_side->jnl_ver > remote_jnl_ver)
{
assert(JNL_VER_EARLIEST_REPL <= remote_jnl_ver);
assert((remote_jnl_ver - JNL_VER_EARLIEST_REPL) < ARRAYSIZE(repl_filter_old2cur));
assert((remote_jnl_ver - JNL_VER_EARLIEST_REPL) < ARRAYSIZE(repl_filter_cur2old));
assert(IF_NONE != repl_filter_old2cur[remote_jnl_ver - JNL_VER_EARLIEST_REPL]);
assert(IF_INVALID != repl_filter_old2cur[remote_jnl_ver - JNL_VER_EARLIEST_REPL]);
/* reverse transformation should exist */
assert(IF_INVALID != repl_filter_cur2old[remote_jnl_ver - JNL_VER_EARLIEST_REPL]);
assert(IF_NONE != repl_filter_cur2old[remote_jnl_ver - JNL_VER_EARLIEST_REPL]);
gtmrecv_filter |= INTERNAL_FILTER;
gtmrecv_alloc_filter_buff(gtmrecv_max_repl_msglen);
} else
{
gtmrecv_filter &= ~INTERNAL_FILTER;
if (NO_FILTER == gtmrecv_filter)
gtmrecv_free_filter_buff();
}
/* Don't send any more stopsourcefilter message */
gtmrecv_options.stopsourcefilter = FALSE;
assert(QWEQ(recvd_jnl_seqno, request_from)
|| (jnlpool->repl_inst_filehdr->is_supplementary
&& !jnlpool->jnlpool_ctl->upd_disabled && strm_index));
assert(this_side->is_supplementary == jnlpool->repl_inst_filehdr->is_supplementary);
if (this_side->is_supplementary && !remote_side->is_supplementary)
{
/* For the non-supplementary -> supplementary replication connection that happens
* for the very first time during the lifetime of this receiver server,
* (i.e. when upd_proc_local->read_jnl_seqno is 0), until this point in time,
* we cannot be sure what the starting point of the transmission is (partly due
* to -UPDATERESYNC but primarily due to -NORESYNC). Update process will be
* waiting for receiver to set read_jnl_seqno to a non-zero value. Finish it now.
* In case this is a reconnect, recvpool.upd_proc_local->read_jnl_seqno already
* reflects how much the update process has processed so that should be untouched
* by the receiver server as it will otherwise confuse the concurrently running
* update process (see <C9J02_003091_updproc_assert_fail_due_to_seqno_gap>).
* In addition do not touch recvpool_ctl->jnl_seqno in this case as we need to
* resume filling in the receive pool from where the previous connection stopped.
*/
assert(0 == GET_STRM_INDEX(recvd_jnl_seqno));
if (0 == recvpool.upd_proc_local->read_jnl_seqno)
{ /* Set read_jnl_seqno after jnl_seqno. Update process reads it in opposite
* order. Have memory barriers in between to ensure no out-of-order reads.
*/
recvpool.recvpool_ctl->jnl_seqno = recvd_jnl_seqno;
SHM_WRITE_MEMORY_BARRIER;
recvpool.upd_proc_local->read_jnl_seqno = recvd_jnl_seqno;
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Wrote upd_proc_local->read_jnl_seqno"
" : "INT8_FMT" "INT8_FMTX"\n", recvd_jnl_seqno, recvd_jnl_seqno);
/* If -NORESYNC was used in receiver startup, we don't want to use it
* anymore as future such connections could cause recvpool_ctl->jnl_seqno
* to be reset further backwards and will confuse the update process.
*/
if (recvpool.gtmrecv_local->noresync)
recvpool.gtmrecv_local->noresync = FALSE;
} else
{
assert(recvpool.recvpool_ctl->jnl_seqno
>= recvpool.upd_proc_local->read_jnl_seqno);
assert(!recvpool.gtmrecv_local->noresync);
}
}
/* Now that recvpool.recvpool_ctl->jnl_seqno has been determined for sure (for
* non-supplementary or non-root-primary instances it is determined even before but
* otherwise it takes until now). Force a log on the first recv. This function uses
* recvpool.recvpool_ctl->jnl_seqno hence the placement here.
*/
gtmrecv_reinit_logseqno();
break;
case REPL_INST_NOHIST:
if (0 == data_len)
{
assert(msg_len == MIN_REPL_MSGLEN - REPL_MSG_HDRLEN);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Originating instance encountered a "
"REPLINSTNOHIST error. JNL_SEQNO of this replicating instance precedes"
" the current history in the originating instance file. "
"Receiver server exiting.\n");
gtmrecv_autoshutdown(); /* should not return */
assert(FALSE);
}
break;
case REPL_LOGFILE_INFO:
if (0 == data_len)
{
logfile_msgp = (repl_logfile_info_msg_t *)(buffp - msg_len - REPL_MSG_HDRLEN);
assert(REPL_PROTO_VER_REMOTE_LOGPATH <= logfile_msgp->proto_ver);
if (remote_side->cross_endian)
{
logfile_msgp->fullpath_len = GTM_BYTESWAP_32(logfile_msgp->fullpath_len);
logfile_msgp->pid = GTM_BYTESWAP_32(logfile_msgp->pid);
}
assert('\0' == logfile_msgp->fullpath[logfile_msgp->fullpath_len - 1]);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Remote side source log file path is %s; "
"Source Server PID = %d\n",
logfile_msgp->fullpath, logfile_msgp->pid);
/* Now, send our logfile path to the source side */
assert(remote_side->endianness_known);
len = repl_logfileinfo_get(recvpool.gtmrecv_local->log_file,
&logfile_msg,
remote_side->cross_endian,
gtmrecv_log_fp);
REPL_SEND_LOOP(gtmrecv_sock_fd, &logfile_msg, len, REPL_POLL_NOWAIT)
{
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
CHECK_REPL_SEND_LOOP_ERROR(status, "REPL_LOGFILE_INFO");
}
break;
# ifdef GTM_TLS
case REPL_NEED_TLS_INFO:
if (0 != data_len)
break;
assert(REPL_PROTO_VER_TLS_SUPPORT <= REPL_PROTO_VER_THIS);
assert(REPL_PROTO_VER_TLS_SUPPORT <= remote_side->proto_ver);
assert(REPL_TLS_REQUESTED);
assert(NULL != tls_ctx);
assert(!repl_tls.enabled);
need_tlsinfo_msgp = (repl_tlsinfo_msg_t *)(buffp - msg_len - REPL_MSG_HDRLEN);
remote_API_ver = need_tlsinfo_msgp->API_version;
remote_lib_ver = need_tlsinfo_msgp->library_version;
if (remote_side->cross_endian)
{
remote_API_ver = GTM_BYTESWAP_32(remote_API_ver);
remote_lib_ver = (uint4)GTM_BYTESWAP_32(remote_lib_ver);
}
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received REPL_NEED_TLS_INFO message\n");
repl_log(gtmrecv_log_fp, TRUE, TRUE, " Remote side API version: 0x%08x\n", remote_API_ver);
repl_log(gtmrecv_log_fp, TRUE, TRUE, " Remote side Library version: 0x%08x\n",
remote_lib_ver);
if (!gtmrecv_exchange_tls_info(remote_API_ver, remote_lib_ver))
{
pha_peer_cert_check = FALSE;
if (repl_connection_reset || gtmrecv_wait_for_jnl_seqno)
return;
assert(PLAINTEXT_FALLBACK);
assert(!repl_tls.enabled);
} else
{ /* From here on, all communications are secured with SSL */
repl_tls.enabled = TRUE;
pha_peer_cert_check = gtm_tls_has_post_hand_shake(repl_tls.sock);
}
break;
case REPL_RENEG_ACK_ME:
if (0 != data_len)
break;
assert(!reneg_ack_sent);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received %s acknowledgement request message\n",
(gtm_tls_does_renegotiate(repl_tls.sock)) ? "renegotiation" : "key-update");
gtmrecv_repl_send(&renegotiate_msg, REPL_RENEG_ACK, MIN_REPL_MSGLEN, "REPL_RENEG_ACK",
MAX_SEQNO);
DEBUG_ONLY(reneg_ack_sent = TRUE);
repl_tls.renegotiate_state = REPLTLS_WAITING_FOR_RENEG_COMPLETE;
break;
case REPL_RENEG_COMPLETE:
if (0 != data_len)
break;
assert(reneg_ack_sent);
repl_log(gtmrecv_log_fp, TRUE, TRUE, "Received SSL/TLS connection %s complete message.\n",
(gtm_tls_does_renegotiate(repl_tls.sock)) ? "renegotiation" : "key-update");
DEBUG_ONLY(reneg_ack_sent = FALSE);
if (gtm_tls_does_renegotiate(repl_tls.sock)) /* Announce results only if renegotiated */
repl_log_tls_info(gtmrecv_log_fp, repl_tls.sock);
repl_tls.renegotiate_state = REPLTLS_RENEG_STATE_NONE;
break;
# endif
default:
WACKY_MESSAGE(msg_type, >mrecv_sock_fd, 3); /* undefined message type */
}
if (repl_connection_reset)
return;
}
assert(0 == ((unsigned long)(buffp) % REPL_MSG_ALIGN));
if (buff_start != buffp)
{ /* We are at the tail of the current received buffer. */
if ((0 != data_len) && !copied_to_recvpool)
{ /* We have a complete header for a control message, but not the complete message. Move the message
* (including the header) to the beginning of the allocated buffer space. Restore the remaining
* length to buff_unprocessed and reset data_len so that the header is read again.
*/
assert(buffp_start);
data_len = 0;
buff_unprocessed += (buffered_data_len + processed_hdrlen);
assert((0 <= buff_unprocessed) && (buff_unprocessed <= recvpool_size));
if (buffp_start != buff_start)
memmove(buff_start, buffp_start, buff_unprocessed);
} else if (0 != buff_unprocessed)
{ /* We have an incomplete header. Move it to the beginning of the allocated buffer space. */
memmove(buff_start, buffp, buff_unprocessed);
}
buffp = buff_start;
}
GTMRECV_POLL_ACTIONS(data_len, buff_unprocessed, buffp);
}
}
void repl_cmp_solve_rcv_timeout(void)
{
assertpro(FALSE);
}
STATICFNDEF void gtmrecv_heartbeat_timer(TID tid, int4 interval_len, int *interval_ptr)
{
assert(0 != gtmrecv_now);
UNIX_ONLY(assert(*interval_ptr == heartbeat_period);) /* interval_len and interval_ptr are dummies on VMS */
gtmrecv_now += heartbeat_period;
REPL_DPRINT2("Starting heartbeat timer with %d s\n", heartbeat_period);
start_timer((TID)gtmrecv_heartbeat_timer, heartbeat_period * 1000, gtmrecv_heartbeat_timer, SIZEOF(heartbeat_period),
&heartbeat_period); /* start_timer expects time interval in milli seconds, heartbeat_period is in seconds */
}
STATICFNDEF void gtmrecv_main_loop(boolean_t crash_restart)
{
assert(FD_INVALID == gtmrecv_sock_fd);
gtmrecv_poll_actions(0, 0, NULL); /* Clear any pending bad trans */
gtmrecv_est_conn();
gtmrecv_bad_trans_sent = FALSE; /* this assignment should be after gtmrecv_est_conn since gtmrecv_est_conn can
* potentially call gtmrecv_poll_actions. If the timing is right,
* gtmrecv_poll_actions might set this variable to TRUE if the update process sets
* bad_trans in the recvpool. When we are (re)establishing connection with the
* source server, there is no point in doing bad trans processing. Also, we have
* to send START_JNL_SEQNO message to the source server. If not, there will be a
* deadlock with the source and receiver servers waiting for each other to send
* a message. */
repl_recv_prev_log_time = gtmrecv_now;
while (!repl_connection_reset)
do_main_loop(crash_restart);
return;
}
void gtmrecv_process(boolean_t crash_restart)
{
recvpool_ctl_ptr_t recvpool_ctl;
upd_proc_local_ptr_t upd_proc_local;
gtmrecv_local_ptr_t gtmrecv_local;
if (ZLIB_CMPLVL_NONE != gtm_zlib_cmp_level)
gtm_zlib_init(); /* Open zlib shared library for compression/decompression */
# ifdef GTM_TLS
if (REPL_TLS_REQUESTED)
{
repl_do_tls_init(gtmrecv_log_fp);
assert(REPL_TLS_REQUESTED || PLAINTEXT_FALLBACK);
}
# endif
recvpool_ctl = recvpool.recvpool_ctl;
upd_proc_local = recvpool.upd_proc_local;
gtmrecv_local = recvpool.gtmrecv_local;
/* Check all message sizes are the same size (32 bytes = MIN_REPL_MSGLEN) except for the REPL_OLD_TRIPLE message
* (repl_histrec_msg_t structure) which is 8 bytes more. Pre-supplementary, the receiver server knew to handle
* different sized messages only for a few messages types REPL_TR_JNL_RECS, REPL_OLD_TRIPLE and REPL_CMP_SOLVE.
* But post-supplementary it knows to handle different sized messages for various additional message types
* (including REPL_NEED_INSTINFO, REPL_INSTINFO, REPL_HISTREC).
*/
assert(MIN_REPL_MSGLEN == SIZEOF(repl_start_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_start_reply_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_resync_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_old_needinst_msg_t));
assert(MIN_REPL_MSGLEN < SIZEOF(repl_needinst_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_needhistinfo_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_old_instinfo_msg_t));
assert(MIN_REPL_MSGLEN < SIZEOF(repl_instinfo_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_histinfo1_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_histinfo2_msg_t));
assert(MIN_REPL_MSGLEN < SIZEOF(repl_histinfo_msg_t));
assert(MIN_REPL_MSGLEN < SIZEOF(repl_old_triple_msg_t));
assert(MIN_REPL_MSGLEN < SIZEOF(repl_histrec_msg_t));
assert(MIN_REPL_MSGLEN == SIZEOF(repl_heartbeat_msg_t));
assert(REPL_POLL_WAIT < MILLISECS_IN_SEC);
recvpool_size = recvpool_ctl->recvpool_size;
recvpool_high_watermark = (long)((float)RECVPOOL_HIGH_WATERMARK_PCTG / 100 * recvpool_size);
recvpool_low_watermark = (long)((float)RECVPOOL_LOW_WATERMARK_PCTG / 100 * recvpool_size);
if ((long)((float)(RECVPOOL_HIGH_WATERMARK_PCTG - RECVPOOL_LOW_WATERMARK_PCTG) / 100 * recvpool_size) >=
RECVPOOL_XON_TRIGGER_SIZE)
{ /* for large receive pools, the difference between high and low watermarks as computed above may be too large that
* we may not send XON quickly enough. Limit the difference to RECVPOOL_XON_TRIGGER_SIZE */
recvpool_low_watermark = recvpool_high_watermark - RECVPOOL_XON_TRIGGER_SIZE;
}
REPL_DPRINT4("RECVPOOL HIGH WATERMARK is %ld, LOW WATERMARK is %ld, Receive pool size is %ld\n",
recvpool_high_watermark, recvpool_low_watermark, recvpool_size);
gtmrecv_alloc_msgbuff();
gtmrecv_now = time(NULL);
heartbeat_period = GTMRECV_HEARTBEAT_PERIOD; /* time keeper, well sorta */
start_timer((TID)gtmrecv_heartbeat_timer, heartbeat_period * 1000, gtmrecv_heartbeat_timer, SIZEOF(heartbeat_period),
&heartbeat_period); /* start_timer expects time interval in milli seconds, heartbeat_period is in seconds */
do
{
gtmrecv_main_loop(crash_restart);
} while (repl_connection_reset);
assertpro(FALSE); /* shouldn't reach here */
return;
}
|