/ Artifact [c230a7a2]
Login

Artifact c230a7a24766320d8414afd087edcd43e499fb45e86361f6f4f464f343d965a9:


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 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825 5826 5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940 5941 5942 5943 5944 5945 5946 5947 5948 5949 5950 5951 5952 5953 5954 5955 5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021 6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342 6343 6344 6345 6346 6347 6348 6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359 6360 6361 6362 6363 6364 6365 6366 6367 6368 6369 6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470 6471 6472 6473 6474 6475 6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494 6495 6496 6497 6498 6499 6500 6501 6502 6503 6504 6505 6506 6507 6508 6509 6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526 6527 6528 6529 6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542 6543 6544 6545 6546 6547 6548 6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564 6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610 6611 6612 6613 6614 6615 6616 6617 6618 6619 6620 6621 6622 6623 6624 6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639 6640 6641 6642 6643 6644 6645 6646 6647 6648 6649 6650 6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667 6668 6669 6670 6671 6672 6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703 6704 6705 6706 6707 6708 6709 6710 6711 6712 6713 6714 6715 6716 6717 6718 6719 6720 6721 6722 6723 6724 6725 6726 6727 6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738 6739 6740 6741 6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781 6782 6783 6784 6785 6786 6787 6788 6789 6790 6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803 6804 6805 6806 6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819 6820 6821 6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859 6860 6861 6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907 6908 6909 6910 6911 6912 6913 6914 6915 6916 6917 6918 6919 6920 6921 6922 6923 6924 6925 6926 6927 6928 6929 6930 6931 6932 6933 6934 6935 6936 6937 6938 6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977 6978 6979 6980 6981 6982 6983 6984 6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995 6996 6997 6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011 7012 7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030 7031 7032 7033 7034 7035 7036 7037 7038 7039 7040 7041 7042 7043 7044 7045 7046 7047 7048 7049 7050 7051 7052 7053 7054 7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066 7067 7068 7069 7070 7071 7072 7073 7074 7075 7076 7077 7078 7079 7080 7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168 7169 7170 7171 7172 7173 7174 7175 7176 7177 7178 7179 7180 7181 7182 7183 7184 7185 7186 7187 7188 7189 7190 7191 7192 7193 7194 7195 7196 7197 7198 7199 7200 7201 7202 7203 7204 7205 7206 7207 7208 7209 7210 7211 7212 7213 7214 7215 7216 7217 7218 7219 7220 7221 7222 7223 7224 7225 7226 7227 7228 7229 7230 7231 7232 7233 7234 7235 7236 7237 7238 7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249 7250 7251 7252 7253 7254 7255 7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268 7269 7270 7271 7272 7273 7274 7275 7276 7277 7278 7279 7280 7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291 7292 7293 7294 7295 7296 7297 7298 7299 7300 7301 7302 7303 7304 7305 7306 7307 7308 7309 7310 7311 7312 7313 7314 7315 7316 7317 7318 7319 7320 7321 7322 7323 7324 7325 7326 7327 7328 7329 7330 7331 7332 7333 7334 7335 7336 7337 7338 7339 7340 7341 7342 7343 7344 7345 7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376 7377 7378 7379 7380 7381 7382 7383 7384 7385 7386 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398 7399 7400 7401 7402 7403 7404 7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415 7416 7417 7418 7419 7420 7421 7422 7423 7424 7425 7426 7427 7428 7429 7430 7431 7432 7433 7434 7435 7436 7437 7438 7439 7440 7441 7442 7443 7444 7445 7446 7447 7448 7449 7450 7451 7452 7453 7454 7455 7456 7457 7458 7459 7460 7461 7462 7463 7464 7465 7466 7467 7468 7469 7470 7471 7472 7473 7474 7475 7476 7477 7478 7479 7480 7481 7482 7483 7484 7485 7486 7487 7488 7489 7490 7491 7492 7493 7494 7495 7496 7497 7498 7499 7500 7501 7502 7503 7504 7505 7506 7507 7508 7509 7510 7511 7512 7513 7514 7515 7516 7517 7518 7519 7520 7521 7522 7523 7524 7525 7526 7527 7528 7529 7530 7531 7532 7533 7534 7535 7536 7537 7538 7539 7540 7541 7542 7543 7544 7545 7546 7547 7548 7549 7550 7551 7552 7553 7554 7555 7556 7557 7558 7559 7560 7561 7562 7563 7564 7565 7566 7567 7568 7569 7570 7571 7572 7573 7574 7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586 7587 7588 7589 7590 7591 7592 7593 7594 7595 7596 7597 7598 7599 7600 7601 7602 7603 7604 7605 7606 7607 7608 7609 7610 7611 7612 7613 7614 7615 7616 7617 7618 7619 7620 7621 7622 7623 7624 7625 7626 7627 7628 7629 7630 7631 7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642 7643 7644 7645 7646 7647 7648 7649 7650 7651 7652 7653 7654 7655 7656 7657 7658 7659 7660 7661 7662 7663 7664 7665 7666 7667 7668 7669 7670 7671 7672 7673 7674 7675 7676 7677 7678 7679 7680 7681 7682 7683 7684 7685 7686 7687 7688 7689 7690 7691 7692 7693 7694 7695 7696 7697 7698 7699 7700 7701 7702 7703 7704 7705 7706 7707 7708 7709 7710 7711 7712 7713 7714 7715 7716 7717 7718 7719 7720 7721 7722 7723 7724 7725 7726 7727 7728 7729 7730 7731 7732 7733 7734 7735 7736 7737 7738 7739 7740 7741 7742 7743 7744 7745 7746 7747 7748 7749 7750 7751 7752 7753 7754 7755 7756 7757 7758 7759 7760 7761 7762 7763 7764 7765 7766 7767 7768 7769 7770 7771 7772 7773 7774 7775 7776 7777 7778 7779 7780 7781 7782 7783 7784 7785 7786 7787 7788
/*
** 2004 May 22
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
******************************************************************************
**
** This file contains the VFS implementation for unix-like operating systems
** include Linux, MacOSX, *BSD, QNX, VxWorks, AIX, HPUX, and others.
**
** There are actually several different VFS implementations in this file.
** The differences are in the way that file locking is done.  The default
** implementation uses Posix Advisory Locks.  Alternative implementations
** use flock(), dot-files, various proprietary locking schemas, or simply
** skip locking all together.
**
** This source file is organized into divisions where the logic for various
** subfunctions is contained within the appropriate division.  PLEASE
** KEEP THE STRUCTURE OF THIS FILE INTACT.  New code should be placed
** in the correct division and should be clearly labeled.
**
** The layout of divisions is as follows:
**
**   *  General-purpose declarations and utility functions.
**   *  Unique file ID logic used by VxWorks.
**   *  Various locking primitive implementations (all except proxy locking):
**      + for Posix Advisory Locks
**      + for no-op locks
**      + for dot-file locks
**      + for flock() locking
**      + for named semaphore locks (VxWorks only)
**      + for AFP filesystem locks (MacOSX only)
**   *  sqlite3_file methods not associated with locking.
**   *  Definitions of sqlite3_io_methods objects for all locking
**      methods plus "finder" functions for each locking method.
**   *  sqlite3_vfs method implementations.
**   *  Locking primitives for the proxy uber-locking-method. (MacOSX only)
**   *  Definitions of sqlite3_vfs objects for all locking methods
**      plus implementations of sqlite3_os_init() and sqlite3_os_end().
*/
#include "sqliteInt.h"
#if SQLITE_OS_UNIX              /* This file is used on unix only */

/*
** There are various methods for file locking used for concurrency
** control:
**
**   1. POSIX locking (the default),
**   2. No locking,
**   3. Dot-file locking,
**   4. flock() locking,
**   5. AFP locking (OSX only),
**   6. Named POSIX semaphores (VXWorks only),
**   7. proxy locking. (OSX only)
**
** Styles 4, 5, and 7 are only available of SQLITE_ENABLE_LOCKING_STYLE
** is defined to 1.  The SQLITE_ENABLE_LOCKING_STYLE also enables automatic
** selection of the appropriate locking style based on the filesystem
** where the database is located.  
*/
#if !defined(SQLITE_ENABLE_LOCKING_STYLE)
#  if defined(__APPLE__)
#    define SQLITE_ENABLE_LOCKING_STYLE 1
#  else
#    define SQLITE_ENABLE_LOCKING_STYLE 0
#  endif
#endif

/* Use pread() and pwrite() if they are available */
#if defined(__APPLE__)
# define HAVE_PREAD 1
# define HAVE_PWRITE 1
#endif
#if defined(HAVE_PREAD64) && defined(HAVE_PWRITE64)
# undef USE_PREAD
# define USE_PREAD64 1
#elif defined(HAVE_PREAD) && defined(HAVE_PWRITE)
# undef USE_PREAD64
# define USE_PREAD 1
#endif

/*
** standard include files.
*/
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <unistd.h>
#include <time.h>
#include <sys/time.h>
#include <errno.h>
#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0
# include <sys/mman.h>
#endif

#if SQLITE_ENABLE_LOCKING_STYLE
# include <sys/ioctl.h>
# include <sys/file.h>
# include <sys/param.h>
#endif /* SQLITE_ENABLE_LOCKING_STYLE */

#if defined(__APPLE__) && ((__MAC_OS_X_VERSION_MIN_REQUIRED > 1050) || \
                           (__IPHONE_OS_VERSION_MIN_REQUIRED > 2000))
#  if (!defined(TARGET_OS_EMBEDDED) || (TARGET_OS_EMBEDDED==0)) \
       && (!defined(TARGET_IPHONE_SIMULATOR) || (TARGET_IPHONE_SIMULATOR==0))
#    define HAVE_GETHOSTUUID 1
#  else
#    warning "gethostuuid() is disabled."
#  endif
#endif


#if OS_VXWORKS
# include <sys/ioctl.h>
# include <semaphore.h>
# include <limits.h>
#endif /* OS_VXWORKS */

#if defined(__APPLE__) || SQLITE_ENABLE_LOCKING_STYLE
# include <sys/mount.h>
#endif

#ifdef HAVE_UTIME
# include <utime.h>
#endif

/*
** Allowed values of unixFile.fsFlags
*/
#define SQLITE_FSFLAGS_IS_MSDOS     0x1

/*
** If we are to be thread-safe, include the pthreads header and define
** the SQLITE_UNIX_THREADS macro.
*/
#if SQLITE_THREADSAFE
# include <pthread.h>
# define SQLITE_UNIX_THREADS 1
#endif

/*
** Default permissions when creating a new file
*/
#ifndef SQLITE_DEFAULT_FILE_PERMISSIONS
# define SQLITE_DEFAULT_FILE_PERMISSIONS 0644
#endif

/*
** Default permissions when creating auto proxy dir
*/
#ifndef SQLITE_DEFAULT_PROXYDIR_PERMISSIONS
# define SQLITE_DEFAULT_PROXYDIR_PERMISSIONS 0755
#endif

/*
** Maximum supported path-length.
*/
#define MAX_PATHNAME 512

/*
** Maximum supported symbolic links
*/
#define SQLITE_MAX_SYMLINKS 100

/* Always cast the getpid() return type for compatibility with
** kernel modules in VxWorks. */
#define osGetpid(X) (pid_t)getpid()

/*
** Only set the lastErrno if the error code is a real error and not 
** a normal expected return code of SQLITE_BUSY or SQLITE_OK
*/
#define IS_LOCK_ERROR(x)  ((x != SQLITE_OK) && (x != SQLITE_BUSY))

/* Forward references */
typedef struct unixShm unixShm;               /* Connection shared memory */
typedef struct unixShmNode unixShmNode;       /* Shared memory instance */
typedef struct unixInodeInfo unixInodeInfo;   /* An i-node */
typedef struct UnixUnusedFd UnixUnusedFd;     /* An unused file descriptor */

/*
** Sometimes, after a file handle is closed by SQLite, the file descriptor
** cannot be closed immediately. In these cases, instances of the following
** structure are used to store the file descriptor while waiting for an
** opportunity to either close or reuse it.
*/
struct UnixUnusedFd {
  int fd;                   /* File descriptor to close */
  int flags;                /* Flags this file descriptor was opened with */
  UnixUnusedFd *pNext;      /* Next unused file descriptor on same file */
};

/*
** The unixFile structure is subclass of sqlite3_file specific to the unix
** VFS implementations.
*/
typedef struct unixFile unixFile;
struct unixFile {
  sqlite3_io_methods const *pMethod;  /* Always the first entry */
  sqlite3_vfs *pVfs;                  /* The VFS that created this unixFile */
  unixInodeInfo *pInode;              /* Info about locks on this inode */
  int h;                              /* The file descriptor */
  unsigned char eFileLock;            /* The type of lock held on this fd */
  unsigned short int ctrlFlags;       /* Behavioral bits.  UNIXFILE_* flags */
  int lastErrno;                      /* The unix errno from last I/O error */
  void *lockingContext;               /* Locking style specific state */
  UnixUnusedFd *pPreallocatedUnused;  /* Pre-allocated UnixUnusedFd */
  const char *zPath;                  /* Name of the file */
  unixShm *pShm;                      /* Shared memory segment information */
  int szChunk;                        /* Configured by FCNTL_CHUNK_SIZE */
#if SQLITE_MAX_MMAP_SIZE>0
  int nFetchOut;                      /* Number of outstanding xFetch refs */
  sqlite3_int64 mmapSize;             /* Usable size of mapping at pMapRegion */
  sqlite3_int64 mmapSizeActual;       /* Actual size of mapping at pMapRegion */
  sqlite3_int64 mmapSizeMax;          /* Configured FCNTL_MMAP_SIZE value */
  void *pMapRegion;                   /* Memory mapped region */
#endif
  int sectorSize;                     /* Device sector size */
  int deviceCharacteristics;          /* Precomputed device characteristics */
#if SQLITE_ENABLE_LOCKING_STYLE
  int openFlags;                      /* The flags specified at open() */
#endif
#if SQLITE_ENABLE_LOCKING_STYLE || defined(__APPLE__)
  unsigned fsFlags;                   /* cached details from statfs() */
#endif
#ifdef SQLITE_ENABLE_SETLK_TIMEOUT
  unsigned iBusyTimeout;              /* Wait this many millisec on locks */
#endif
#if OS_VXWORKS
  struct vxworksFileId *pId;          /* Unique file ID */
#endif
#ifdef SQLITE_DEBUG
  /* The next group of variables are used to track whether or not the
  ** transaction counter in bytes 24-27 of database files are updated
  ** whenever any part of the database changes.  An assertion fault will
  ** occur if a file is updated without also updating the transaction
  ** counter.  This test is made to avoid new problems similar to the
  ** one described by ticket #3584. 
  */
  unsigned char transCntrChng;   /* True if the transaction counter changed */
  unsigned char dbUpdate;        /* True if any part of database file changed */
  unsigned char inNormalWrite;   /* True if in a normal write operation */

#endif

#ifdef SQLITE_TEST
  /* In test mode, increase the size of this structure a bit so that 
  ** it is larger than the struct CrashFile defined in test6.c.
  */
  char aPadding[32];
#endif
};

/* This variable holds the process id (pid) from when the xRandomness()
** method was called.  If xOpen() is called from a different process id,
** indicating that a fork() has occurred, the PRNG will be reset.
*/
static pid_t randomnessPid = 0;

/*
** Allowed values for the unixFile.ctrlFlags bitmask:
*/
#define UNIXFILE_EXCL        0x01     /* Connections from one process only */
#define UNIXFILE_RDONLY      0x02     /* Connection is read only */
#define UNIXFILE_PERSIST_WAL 0x04     /* Persistent WAL mode */
#ifndef SQLITE_DISABLE_DIRSYNC
# define UNIXFILE_DIRSYNC    0x08     /* Directory sync needed */
#else
# define UNIXFILE_DIRSYNC    0x00
#endif
#define UNIXFILE_PSOW        0x10     /* SQLITE_IOCAP_POWERSAFE_OVERWRITE */
#define UNIXFILE_DELETE      0x20     /* Delete on close */
#define UNIXFILE_URI         0x40     /* Filename might have query parameters */
#define UNIXFILE_NOLOCK      0x80     /* Do no file locking */

/*
** Include code that is common to all os_*.c files
*/
#include "os_common.h"

/*
** Define various macros that are missing from some systems.
*/
#ifndef O_LARGEFILE
# define O_LARGEFILE 0
#endif
#ifdef SQLITE_DISABLE_LFS
# undef O_LARGEFILE
# define O_LARGEFILE 0
#endif
#ifndef O_NOFOLLOW
# define O_NOFOLLOW 0
#endif
#ifndef O_BINARY
# define O_BINARY 0
#endif

/*
** The threadid macro resolves to the thread-id or to 0.  Used for
** testing and debugging only.
*/
#if SQLITE_THREADSAFE
#define threadid pthread_self()
#else
#define threadid 0
#endif

/*
** HAVE_MREMAP defaults to true on Linux and false everywhere else.
*/
#if !defined(HAVE_MREMAP)
# if defined(__linux__) && defined(_GNU_SOURCE)
#  define HAVE_MREMAP 1
# else
#  define HAVE_MREMAP 0
# endif
#endif

/*
** Explicitly call the 64-bit version of lseek() on Android. Otherwise, lseek()
** is the 32-bit version, even if _FILE_OFFSET_BITS=64 is defined.
*/
#ifdef __ANDROID__
# define lseek lseek64
#endif

#ifdef __linux__
/*
** Linux-specific IOCTL magic numbers used for controlling F2FS
*/
#define F2FS_IOCTL_MAGIC        0xf5
#define F2FS_IOC_START_ATOMIC_WRITE     _IO(F2FS_IOCTL_MAGIC, 1)
#define F2FS_IOC_COMMIT_ATOMIC_WRITE    _IO(F2FS_IOCTL_MAGIC, 2)
#define F2FS_IOC_START_VOLATILE_WRITE   _IO(F2FS_IOCTL_MAGIC, 3)
#define F2FS_IOC_ABORT_VOLATILE_WRITE   _IO(F2FS_IOCTL_MAGIC, 5)
#define F2FS_IOC_GET_FEATURES           _IOR(F2FS_IOCTL_MAGIC, 12, u32)
#define F2FS_FEATURE_ATOMIC_WRITE 0x0004
#endif /* __linux__ */


/*
** Different Unix systems declare open() in different ways.  Same use
** open(const char*,int,mode_t).  Others use open(const char*,int,...).
** The difference is important when using a pointer to the function.
**
** The safest way to deal with the problem is to always use this wrapper
** which always has the same well-defined interface.
*/
static int posixOpen(const char *zFile, int flags, int mode){
  return open(zFile, flags, mode);
}

/* Forward reference */
static int openDirectory(const char*, int*);
static int unixGetpagesize(void);

/*
** Many system calls are accessed through pointer-to-functions so that
** they may be overridden at runtime to facilitate fault injection during
** testing and sandboxing.  The following array holds the names and pointers
** to all overrideable system calls.
*/
static struct unix_syscall {
  const char *zName;            /* Name of the system call */
  sqlite3_syscall_ptr pCurrent; /* Current value of the system call */
  sqlite3_syscall_ptr pDefault; /* Default value */
} aSyscall[] = {
  { "open",         (sqlite3_syscall_ptr)posixOpen,  0  },
#define osOpen      ((int(*)(const char*,int,int))aSyscall[0].pCurrent)

  { "close",        (sqlite3_syscall_ptr)close,      0  },
#define osClose     ((int(*)(int))aSyscall[1].pCurrent)

  { "access",       (sqlite3_syscall_ptr)access,     0  },
#define osAccess    ((int(*)(const char*,int))aSyscall[2].pCurrent)

  { "getcwd",       (sqlite3_syscall_ptr)getcwd,     0  },
#define osGetcwd    ((char*(*)(char*,size_t))aSyscall[3].pCurrent)

  { "stat",         (sqlite3_syscall_ptr)stat,       0  },
#define osStat      ((int(*)(const char*,struct stat*))aSyscall[4].pCurrent)

/*
** The DJGPP compiler environment looks mostly like Unix, but it
** lacks the fcntl() system call.  So redefine fcntl() to be something
** that always succeeds.  This means that locking does not occur under
** DJGPP.  But it is DOS - what did you expect?
*/
#ifdef __DJGPP__
  { "fstat",        0,                 0  },
#define osFstat(a,b,c)    0
#else     
  { "fstat",        (sqlite3_syscall_ptr)fstat,      0  },
#define osFstat     ((int(*)(int,struct stat*))aSyscall[5].pCurrent)
#endif

  { "ftruncate",    (sqlite3_syscall_ptr)ftruncate,  0  },
#define osFtruncate ((int(*)(int,off_t))aSyscall[6].pCurrent)

  { "fcntl",        (sqlite3_syscall_ptr)fcntl,      0  },
#define osFcntl     ((int(*)(int,int,...))aSyscall[7].pCurrent)

  { "read",         (sqlite3_syscall_ptr)read,       0  },
#define osRead      ((ssize_t(*)(int,void*,size_t))aSyscall[8].pCurrent)

#if defined(USE_PREAD) || SQLITE_ENABLE_LOCKING_STYLE
  { "pread",        (sqlite3_syscall_ptr)pread,      0  },
#else
  { "pread",        (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPread     ((ssize_t(*)(int,void*,size_t,off_t))aSyscall[9].pCurrent)

#if defined(USE_PREAD64)
  { "pread64",      (sqlite3_syscall_ptr)pread64,    0  },
#else
  { "pread64",      (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPread64 ((ssize_t(*)(int,void*,size_t,off64_t))aSyscall[10].pCurrent)

  { "write",        (sqlite3_syscall_ptr)write,      0  },
#define osWrite     ((ssize_t(*)(int,const void*,size_t))aSyscall[11].pCurrent)

#if defined(USE_PREAD) || SQLITE_ENABLE_LOCKING_STYLE
  { "pwrite",       (sqlite3_syscall_ptr)pwrite,     0  },
#else
  { "pwrite",       (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPwrite    ((ssize_t(*)(int,const void*,size_t,off_t))\
                    aSyscall[12].pCurrent)

#if defined(USE_PREAD64)
  { "pwrite64",     (sqlite3_syscall_ptr)pwrite64,   0  },
#else
  { "pwrite64",     (sqlite3_syscall_ptr)0,          0  },
#endif
#define osPwrite64  ((ssize_t(*)(int,const void*,size_t,off64_t))\
                    aSyscall[13].pCurrent)

  { "fchmod",       (sqlite3_syscall_ptr)fchmod,          0  },
#define osFchmod    ((int(*)(int,mode_t))aSyscall[14].pCurrent)

#if defined(HAVE_POSIX_FALLOCATE) && HAVE_POSIX_FALLOCATE
  { "fallocate",    (sqlite3_syscall_ptr)posix_fallocate,  0 },
#else
  { "fallocate",    (sqlite3_syscall_ptr)0,                0 },
#endif
#define osFallocate ((int(*)(int,off_t,off_t))aSyscall[15].pCurrent)

  { "unlink",       (sqlite3_syscall_ptr)unlink,           0 },
#define osUnlink    ((int(*)(const char*))aSyscall[16].pCurrent)

  { "openDirectory",    (sqlite3_syscall_ptr)openDirectory,      0 },
#define osOpenDirectory ((int(*)(const char*,int*))aSyscall[17].pCurrent)

  { "mkdir",        (sqlite3_syscall_ptr)mkdir,           0 },
#define osMkdir     ((int(*)(const char*,mode_t))aSyscall[18].pCurrent)

  { "rmdir",        (sqlite3_syscall_ptr)rmdir,           0 },
#define osRmdir     ((int(*)(const char*))aSyscall[19].pCurrent)

#if defined(HAVE_FCHOWN)
  { "fchown",       (sqlite3_syscall_ptr)fchown,          0 },
#else
  { "fchown",       (sqlite3_syscall_ptr)0,               0 },
#endif
#define osFchown    ((int(*)(int,uid_t,gid_t))aSyscall[20].pCurrent)

#if defined(HAVE_FCHOWN)
  { "geteuid",      (sqlite3_syscall_ptr)geteuid,         0 },
#else
  { "geteuid",      (sqlite3_syscall_ptr)0,               0 },
#endif
#define osGeteuid   ((uid_t(*)(void))aSyscall[21].pCurrent)

#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0
  { "mmap",         (sqlite3_syscall_ptr)mmap,            0 },
#else
  { "mmap",         (sqlite3_syscall_ptr)0,               0 },
#endif
#define osMmap ((void*(*)(void*,size_t,int,int,int,off_t))aSyscall[22].pCurrent)

#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0
  { "munmap",       (sqlite3_syscall_ptr)munmap,          0 },
#else
  { "munmap",       (sqlite3_syscall_ptr)0,               0 },
#endif
#define osMunmap ((int(*)(void*,size_t))aSyscall[23].pCurrent)

#if HAVE_MREMAP && (!defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0)
  { "mremap",       (sqlite3_syscall_ptr)mremap,          0 },
#else
  { "mremap",       (sqlite3_syscall_ptr)0,               0 },
#endif
#define osMremap ((void*(*)(void*,size_t,size_t,int,...))aSyscall[24].pCurrent)

#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0
  { "getpagesize",  (sqlite3_syscall_ptr)unixGetpagesize, 0 },
#else
  { "getpagesize",  (sqlite3_syscall_ptr)0,               0 },
#endif
#define osGetpagesize ((int(*)(void))aSyscall[25].pCurrent)

#if defined(HAVE_READLINK)
  { "readlink",     (sqlite3_syscall_ptr)readlink,        0 },
#else
  { "readlink",     (sqlite3_syscall_ptr)0,               0 },
#endif
#define osReadlink ((ssize_t(*)(const char*,char*,size_t))aSyscall[26].pCurrent)

#if defined(HAVE_LSTAT)
  { "lstat",         (sqlite3_syscall_ptr)lstat,          0 },
#else
  { "lstat",         (sqlite3_syscall_ptr)0,              0 },
#endif
#define osLstat      ((int(*)(const char*,struct stat*))aSyscall[27].pCurrent)

#if defined(__linux__) && defined(SQLITE_ENABLE_BATCH_ATOMIC_WRITE)
  { "ioctl",         (sqlite3_syscall_ptr)ioctl,          0 },
#else
  { "ioctl",         (sqlite3_syscall_ptr)0,              0 },
#endif
#define osIoctl ((int(*)(int,int,...))aSyscall[28].pCurrent)

}; /* End of the overrideable system calls */


/*
** On some systems, calls to fchown() will trigger a message in a security
** log if they come from non-root processes.  So avoid calling fchown() if
** we are not running as root.
*/
static int robustFchown(int fd, uid_t uid, gid_t gid){
#if defined(HAVE_FCHOWN)
  return osGeteuid() ? 0 : osFchown(fd,uid,gid);
#else
  return 0;
#endif
}

/*
** This is the xSetSystemCall() method of sqlite3_vfs for all of the
** "unix" VFSes.  Return SQLITE_OK opon successfully updating the
** system call pointer, or SQLITE_NOTFOUND if there is no configurable
** system call named zName.
*/
static int unixSetSystemCall(
  sqlite3_vfs *pNotUsed,        /* The VFS pointer.  Not used */
  const char *zName,            /* Name of system call to override */
  sqlite3_syscall_ptr pNewFunc  /* Pointer to new system call value */
){
  unsigned int i;
  int rc = SQLITE_NOTFOUND;

  UNUSED_PARAMETER(pNotUsed);
  if( zName==0 ){
    /* If no zName is given, restore all system calls to their default
    ** settings and return NULL
    */
    rc = SQLITE_OK;
    for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
      if( aSyscall[i].pDefault ){
        aSyscall[i].pCurrent = aSyscall[i].pDefault;
      }
    }
  }else{
    /* If zName is specified, operate on only the one system call
    ** specified.
    */
    for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
      if( strcmp(zName, aSyscall[i].zName)==0 ){
        if( aSyscall[i].pDefault==0 ){
          aSyscall[i].pDefault = aSyscall[i].pCurrent;
        }
        rc = SQLITE_OK;
        if( pNewFunc==0 ) pNewFunc = aSyscall[i].pDefault;
        aSyscall[i].pCurrent = pNewFunc;
        break;
      }
    }
  }
  return rc;
}

/*
** Return the value of a system call.  Return NULL if zName is not a
** recognized system call name.  NULL is also returned if the system call
** is currently undefined.
*/
static sqlite3_syscall_ptr unixGetSystemCall(
  sqlite3_vfs *pNotUsed,
  const char *zName
){
  unsigned int i;

  UNUSED_PARAMETER(pNotUsed);
  for(i=0; i<sizeof(aSyscall)/sizeof(aSyscall[0]); i++){
    if( strcmp(zName, aSyscall[i].zName)==0 ) return aSyscall[i].pCurrent;
  }
  return 0;
}

/*
** Return the name of the first system call after zName.  If zName==NULL
** then return the name of the first system call.  Return NULL if zName
** is the last system call or if zName is not the name of a valid
** system call.
*/
static const char *unixNextSystemCall(sqlite3_vfs *p, const char *zName){
  int i = -1;

  UNUSED_PARAMETER(p);
  if( zName ){
    for(i=0; i<ArraySize(aSyscall)-1; i++){
      if( strcmp(zName, aSyscall[i].zName)==0 ) break;
    }
  }
  for(i++; i<ArraySize(aSyscall); i++){
    if( aSyscall[i].pCurrent!=0 ) return aSyscall[i].zName;
  }
  return 0;
}

/*
** Do not accept any file descriptor less than this value, in order to avoid
** opening database file using file descriptors that are commonly used for 
** standard input, output, and error.
*/
#ifndef SQLITE_MINIMUM_FILE_DESCRIPTOR
# define SQLITE_MINIMUM_FILE_DESCRIPTOR 3
#endif

/*
** Invoke open().  Do so multiple times, until it either succeeds or
** fails for some reason other than EINTR.
**
** If the file creation mode "m" is 0 then set it to the default for
** SQLite.  The default is SQLITE_DEFAULT_FILE_PERMISSIONS (normally
** 0644) as modified by the system umask.  If m is not 0, then
** make the file creation mode be exactly m ignoring the umask.
**
** The m parameter will be non-zero only when creating -wal, -journal,
** and -shm files.  We want those files to have *exactly* the same
** permissions as their original database, unadulterated by the umask.
** In that way, if a database file is -rw-rw-rw or -rw-rw-r-, and a
** transaction crashes and leaves behind hot journals, then any
** process that is able to write to the database will also be able to
** recover the hot journals.
*/
static int robust_open(const char *z, int f, mode_t m){
  int fd;
  mode_t m2 = m ? m : SQLITE_DEFAULT_FILE_PERMISSIONS;
  while(1){
#if defined(O_CLOEXEC)
    fd = osOpen(z,f|O_CLOEXEC,m2);
#else
    fd = osOpen(z,f,m2);
#endif
    if( fd<0 ){
      if( errno==EINTR ) continue;
      break;
    }
    if( fd>=SQLITE_MINIMUM_FILE_DESCRIPTOR ) break;
    osClose(fd);
    sqlite3_log(SQLITE_WARNING, 
                "attempt to open \"%s\" as file descriptor %d", z, fd);
    fd = -1;
    if( osOpen("/dev/null", f, m)<0 ) break;
  }
  if( fd>=0 ){
    if( m!=0 ){
      struct stat statbuf;
      if( osFstat(fd, &statbuf)==0 
       && statbuf.st_size==0
       && (statbuf.st_mode&0777)!=m 
      ){
        osFchmod(fd, m);
      }
    }
#if defined(FD_CLOEXEC) && (!defined(O_CLOEXEC) || O_CLOEXEC==0)
    osFcntl(fd, F_SETFD, osFcntl(fd, F_GETFD, 0) | FD_CLOEXEC);
#endif
  }
  return fd;
}

/*
** Helper functions to obtain and relinquish the global mutex. The
** global mutex is used to protect the unixInodeInfo and
** vxworksFileId objects used by this file, all of which may be 
** shared by multiple threads.
**
** Function unixMutexHeld() is used to assert() that the global mutex 
** is held when required. This function is only used as part of assert() 
** statements. e.g.
**
**   unixEnterMutex()
**     assert( unixMutexHeld() );
**   unixEnterLeave()
*/
static sqlite3_mutex *unixBigLock = 0;
static void unixEnterMutex(void){
  sqlite3_mutex_enter(unixBigLock);
}
static void unixLeaveMutex(void){
  sqlite3_mutex_leave(unixBigLock);
}
#ifdef SQLITE_DEBUG
static int unixMutexHeld(void) {
  return sqlite3_mutex_held(unixBigLock);
}
#endif


#ifdef SQLITE_HAVE_OS_TRACE
/*
** Helper function for printing out trace information from debugging
** binaries. This returns the string representation of the supplied
** integer lock-type.
*/
static const char *azFileLock(int eFileLock){
  switch( eFileLock ){
    case NO_LOCK: return "NONE";
    case SHARED_LOCK: return "SHARED";
    case RESERVED_LOCK: return "RESERVED";
    case PENDING_LOCK: return "PENDING";
    case EXCLUSIVE_LOCK: return "EXCLUSIVE";
  }
  return "ERROR";
}
#endif

#ifdef SQLITE_LOCK_TRACE
/*
** Print out information about all locking operations.
**
** This routine is used for troubleshooting locks on multithreaded
** platforms.  Enable by compiling with the -DSQLITE_LOCK_TRACE
** command-line option on the compiler.  This code is normally
** turned off.
*/
static int lockTrace(int fd, int op, struct flock *p){
  char *zOpName, *zType;
  int s;
  int savedErrno;
  if( op==F_GETLK ){
    zOpName = "GETLK";
  }else if( op==F_SETLK ){
    zOpName = "SETLK";
  }else{
    s = osFcntl(fd, op, p);
    sqlite3DebugPrintf("fcntl unknown %d %d %d\n", fd, op, s);
    return s;
  }
  if( p->l_type==F_RDLCK ){
    zType = "RDLCK";
  }else if( p->l_type==F_WRLCK ){
    zType = "WRLCK";
  }else if( p->l_type==F_UNLCK ){
    zType = "UNLCK";
  }else{
    assert( 0 );
  }
  assert( p->l_whence==SEEK_SET );
  s = osFcntl(fd, op, p);
  savedErrno = errno;
  sqlite3DebugPrintf("fcntl %d %d %s %s %d %d %d %d\n",
     threadid, fd, zOpName, zType, (int)p->l_start, (int)p->l_len,
     (int)p->l_pid, s);
  if( s==(-1) && op==F_SETLK && (p->l_type==F_RDLCK || p->l_type==F_WRLCK) ){
    struct flock l2;
    l2 = *p;
    osFcntl(fd, F_GETLK, &l2);
    if( l2.l_type==F_RDLCK ){
      zType = "RDLCK";
    }else if( l2.l_type==F_WRLCK ){
      zType = "WRLCK";
    }else if( l2.l_type==F_UNLCK ){
      zType = "UNLCK";
    }else{
      assert( 0 );
    }
    sqlite3DebugPrintf("fcntl-failure-reason: %s %d %d %d\n",
       zType, (int)l2.l_start, (int)l2.l_len, (int)l2.l_pid);
  }
  errno = savedErrno;
  return s;
}
#undef osFcntl
#define osFcntl lockTrace
#endif /* SQLITE_LOCK_TRACE */

/*
** Retry ftruncate() calls that fail due to EINTR
**
** All calls to ftruncate() within this file should be made through
** this wrapper.  On the Android platform, bypassing the logic below
** could lead to a corrupt database.
*/
static int robust_ftruncate(int h, sqlite3_int64 sz){
  int rc;
#ifdef __ANDROID__
  /* On Android, ftruncate() always uses 32-bit offsets, even if 
  ** _FILE_OFFSET_BITS=64 is defined. This means it is unsafe to attempt to
  ** truncate a file to any size larger than 2GiB. Silently ignore any
  ** such attempts.  */
  if( sz>(sqlite3_int64)0x7FFFFFFF ){
    rc = SQLITE_OK;
  }else
#endif
  do{ rc = osFtruncate(h,sz); }while( rc<0 && errno==EINTR );
  return rc;
}

/*
** This routine translates a standard POSIX errno code into something
** useful to the clients of the sqlite3 functions.  Specifically, it is
** intended to translate a variety of "try again" errors into SQLITE_BUSY
** and a variety of "please close the file descriptor NOW" errors into 
** SQLITE_IOERR
** 
** Errors during initialization of locks, or file system support for locks,
** should handle ENOLCK, ENOTSUP, EOPNOTSUPP separately.
*/
static int sqliteErrorFromPosixError(int posixError, int sqliteIOErr) {
  assert( (sqliteIOErr == SQLITE_IOERR_LOCK) || 
          (sqliteIOErr == SQLITE_IOERR_UNLOCK) || 
          (sqliteIOErr == SQLITE_IOERR_RDLOCK) ||
          (sqliteIOErr == SQLITE_IOERR_CHECKRESERVEDLOCK) );
  switch (posixError) {
  case EACCES: 
  case EAGAIN:
  case ETIMEDOUT:
  case EBUSY:
  case EINTR:
  case ENOLCK:  
    /* random NFS retry error, unless during file system support 
     * introspection, in which it actually means what it says */
    return SQLITE_BUSY;
    
  case EPERM: 
    return SQLITE_PERM;
    
  default: 
    return sqliteIOErr;
  }
}


/******************************************************************************
****************** Begin Unique File ID Utility Used By VxWorks ***************
**
** On most versions of unix, we can get a unique ID for a file by concatenating
** the device number and the inode number.  But this does not work on VxWorks.
** On VxWorks, a unique file id must be based on the canonical filename.
**
** A pointer to an instance of the following structure can be used as a
** unique file ID in VxWorks.  Each instance of this structure contains
** a copy of the canonical filename.  There is also a reference count.  
** The structure is reclaimed when the number of pointers to it drops to
** zero.
**
** There are never very many files open at one time and lookups are not
** a performance-critical path, so it is sufficient to put these
** structures on a linked list.
*/
struct vxworksFileId {
  struct vxworksFileId *pNext;  /* Next in a list of them all */
  int nRef;                     /* Number of references to this one */
  int nName;                    /* Length of the zCanonicalName[] string */
  char *zCanonicalName;         /* Canonical filename */
};

#if OS_VXWORKS
/* 
** All unique filenames are held on a linked list headed by this
** variable:
*/
static struct vxworksFileId *vxworksFileList = 0;

/*
** Simplify a filename into its canonical form
** by making the following changes:
**
**  * removing any trailing and duplicate /
**  * convert /./ into just /
**  * convert /A/../ where A is any simple name into just /
**
** Changes are made in-place.  Return the new name length.
**
** The original filename is in z[0..n-1].  Return the number of
** characters in the simplified name.
*/
static int vxworksSimplifyName(char *z, int n){
  int i, j;
  while( n>1 && z[n-1]=='/' ){ n--; }
  for(i=j=0; i<n; i++){
    if( z[i]=='/' ){
      if( z[i+1]=='/' ) continue;
      if( z[i+1]=='.' && i+2<n && z[i+2]=='/' ){
        i += 1;
        continue;
      }
      if( z[i+1]=='.' && i+3<n && z[i+2]=='.' && z[i+3]=='/' ){
        while( j>0 && z[j-1]!='/' ){ j--; }
        if( j>0 ){ j--; }
        i += 2;
        continue;
      }
    }
    z[j++] = z[i];
  }
  z[j] = 0;
  return j;
}

/*
** Find a unique file ID for the given absolute pathname.  Return
** a pointer to the vxworksFileId object.  This pointer is the unique
** file ID.
**
** The nRef field of the vxworksFileId object is incremented before
** the object is returned.  A new vxworksFileId object is created
** and added to the global list if necessary.
**
** If a memory allocation error occurs, return NULL.
*/
static struct vxworksFileId *vxworksFindFileId(const char *zAbsoluteName){
  struct vxworksFileId *pNew;         /* search key and new file ID */
  struct vxworksFileId *pCandidate;   /* For looping over existing file IDs */
  int n;                              /* Length of zAbsoluteName string */

  assert( zAbsoluteName[0]=='/' );
  n = (int)strlen(zAbsoluteName);
  pNew = sqlite3_malloc64( sizeof(*pNew) + (n+1) );
  if( pNew==0 ) return 0;
  pNew->zCanonicalName = (char*)&pNew[1];
  memcpy(pNew->zCanonicalName, zAbsoluteName, n+1);
  n = vxworksSimplifyName(pNew->zCanonicalName, n);

  /* Search for an existing entry that matching the canonical name.
  ** If found, increment the reference count and return a pointer to
  ** the existing file ID.
  */
  unixEnterMutex();
  for(pCandidate=vxworksFileList; pCandidate; pCandidate=pCandidate->pNext){
    if( pCandidate->nName==n 
     && memcmp(pCandidate->zCanonicalName, pNew->zCanonicalName, n)==0
    ){
       sqlite3_free(pNew);
       pCandidate->nRef++;
       unixLeaveMutex();
       return pCandidate;
    }
  }

  /* No match was found.  We will make a new file ID */
  pNew->nRef = 1;
  pNew->nName = n;
  pNew->pNext = vxworksFileList;
  vxworksFileList = pNew;
  unixLeaveMutex();
  return pNew;
}

/*
** Decrement the reference count on a vxworksFileId object.  Free
** the object when the reference count reaches zero.
*/
static void vxworksReleaseFileId(struct vxworksFileId *pId){
  unixEnterMutex();
  assert( pId->nRef>0 );
  pId->nRef--;
  if( pId->nRef==0 ){
    struct vxworksFileId **pp;
    for(pp=&vxworksFileList; *pp && *pp!=pId; pp = &((*pp)->pNext)){}
    assert( *pp==pId );
    *pp = pId->pNext;
    sqlite3_free(pId);
  }
  unixLeaveMutex();
}
#endif /* OS_VXWORKS */
/*************** End of Unique File ID Utility Used By VxWorks ****************
******************************************************************************/


/******************************************************************************
*************************** Posix Advisory Locking ****************************
**
** POSIX advisory locks are broken by design.  ANSI STD 1003.1 (1996)
** section 6.5.2.2 lines 483 through 490 specify that when a process
** sets or clears a lock, that operation overrides any prior locks set
** by the same process.  It does not explicitly say so, but this implies
** that it overrides locks set by the same process using a different
** file descriptor.  Consider this test case:
**
**       int fd1 = open("./file1", O_RDWR|O_CREAT, 0644);
**       int fd2 = open("./file2", O_RDWR|O_CREAT, 0644);
**
** Suppose ./file1 and ./file2 are really the same file (because
** one is a hard or symbolic link to the other) then if you set
** an exclusive lock on fd1, then try to get an exclusive lock
** on fd2, it works.  I would have expected the second lock to
** fail since there was already a lock on the file due to fd1.
** But not so.  Since both locks came from the same process, the
** second overrides the first, even though they were on different
** file descriptors opened on different file names.
**
** This means that we cannot use POSIX locks to synchronize file access
** among competing threads of the same process.  POSIX locks will work fine
** to synchronize access for threads in separate processes, but not
** threads within the same process.
**
** To work around the problem, SQLite has to manage file locks internally
** on its own.  Whenever a new database is opened, we have to find the
** specific inode of the database file (the inode is determined by the
** st_dev and st_ino fields of the stat structure that fstat() fills in)
** and check for locks already existing on that inode.  When locks are
** created or removed, we have to look at our own internal record of the
** locks to see if another thread has previously set a lock on that same
** inode.
**
** (Aside: The use of inode numbers as unique IDs does not work on VxWorks.
** For VxWorks, we have to use the alternative unique ID system based on
** canonical filename and implemented in the previous division.)
**
** The sqlite3_file structure for POSIX is no longer just an integer file
** descriptor.  It is now a structure that holds the integer file
** descriptor and a pointer to a structure that describes the internal
** locks on the corresponding inode.  There is one locking structure
** per inode, so if the same inode is opened twice, both unixFile structures
** point to the same locking structure.  The locking structure keeps
** a reference count (so we will know when to delete it) and a "cnt"
** field that tells us its internal lock status.  cnt==0 means the
** file is unlocked.  cnt==-1 means the file has an exclusive lock.
** cnt>0 means there are cnt shared locks on the file.
**
** Any attempt to lock or unlock a file first checks the locking
** structure.  The fcntl() system call is only invoked to set a 
** POSIX lock if the internal lock structure transitions between
** a locked and an unlocked state.
**
** But wait:  there are yet more problems with POSIX advisory locks.
**
** If you close a file descriptor that points to a file that has locks,
** all locks on that file that are owned by the current process are
** released.  To work around this problem, each unixInodeInfo object
** maintains a count of the number of pending locks on tha inode.
** When an attempt is made to close an unixFile, if there are
** other unixFile open on the same inode that are holding locks, the call
** to close() the file descriptor is deferred until all of the locks clear.
** The unixInodeInfo structure keeps a list of file descriptors that need to
** be closed and that list is walked (and cleared) when the last lock
** clears.
**
** Yet another problem:  LinuxThreads do not play well with posix locks.
**
** Many older versions of linux use the LinuxThreads library which is
** not posix compliant.  Under LinuxThreads, a lock created by thread
** A cannot be modified or overridden by a different thread B.
** Only thread A can modify the lock.  Locking behavior is correct
** if the appliation uses the newer Native Posix Thread Library (NPTL)
** on linux - with NPTL a lock created by thread A can override locks
** in thread B.  But there is no way to know at compile-time which
** threading library is being used.  So there is no way to know at
** compile-time whether or not thread A can override locks on thread B.
** One has to do a run-time check to discover the behavior of the
** current process.
**
** SQLite used to support LinuxThreads.  But support for LinuxThreads
** was dropped beginning with version 3.7.0.  SQLite will still work with
** LinuxThreads provided that (1) there is no more than one connection 
** per database file in the same process and (2) database connections
** do not move across threads.
*/

/*
** An instance of the following structure serves as the key used
** to locate a particular unixInodeInfo object.
*/
struct unixFileId {
  dev_t dev;                  /* Device number */
#if OS_VXWORKS
  struct vxworksFileId *pId;  /* Unique file ID for vxworks. */
#else
  /* We are told that some versions of Android contain a bug that
  ** sizes ino_t at only 32-bits instead of 64-bits. (See
  ** https://android-review.googlesource.com/#/c/115351/3/dist/sqlite3.c)
  ** To work around this, always allocate 64-bits for the inode number.  
  ** On small machines that only have 32-bit inodes, this wastes 4 bytes,
  ** but that should not be a big deal. */
  /* WAS:  ino_t ino;   */
  u64 ino;                   /* Inode number */
#endif
};

/*
** An instance of the following structure is allocated for each open
** inode.  Or, on LinuxThreads, there is one of these structures for
** each inode opened by each thread.
**
** A single inode can have multiple file descriptors, so each unixFile
** structure contains a pointer to an instance of this object and this
** object keeps a count of the number of unixFile pointing to it.
*/
struct unixInodeInfo {
  struct unixFileId fileId;       /* The lookup key */
  int nShared;                    /* Number of SHARED locks held */
  unsigned char eFileLock;        /* One of SHARED_LOCK, RESERVED_LOCK etc. */
  unsigned char bProcessLock;     /* An exclusive process lock is held */
  int nRef;                       /* Number of pointers to this structure */
  unixShmNode *pShmNode;          /* Shared memory associated with this inode */
  int nLock;                      /* Number of outstanding file locks */
  UnixUnusedFd *pUnused;          /* Unused file descriptors to close */
  unixInodeInfo *pNext;           /* List of all unixInodeInfo objects */
  unixInodeInfo *pPrev;           /*    .... doubly linked */
#if SQLITE_ENABLE_LOCKING_STYLE
  unsigned long long sharedByte;  /* for AFP simulated shared lock */
#endif
#if OS_VXWORKS
  sem_t *pSem;                    /* Named POSIX semaphore */
  char aSemName[MAX_PATHNAME+2];  /* Name of that semaphore */
#endif
};

/*
** A lists of all unixInodeInfo objects.
*/
static unixInodeInfo *inodeList = 0;  /* All unixInodeInfo objects */
static unsigned int nUnusedFd = 0;    /* Total unused file descriptors */

/*
**
** This function - unixLogErrorAtLine(), is only ever called via the macro
** unixLogError().
**
** It is invoked after an error occurs in an OS function and errno has been
** set. It logs a message using sqlite3_log() containing the current value of
** errno and, if possible, the human-readable equivalent from strerror() or
** strerror_r().
**
** The first argument passed to the macro should be the error code that
** will be returned to SQLite (e.g. SQLITE_IOERR_DELETE, SQLITE_CANTOPEN). 
** The two subsequent arguments should be the name of the OS function that
** failed (e.g. "unlink", "open") and the associated file-system path,
** if any.
*/
#define unixLogError(a,b,c)     unixLogErrorAtLine(a,b,c,__LINE__)
static int unixLogErrorAtLine(
  int errcode,                    /* SQLite error code */
  const char *zFunc,              /* Name of OS function that failed */
  const char *zPath,              /* File path associated with error */
  int iLine                       /* Source line number where error occurred */
){
  char *zErr;                     /* Message from strerror() or equivalent */
  int iErrno = errno;             /* Saved syscall error number */

  /* If this is not a threadsafe build (SQLITE_THREADSAFE==0), then use
  ** the strerror() function to obtain the human-readable error message
  ** equivalent to errno. Otherwise, use strerror_r().
  */ 
#if SQLITE_THREADSAFE && defined(HAVE_STRERROR_R)
  char aErr[80];
  memset(aErr, 0, sizeof(aErr));
  zErr = aErr;

  /* If STRERROR_R_CHAR_P (set by autoconf scripts) or __USE_GNU is defined,
  ** assume that the system provides the GNU version of strerror_r() that
  ** returns a pointer to a buffer containing the error message. That pointer 
  ** may point to aErr[], or it may point to some static storage somewhere. 
  ** Otherwise, assume that the system provides the POSIX version of 
  ** strerror_r(), which always writes an error message into aErr[].
  **
  ** If the code incorrectly assumes that it is the POSIX version that is
  ** available, the error message will often be an empty string. Not a
  ** huge problem. Incorrectly concluding that the GNU version is available 
  ** could lead to a segfault though.
  */
#if defined(STRERROR_R_CHAR_P) || defined(__USE_GNU)
  zErr = 
# endif
  strerror_r(iErrno, aErr, sizeof(aErr)-1);

#elif SQLITE_THREADSAFE
  /* This is a threadsafe build, but strerror_r() is not available. */
  zErr = "";
#else
  /* Non-threadsafe build, use strerror(). */
  zErr = strerror(iErrno);
#endif

  if( zPath==0 ) zPath = "";
  sqlite3_log(errcode,
      "os_unix.c:%d: (%d) %s(%s) - %s",
      iLine, iErrno, zFunc, zPath, zErr
  );

  return errcode;
}

/*
** Close a file descriptor.
**
** We assume that close() almost always works, since it is only in a
** very sick application or on a very sick platform that it might fail.
** If it does fail, simply leak the file descriptor, but do log the
** error.
**
** Note that it is not safe to retry close() after EINTR since the
** file descriptor might have already been reused by another thread.
** So we don't even try to recover from an EINTR.  Just log the error
** and move on.
*/
static void robust_close(unixFile *pFile, int h, int lineno){
  if( osClose(h) ){
    unixLogErrorAtLine(SQLITE_IOERR_CLOSE, "close",
                       pFile ? pFile->zPath : 0, lineno);
  }
}

/*
** Set the pFile->lastErrno.  Do this in a subroutine as that provides
** a convenient place to set a breakpoint.
*/
static void storeLastErrno(unixFile *pFile, int error){
  pFile->lastErrno = error;
}

/*
** Close all file descriptors accumuated in the unixInodeInfo->pUnused list.
*/ 
static void closePendingFds(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *p;
  UnixUnusedFd *pNext;
  for(p=pInode->pUnused; p; p=pNext){
    pNext = p->pNext;
    robust_close(pFile, p->fd, __LINE__);
    sqlite3_free(p);
    nUnusedFd--;
  }
  pInode->pUnused = 0;
}

/*
** Release a unixInodeInfo structure previously allocated by findInodeInfo().
**
** The mutex entered using the unixEnterMutex() function must be held
** when this function is called.
*/
static void releaseInodeInfo(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  assert( unixMutexHeld() );
  if( ALWAYS(pInode) ){
    pInode->nRef--;
    if( pInode->nRef==0 ){
      assert( pInode->pShmNode==0 );
      closePendingFds(pFile);
      if( pInode->pPrev ){
        assert( pInode->pPrev->pNext==pInode );
        pInode->pPrev->pNext = pInode->pNext;
      }else{
        assert( inodeList==pInode );
        inodeList = pInode->pNext;
      }
      if( pInode->pNext ){
        assert( pInode->pNext->pPrev==pInode );
        pInode->pNext->pPrev = pInode->pPrev;
      }
      sqlite3_free(pInode);
    }
  }
  assert( inodeList!=0 || nUnusedFd==0 );
}

/*
** Given a file descriptor, locate the unixInodeInfo object that
** describes that file descriptor.  Create a new one if necessary.  The
** return value might be uninitialized if an error occurs.
**
** The mutex entered using the unixEnterMutex() function must be held
** when this function is called.
**
** Return an appropriate error code.
*/
static int findInodeInfo(
  unixFile *pFile,               /* Unix file with file desc used in the key */
  unixInodeInfo **ppInode        /* Return the unixInodeInfo object here */
){
  int rc;                        /* System call return code */
  int fd;                        /* The file descriptor for pFile */
  struct unixFileId fileId;      /* Lookup key for the unixInodeInfo */
  struct stat statbuf;           /* Low-level file information */
  unixInodeInfo *pInode = 0;     /* Candidate unixInodeInfo object */

  assert( unixMutexHeld() );

  /* Get low-level information about the file that we can used to
  ** create a unique name for the file.
  */
  fd = pFile->h;
  rc = osFstat(fd, &statbuf);
  if( rc!=0 ){
    storeLastErrno(pFile, errno);
#if defined(EOVERFLOW) && defined(SQLITE_DISABLE_LFS)
    if( pFile->lastErrno==EOVERFLOW ) return SQLITE_NOLFS;
#endif
    return SQLITE_IOERR;
  }

#ifdef __APPLE__
  /* On OS X on an msdos filesystem, the inode number is reported
  ** incorrectly for zero-size files.  See ticket #3260.  To work
  ** around this problem (we consider it a bug in OS X, not SQLite)
  ** we always increase the file size to 1 by writing a single byte
  ** prior to accessing the inode number.  The one byte written is
  ** an ASCII 'S' character which also happens to be the first byte
  ** in the header of every SQLite database.  In this way, if there
  ** is a race condition such that another thread has already populated
  ** the first page of the database, no damage is done.
  */
  if( statbuf.st_size==0 && (pFile->fsFlags & SQLITE_FSFLAGS_IS_MSDOS)!=0 ){
    do{ rc = osWrite(fd, "S", 1); }while( rc<0 && errno==EINTR );
    if( rc!=1 ){
      storeLastErrno(pFile, errno);
      return SQLITE_IOERR;
    }
    rc = osFstat(fd, &statbuf);
    if( rc!=0 ){
      storeLastErrno(pFile, errno);
      return SQLITE_IOERR;
    }
  }
#endif

  memset(&fileId, 0, sizeof(fileId));
  fileId.dev = statbuf.st_dev;
#if OS_VXWORKS
  fileId.pId = pFile->pId;
#else
  fileId.ino = (u64)statbuf.st_ino;
#endif
  assert( inodeList!=0 || nUnusedFd==0 );
  pInode = inodeList;
  while( pInode && memcmp(&fileId, &pInode->fileId, sizeof(fileId)) ){
    pInode = pInode->pNext;
  }
  if( pInode==0 ){
    pInode = sqlite3_malloc64( sizeof(*pInode) );
    if( pInode==0 ){
      return SQLITE_NOMEM_BKPT;
    }
    memset(pInode, 0, sizeof(*pInode));
    memcpy(&pInode->fileId, &fileId, sizeof(fileId));
    pInode->nRef = 1;
    pInode->pNext = inodeList;
    pInode->pPrev = 0;
    if( inodeList ) inodeList->pPrev = pInode;
    inodeList = pInode;
  }else{
    pInode->nRef++;
  }
  *ppInode = pInode;
  return SQLITE_OK;
}

/*
** Return TRUE if pFile has been renamed or unlinked since it was first opened.
*/
static int fileHasMoved(unixFile *pFile){
#if OS_VXWORKS
  return pFile->pInode!=0 && pFile->pId!=pFile->pInode->fileId.pId;
#else
  struct stat buf;
  return pFile->pInode!=0 &&
      (osStat(pFile->zPath, &buf)!=0 
         || (u64)buf.st_ino!=pFile->pInode->fileId.ino);
#endif
}


/*
** Check a unixFile that is a database.  Verify the following:
**
** (1) There is exactly one hard link on the file
** (2) The file is not a symbolic link
** (3) The file has not been renamed or unlinked
**
** Issue sqlite3_log(SQLITE_WARNING,...) messages if anything is not right.
*/
static void verifyDbFile(unixFile *pFile){
  struct stat buf;
  int rc;

  /* These verifications occurs for the main database only */
  if( pFile->ctrlFlags & UNIXFILE_NOLOCK ) return;

  rc = osFstat(pFile->h, &buf);
  if( rc!=0 ){
    sqlite3_log(SQLITE_WARNING, "cannot fstat db file %s", pFile->zPath);
    return;
  }
  if( buf.st_nlink==0 ){
    sqlite3_log(SQLITE_WARNING, "file unlinked while open: %s", pFile->zPath);
    return;
  }
  if( buf.st_nlink>1 ){
    sqlite3_log(SQLITE_WARNING, "multiple links to file: %s", pFile->zPath);
    return;
  }
  if( fileHasMoved(pFile) ){
    sqlite3_log(SQLITE_WARNING, "file renamed while open: %s", pFile->zPath);
    return;
  }
}


/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, set *pResOut
** to a non-zero value otherwise *pResOut is set to zero.  The return value
** is set to SQLITE_OK unless an I/O error occurs during lock checking.
*/
static int unixCheckReservedLock(sqlite3_file *id, int *pResOut){
  int rc = SQLITE_OK;
  int reserved = 0;
  unixFile *pFile = (unixFile*)id;

  SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );

  assert( pFile );
  assert( pFile->eFileLock<=SHARED_LOCK );
  unixEnterMutex(); /* Because pFile->pInode is shared across threads */

  /* Check if a thread in this process holds such a lock */
  if( pFile->pInode->eFileLock>SHARED_LOCK ){
    reserved = 1;
  }

  /* Otherwise see if some other process holds it.
  */
#ifndef __DJGPP__
  if( !reserved && !pFile->pInode->bProcessLock ){
    struct flock lock;
    lock.l_whence = SEEK_SET;
    lock.l_start = RESERVED_BYTE;
    lock.l_len = 1;
    lock.l_type = F_WRLCK;
    if( osFcntl(pFile->h, F_GETLK, &lock) ){
      rc = SQLITE_IOERR_CHECKRESERVEDLOCK;
      storeLastErrno(pFile, errno);
    } else if( lock.l_type!=F_UNLCK ){
      reserved = 1;
    }
  }
#endif
  
  unixLeaveMutex();
  OSTRACE(("TEST WR-LOCK %d %d %d (unix)\n", pFile->h, rc, reserved));

  *pResOut = reserved;
  return rc;
}

/*
** Set a posix-advisory-lock.
**
** There are two versions of this routine.  If compiled with
** SQLITE_ENABLE_SETLK_TIMEOUT then the routine has an extra parameter
** which is a pointer to a unixFile.  If the unixFile->iBusyTimeout
** value is set, then it is the number of milliseconds to wait before
** failing the lock.  The iBusyTimeout value is always reset back to
** zero on each call.
**
** If SQLITE_ENABLE_SETLK_TIMEOUT is not defined, then do a non-blocking
** attempt to set the lock.
*/
#ifndef SQLITE_ENABLE_SETLK_TIMEOUT
# define osSetPosixAdvisoryLock(h,x,t) osFcntl(h,F_SETLK,x)
#else
static int osSetPosixAdvisoryLock(
  int h,                /* The file descriptor on which to take the lock */
  struct flock *pLock,  /* The description of the lock */
  unixFile *pFile       /* Structure holding timeout value */
){
  int rc = osFcntl(h,F_SETLK,pLock);
  while( rc<0 && pFile->iBusyTimeout>0 ){
    /* On systems that support some kind of blocking file lock with a timeout,
    ** make appropriate changes here to invoke that blocking file lock.  On
    ** generic posix, however, there is no such API.  So we simply try the
    ** lock once every millisecond until either the timeout expires, or until
    ** the lock is obtained. */
    usleep(1000);
    rc = osFcntl(h,F_SETLK,pLock);
    pFile->iBusyTimeout--;
  }
  return rc;
}
#endif /* SQLITE_ENABLE_SETLK_TIMEOUT */


/*
** Attempt to set a system-lock on the file pFile.  The lock is 
** described by pLock.
**
** If the pFile was opened read/write from unix-excl, then the only lock
** ever obtained is an exclusive lock, and it is obtained exactly once
** the first time any lock is attempted.  All subsequent system locking
** operations become no-ops.  Locking operations still happen internally,
** in order to coordinate access between separate database connections
** within this process, but all of that is handled in memory and the
** operating system does not participate.
**
** This function is a pass-through to fcntl(F_SETLK) if pFile is using
** any VFS other than "unix-excl" or if pFile is opened on "unix-excl"
** and is read-only.
**
** Zero is returned if the call completes successfully, or -1 if a call
** to fcntl() fails. In this case, errno is set appropriately (by fcntl()).
*/
static int unixFileLock(unixFile *pFile, struct flock *pLock){
  int rc;
  unixInodeInfo *pInode = pFile->pInode;
  assert( unixMutexHeld() );
  assert( pInode!=0 );
  if( (pFile->ctrlFlags & (UNIXFILE_EXCL|UNIXFILE_RDONLY))==UNIXFILE_EXCL ){
    if( pInode->bProcessLock==0 ){
      struct flock lock;
      assert( pInode->nLock==0 );
      lock.l_whence = SEEK_SET;
      lock.l_start = SHARED_FIRST;
      lock.l_len = SHARED_SIZE;
      lock.l_type = F_WRLCK;
      rc = osSetPosixAdvisoryLock(pFile->h, &lock, pFile);
      if( rc<0 ) return rc;
      pInode->bProcessLock = 1;
      pInode->nLock++;
    }else{
      rc = 0;
    }
  }else{
    rc = osSetPosixAdvisoryLock(pFile->h, pLock, pFile);
  }
  return rc;
}

/*
** Lock the file with the lock specified by parameter eFileLock - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK
**     (3) PENDING_LOCK
**     (4) EXCLUSIVE_LOCK
**
** Sometimes when requesting one lock state, additional lock states
** are inserted in between.  The locking might fail on one of the later
** transitions leaving the lock state different from what it started but
** still short of its goal.  The following chart shows the allowed
** transitions and the inserted intermediate states:
**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int unixLock(sqlite3_file *id, int eFileLock){
  /* The following describes the implementation of the various locks and
  ** lock transitions in terms of the POSIX advisory shared and exclusive
  ** lock primitives (called read-locks and write-locks below, to avoid
  ** confusion with SQLite lock names). The algorithms are complicated
  ** slightly in order to be compatible with Windows95 systems simultaneously
  ** accessing the same database file, in case that is ever required.
  **
  ** Symbols defined in os.h indentify the 'pending byte' and the 'reserved
  ** byte', each single bytes at well known offsets, and the 'shared byte
  ** range', a range of 510 bytes at a well known offset.
  **
  ** To obtain a SHARED lock, a read-lock is obtained on the 'pending
  ** byte'.  If this is successful, 'shared byte range' is read-locked
  ** and the lock on the 'pending byte' released.  (Legacy note:  When
  ** SQLite was first developed, Windows95 systems were still very common,
  ** and Widnows95 lacks a shared-lock capability.  So on Windows95, a
  ** single randomly selected by from the 'shared byte range' is locked.
  ** Windows95 is now pretty much extinct, but this work-around for the
  ** lack of shared-locks on Windows95 lives on, for backwards
  ** compatibility.)
  **
  ** A process may only obtain a RESERVED lock after it has a SHARED lock.
  ** A RESERVED lock is implemented by grabbing a write-lock on the
  ** 'reserved byte'. 
  **
  ** A process may only obtain a PENDING lock after it has obtained a
  ** SHARED lock. A PENDING lock is implemented by obtaining a write-lock
  ** on the 'pending byte'. This ensures that no new SHARED locks can be
  ** obtained, but existing SHARED locks are allowed to persist. A process
  ** does not have to obtain a RESERVED lock on the way to a PENDING lock.
  ** This property is used by the algorithm for rolling back a journal file
  ** after a crash.
  **
  ** An EXCLUSIVE lock, obtained after a PENDING lock is held, is
  ** implemented by obtaining a write-lock on the entire 'shared byte
  ** range'. Since all other locks require a read-lock on one of the bytes
  ** within this range, this ensures that no other locks are held on the
  ** database. 
  */
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode;
  struct flock lock;
  int tErrno = 0;

  assert( pFile );
  OSTRACE(("LOCK    %d %s was %s(%s,%d) pid=%d (unix)\n", pFile->h,
      azFileLock(eFileLock), azFileLock(pFile->eFileLock),
      azFileLock(pFile->pInode->eFileLock), pFile->pInode->nShared,
      osGetpid(0)));

  /* If there is already a lock of this type or more restrictive on the
  ** unixFile, do nothing. Don't use the end_lock: exit path, as
  ** unixEnterMutex() hasn't been called yet.
  */
  if( pFile->eFileLock>=eFileLock ){
    OSTRACE(("LOCK    %d %s ok (already held) (unix)\n", pFile->h,
            azFileLock(eFileLock)));
    return SQLITE_OK;
  }

  /* Make sure the locking sequence is correct.
  **  (1) We never move from unlocked to anything higher than shared lock.
  **  (2) SQLite never explicitly requests a pendig lock.
  **  (3) A shared lock is always held when a reserve lock is requested.
  */
  assert( pFile->eFileLock!=NO_LOCK || eFileLock==SHARED_LOCK );
  assert( eFileLock!=PENDING_LOCK );
  assert( eFileLock!=RESERVED_LOCK || pFile->eFileLock==SHARED_LOCK );

  /* This mutex is needed because pFile->pInode is shared across threads
  */
  unixEnterMutex();
  pInode = pFile->pInode;

  /* If some thread using this PID has a lock via a different unixFile*
  ** handle that precludes the requested lock, return BUSY.
  */
  if( (pFile->eFileLock!=pInode->eFileLock && 
          (pInode->eFileLock>=PENDING_LOCK || eFileLock>SHARED_LOCK))
  ){
    rc = SQLITE_BUSY;
    goto end_lock;
  }

  /* If a SHARED lock is requested, and some thread using this PID already
  ** has a SHARED or RESERVED lock, then increment reference counts and
  ** return SQLITE_OK.
  */
  if( eFileLock==SHARED_LOCK && 
      (pInode->eFileLock==SHARED_LOCK || pInode->eFileLock==RESERVED_LOCK) ){
    assert( eFileLock==SHARED_LOCK );
    assert( pFile->eFileLock==0 );
    assert( pInode->nShared>0 );
    pFile->eFileLock = SHARED_LOCK;
    pInode->nShared++;
    pInode->nLock++;
    goto end_lock;
  }


  /* A PENDING lock is needed before acquiring a SHARED lock and before
  ** acquiring an EXCLUSIVE lock.  For the SHARED lock, the PENDING will
  ** be released.
  */
  lock.l_len = 1L;
  lock.l_whence = SEEK_SET;
  if( eFileLock==SHARED_LOCK 
      || (eFileLock==EXCLUSIVE_LOCK && pFile->eFileLock<PENDING_LOCK)
  ){
    lock.l_type = (eFileLock==SHARED_LOCK?F_RDLCK:F_WRLCK);
    lock.l_start = PENDING_BYTE;
    if( unixFileLock(pFile, &lock) ){
      tErrno = errno;
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
      if( rc!=SQLITE_BUSY ){
        storeLastErrno(pFile, tErrno);
      }
      goto end_lock;
    }
  }


  /* If control gets to this point, then actually go ahead and make
  ** operating system calls for the specified lock.
  */
  if( eFileLock==SHARED_LOCK ){
    assert( pInode->nShared==0 );
    assert( pInode->eFileLock==0 );
    assert( rc==SQLITE_OK );

    /* Now get the read-lock */
    lock.l_start = SHARED_FIRST;
    lock.l_len = SHARED_SIZE;
    if( unixFileLock(pFile, &lock) ){
      tErrno = errno;
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
    }

    /* Drop the temporary PENDING lock */
    lock.l_start = PENDING_BYTE;
    lock.l_len = 1L;
    lock.l_type = F_UNLCK;
    if( unixFileLock(pFile, &lock) && rc==SQLITE_OK ){
      /* This could happen with a network mount */
      tErrno = errno;
      rc = SQLITE_IOERR_UNLOCK; 
    }

    if( rc ){
      if( rc!=SQLITE_BUSY ){
        storeLastErrno(pFile, tErrno);
      }
      goto end_lock;
    }else{
      pFile->eFileLock = SHARED_LOCK;
      pInode->nLock++;
      pInode->nShared = 1;
    }
  }else if( eFileLock==EXCLUSIVE_LOCK && pInode->nShared>1 ){
    /* We are trying for an exclusive lock but another thread in this
    ** same process is still holding a shared lock. */
    rc = SQLITE_BUSY;
  }else{
    /* The request was for a RESERVED or EXCLUSIVE lock.  It is
    ** assumed that there is a SHARED or greater lock on the file
    ** already.
    */
    assert( 0!=pFile->eFileLock );
    lock.l_type = F_WRLCK;

    assert( eFileLock==RESERVED_LOCK || eFileLock==EXCLUSIVE_LOCK );
    if( eFileLock==RESERVED_LOCK ){
      lock.l_start = RESERVED_BYTE;
      lock.l_len = 1L;
    }else{
      lock.l_start = SHARED_FIRST;
      lock.l_len = SHARED_SIZE;
    }

    if( unixFileLock(pFile, &lock) ){
      tErrno = errno;
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
      if( rc!=SQLITE_BUSY ){
        storeLastErrno(pFile, tErrno);
      }
    }
  }
  

#ifdef SQLITE_DEBUG
  /* Set up the transaction-counter change checking flags when
  ** transitioning from a SHARED to a RESERVED lock.  The change
  ** from SHARED to RESERVED marks the beginning of a normal
  ** write operation (not a hot journal rollback).
  */
  if( rc==SQLITE_OK
   && pFile->eFileLock<=SHARED_LOCK
   && eFileLock==RESERVED_LOCK
  ){
    pFile->transCntrChng = 0;
    pFile->dbUpdate = 0;
    pFile->inNormalWrite = 1;
  }
#endif


  if( rc==SQLITE_OK ){
    pFile->eFileLock = eFileLock;
    pInode->eFileLock = eFileLock;
  }else if( eFileLock==EXCLUSIVE_LOCK ){
    pFile->eFileLock = PENDING_LOCK;
    pInode->eFileLock = PENDING_LOCK;
  }

end_lock:
  unixLeaveMutex();
  OSTRACE(("LOCK    %d %s %s (unix)\n", pFile->h, azFileLock(eFileLock), 
      rc==SQLITE_OK ? "ok" : "failed"));
  return rc;
}

/*
** Add the file descriptor used by file handle pFile to the corresponding
** pUnused list.
*/
static void setPendingFd(unixFile *pFile){
  unixInodeInfo *pInode = pFile->pInode;
  UnixUnusedFd *p = pFile->pPreallocatedUnused;
  p->pNext = pInode->pUnused;
  pInode->pUnused = p;
  pFile->h = -1;
  pFile->pPreallocatedUnused = 0;
  nUnusedFd++;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
** 
** If handleNFSUnlock is true, then on downgrading an EXCLUSIVE_LOCK to SHARED
** the byte range is divided into 2 parts and the first part is unlocked then
** set to a read lock, then the other part is simply unlocked.  This works 
** around a bug in BSD NFS lockd (also seen on MacOSX 10.3+) that fails to 
** remove the write lock on a region when a read lock is set.
*/
static int posixUnlock(sqlite3_file *id, int eFileLock, int handleNFSUnlock){
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode;
  struct flock lock;
  int rc = SQLITE_OK;

  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (unix)\n", pFile->h, eFileLock,
      pFile->eFileLock, pFile->pInode->eFileLock, pFile->pInode->nShared,
      osGetpid(0)));

  assert( eFileLock<=SHARED_LOCK );
  if( pFile->eFileLock<=eFileLock ){
    return SQLITE_OK;
  }
  unixEnterMutex();
  pInode = pFile->pInode;
  assert( pInode->nShared!=0 );
  if( pFile->eFileLock>SHARED_LOCK ){
    assert( pInode->eFileLock==pFile->eFileLock );

#ifdef SQLITE_DEBUG
    /* When reducing a lock such that other processes can start
    ** reading the database file again, make sure that the
    ** transaction counter was updated if any part of the database
    ** file changed.  If the transaction counter is not updated,
    ** other connections to the same file might not realize that
    ** the file has changed and hence might not know to flush their
    ** cache.  The use of a stale cache can lead to database corruption.
    */
    pFile->inNormalWrite = 0;
#endif

    /* downgrading to a shared lock on NFS involves clearing the write lock
    ** before establishing the readlock - to avoid a race condition we downgrade
    ** the lock in 2 blocks, so that part of the range will be covered by a 
    ** write lock until the rest is covered by a read lock:
    **  1:   [WWWWW]
    **  2:   [....W]
    **  3:   [RRRRW]
    **  4:   [RRRR.]
    */
    if( eFileLock==SHARED_LOCK ){
#if !defined(__APPLE__) || !SQLITE_ENABLE_LOCKING_STYLE
      (void)handleNFSUnlock;
      assert( handleNFSUnlock==0 );
#endif
#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
      if( handleNFSUnlock ){
        int tErrno;               /* Error code from system call errors */
        off_t divSize = SHARED_SIZE - 1;
        
        lock.l_type = F_UNLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST;
        lock.l_len = divSize;
        if( unixFileLock(pFile, &lock)==(-1) ){
          tErrno = errno;
          rc = SQLITE_IOERR_UNLOCK;
          storeLastErrno(pFile, tErrno);
          goto end_unlock;
        }
        lock.l_type = F_RDLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST;
        lock.l_len = divSize;
        if( unixFileLock(pFile, &lock)==(-1) ){
          tErrno = errno;
          rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_RDLOCK);
          if( IS_LOCK_ERROR(rc) ){
            storeLastErrno(pFile, tErrno);
          }
          goto end_unlock;
        }
        lock.l_type = F_UNLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST+divSize;
        lock.l_len = SHARED_SIZE-divSize;
        if( unixFileLock(pFile, &lock)==(-1) ){
          tErrno = errno;
          rc = SQLITE_IOERR_UNLOCK;
          storeLastErrno(pFile, tErrno);
          goto end_unlock;
        }
      }else
#endif /* defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE */
      {
        lock.l_type = F_RDLCK;
        lock.l_whence = SEEK_SET;
        lock.l_start = SHARED_FIRST;
        lock.l_len = SHARED_SIZE;
        if( unixFileLock(pFile, &lock) ){
          /* In theory, the call to unixFileLock() cannot fail because another
          ** process is holding an incompatible lock. If it does, this 
          ** indicates that the other process is not following the locking
          ** protocol. If this happens, return SQLITE_IOERR_RDLOCK. Returning
          ** SQLITE_BUSY would confuse the upper layer (in practice it causes 
          ** an assert to fail). */ 
          rc = SQLITE_IOERR_RDLOCK;
          storeLastErrno(pFile, errno);
          goto end_unlock;
        }
      }
    }
    lock.l_type = F_UNLCK;
    lock.l_whence = SEEK_SET;
    lock.l_start = PENDING_BYTE;
    lock.l_len = 2L;  assert( PENDING_BYTE+1==RESERVED_BYTE );
    if( unixFileLock(pFile, &lock)==0 ){
      pInode->eFileLock = SHARED_LOCK;
    }else{
      rc = SQLITE_IOERR_UNLOCK;
      storeLastErrno(pFile, errno);
      goto end_unlock;
    }
  }
  if( eFileLock==NO_LOCK ){
    /* Decrement the shared lock counter.  Release the lock using an
    ** OS call only when all threads in this same process have released
    ** the lock.
    */
    pInode->nShared--;
    if( pInode->nShared==0 ){
      lock.l_type = F_UNLCK;
      lock.l_whence = SEEK_SET;
      lock.l_start = lock.l_len = 0L;
      if( unixFileLock(pFile, &lock)==0 ){
        pInode->eFileLock = NO_LOCK;
      }else{
        rc = SQLITE_IOERR_UNLOCK;
        storeLastErrno(pFile, errno);
        pInode->eFileLock = NO_LOCK;
        pFile->eFileLock = NO_LOCK;
      }
    }

    /* Decrement the count of locks against this same file.  When the
    ** count reaches zero, close any other file descriptors whose close
    ** was deferred because of outstanding locks.
    */
    pInode->nLock--;
    assert( pInode->nLock>=0 );
    if( pInode->nLock==0 ){
      closePendingFds(pFile);
    }
  }

end_unlock:
  unixLeaveMutex();
  if( rc==SQLITE_OK ) pFile->eFileLock = eFileLock;
  return rc;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int unixUnlock(sqlite3_file *id, int eFileLock){
#if SQLITE_MAX_MMAP_SIZE>0
  assert( eFileLock==SHARED_LOCK || ((unixFile *)id)->nFetchOut==0 );
#endif
  return posixUnlock(id, eFileLock, 0);
}

#if SQLITE_MAX_MMAP_SIZE>0
static int unixMapfile(unixFile *pFd, i64 nByte);
static void unixUnmapfile(unixFile *pFd);
#endif

/*
** This function performs the parts of the "close file" operation 
** common to all locking schemes. It closes the directory and file
** handles, if they are valid, and sets all fields of the unixFile
** structure to 0.
**
** It is *not* necessary to hold the mutex when this routine is called,
** even on VxWorks.  A mutex will be acquired on VxWorks by the
** vxworksReleaseFileId() routine.
*/
static int closeUnixFile(sqlite3_file *id){
  unixFile *pFile = (unixFile*)id;
#if SQLITE_MAX_MMAP_SIZE>0
  unixUnmapfile(pFile);
#endif
  if( pFile->h>=0 ){
    robust_close(pFile, pFile->h, __LINE__);
    pFile->h = -1;
  }
#if OS_VXWORKS
  if( pFile->pId ){
    if( pFile->ctrlFlags & UNIXFILE_DELETE ){
      osUnlink(pFile->pId->zCanonicalName);
    }
    vxworksReleaseFileId(pFile->pId);
    pFile->pId = 0;
  }
#endif
#ifdef SQLITE_UNLINK_AFTER_CLOSE
  if( pFile->ctrlFlags & UNIXFILE_DELETE ){
    osUnlink(pFile->zPath);
    sqlite3_free(*(char**)&pFile->zPath);
    pFile->zPath = 0;
  }
#endif
  OSTRACE(("CLOSE   %-3d\n", pFile->h));
  OpenCounter(-1);
  sqlite3_free(pFile->pPreallocatedUnused);
  memset(pFile, 0, sizeof(unixFile));
  return SQLITE_OK;
}

/*
** Close a file.
*/
static int unixClose(sqlite3_file *id){
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile *)id;
  verifyDbFile(pFile);
  unixUnlock(id, NO_LOCK);
  unixEnterMutex();

  /* unixFile.pInode is always valid here. Otherwise, a different close
  ** routine (e.g. nolockClose()) would be called instead.
  */
  assert( pFile->pInode->nLock>0 || pFile->pInode->bProcessLock==0 );
  if( ALWAYS(pFile->pInode) && pFile->pInode->nLock ){
    /* If there are outstanding locks, do not actually close the file just
    ** yet because that would clear those locks.  Instead, add the file
    ** descriptor to pInode->pUnused list.  It will be automatically closed 
    ** when the last lock is cleared.
    */
    setPendingFd(pFile);
  }
  releaseInodeInfo(pFile);
  rc = closeUnixFile(id);
  unixLeaveMutex();
  return rc;
}

/************** End of the posix advisory lock implementation *****************
******************************************************************************/

/******************************************************************************
****************************** No-op Locking **********************************
**
** Of the various locking implementations available, this is by far the
** simplest:  locking is ignored.  No attempt is made to lock the database
** file for reading or writing.
**
** This locking mode is appropriate for use on read-only databases
** (ex: databases that are burned into CD-ROM, for example.)  It can
** also be used if the application employs some external mechanism to
** prevent simultaneous access of the same database by two or more
** database connections.  But there is a serious risk of database
** corruption if this locking mode is used in situations where multiple
** database connections are accessing the same database file at the same
** time and one or more of those connections are writing.
*/

static int nolockCheckReservedLock(sqlite3_file *NotUsed, int *pResOut){
  UNUSED_PARAMETER(NotUsed);
  *pResOut = 0;
  return SQLITE_OK;
}
static int nolockLock(sqlite3_file *NotUsed, int NotUsed2){
  UNUSED_PARAMETER2(NotUsed, NotUsed2);
  return SQLITE_OK;
}
static int nolockUnlock(sqlite3_file *NotUsed, int NotUsed2){
  UNUSED_PARAMETER2(NotUsed, NotUsed2);
  return SQLITE_OK;
}

/*
** Close the file.
*/
static int nolockClose(sqlite3_file *id) {
  return closeUnixFile(id);
}

/******************* End of the no-op lock implementation *********************
******************************************************************************/

/******************************************************************************
************************* Begin dot-file Locking ******************************
**
** The dotfile locking implementation uses the existence of separate lock
** files (really a directory) to control access to the database.  This works
** on just about every filesystem imaginable.  But there are serious downsides:
**
**    (1)  There is zero concurrency.  A single reader blocks all other
**         connections from reading or writing the database.
**
**    (2)  An application crash or power loss can leave stale lock files
**         sitting around that need to be cleared manually.
**
** Nevertheless, a dotlock is an appropriate locking mode for use if no
** other locking strategy is available.
**
** Dotfile locking works by creating a subdirectory in the same directory as
** the database and with the same name but with a ".lock" extension added.
** The existence of a lock directory implies an EXCLUSIVE lock.  All other
** lock types (SHARED, RESERVED, PENDING) are mapped into EXCLUSIVE.
*/

/*
** The file suffix added to the data base filename in order to create the
** lock directory.
*/
#define DOTLOCK_SUFFIX ".lock"

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, set *pResOut
** to a non-zero value otherwise *pResOut is set to zero.  The return value
** is set to SQLITE_OK unless an I/O error occurs during lock checking.
**
** In dotfile locking, either a lock exists or it does not.  So in this
** variation of CheckReservedLock(), *pResOut is set to true if any lock
** is held on the file and false if the file is unlocked.
*/
static int dotlockCheckReservedLock(sqlite3_file *id, int *pResOut) {
  int rc = SQLITE_OK;
  int reserved = 0;
  unixFile *pFile = (unixFile*)id;

  SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );
  
  assert( pFile );
  reserved = osAccess((const char*)pFile->lockingContext, 0)==0;
  OSTRACE(("TEST WR-LOCK %d %d %d (dotlock)\n", pFile->h, rc, reserved));
  *pResOut = reserved;
  return rc;
}

/*
** Lock the file with the lock specified by parameter eFileLock - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK
**     (3) PENDING_LOCK
**     (4) EXCLUSIVE_LOCK
**
** Sometimes when requesting one lock state, additional lock states
** are inserted in between.  The locking might fail on one of the later
** transitions leaving the lock state different from what it started but
** still short of its goal.  The following chart shows the allowed
** transitions and the inserted intermediate states:
**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
**
** With dotfile locking, we really only support state (4): EXCLUSIVE.
** But we track the other locking levels internally.
*/
static int dotlockLock(sqlite3_file *id, int eFileLock) {
  unixFile *pFile = (unixFile*)id;
  char *zLockFile = (char *)pFile->lockingContext;
  int rc = SQLITE_OK;


  /* If we have any lock, then the lock file already exists.  All we have
  ** to do is adjust our internal record of the lock level.
  */
  if( pFile->eFileLock > NO_LOCK ){
    pFile->eFileLock = eFileLock;
    /* Always update the timestamp on the old file */
#ifdef HAVE_UTIME
    utime(zLockFile, NULL);
#else
    utimes(zLockFile, NULL);
#endif
    return SQLITE_OK;
  }
  
  /* grab an exclusive lock */
  rc = osMkdir(zLockFile, 0777);
  if( rc<0 ){
    /* failed to open/create the lock directory */
    int tErrno = errno;
    if( EEXIST == tErrno ){
      rc = SQLITE_BUSY;
    } else {
      rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
      if( rc!=SQLITE_BUSY ){
        storeLastErrno(pFile, tErrno);
      }
    }
    return rc;
  } 
  
  /* got it, set the type and return ok */
  pFile->eFileLock = eFileLock;
  return rc;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
**
** When the locking level reaches NO_LOCK, delete the lock file.
*/
static int dotlockUnlock(sqlite3_file *id, int eFileLock) {
  unixFile *pFile = (unixFile*)id;
  char *zLockFile = (char *)pFile->lockingContext;
  int rc;

  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d pid=%d (dotlock)\n", pFile->h, eFileLock,
           pFile->eFileLock, osGetpid(0)));
  assert( eFileLock<=SHARED_LOCK );
  
  /* no-op if possible */
  if( pFile->eFileLock==eFileLock ){
    return SQLITE_OK;
  }

  /* To downgrade to shared, simply update our internal notion of the
  ** lock state.  No need to mess with the file on disk.
  */
  if( eFileLock==SHARED_LOCK ){
    pFile->eFileLock = SHARED_LOCK;
    return SQLITE_OK;
  }
  
  /* To fully unlock the database, delete the lock file */
  assert( eFileLock==NO_LOCK );
  rc = osRmdir(zLockFile);
  if( rc<0 ){
    int tErrno = errno;
    if( tErrno==ENOENT ){
      rc = SQLITE_OK;
    }else{
      rc = SQLITE_IOERR_UNLOCK;
      storeLastErrno(pFile, tErrno);
    }
    return rc; 
  }
  pFile->eFileLock = NO_LOCK;
  return SQLITE_OK;
}

/*
** Close a file.  Make sure the lock has been released before closing.
*/
static int dotlockClose(sqlite3_file *id) {
  unixFile *pFile = (unixFile*)id;
  assert( id!=0 );
  dotlockUnlock(id, NO_LOCK);
  sqlite3_free(pFile->lockingContext);
  return closeUnixFile(id);
}
/****************** End of the dot-file lock implementation *******************
******************************************************************************/

/******************************************************************************
************************** Begin flock Locking ********************************
**
** Use the flock() system call to do file locking.
**
** flock() locking is like dot-file locking in that the various
** fine-grain locking levels supported by SQLite are collapsed into
** a single exclusive lock.  In other words, SHARED, RESERVED, and
** PENDING locks are the same thing as an EXCLUSIVE lock.  SQLite
** still works when you do this, but concurrency is reduced since
** only a single process can be reading the database at a time.
**
** Omit this section if SQLITE_ENABLE_LOCKING_STYLE is turned off
*/
#if SQLITE_ENABLE_LOCKING_STYLE

/*
** Retry flock() calls that fail with EINTR
*/
#ifdef EINTR
static int robust_flock(int fd, int op){
  int rc;
  do{ rc = flock(fd,op); }while( rc<0 && errno==EINTR );
  return rc;
}
#else
# define robust_flock(a,b) flock(a,b)
#endif
     

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, set *pResOut
** to a non-zero value otherwise *pResOut is set to zero.  The return value
** is set to SQLITE_OK unless an I/O error occurs during lock checking.
*/
static int flockCheckReservedLock(sqlite3_file *id, int *pResOut){
  int rc = SQLITE_OK;
  int reserved = 0;
  unixFile *pFile = (unixFile*)id;
  
  SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );
  
  assert( pFile );
  
  /* Check if a thread in this process holds such a lock */
  if( pFile->eFileLock>SHARED_LOCK ){
    reserved = 1;
  }
  
  /* Otherwise see if some other process holds it. */
  if( !reserved ){
    /* attempt to get the lock */
    int lrc = robust_flock(pFile->h, LOCK_EX | LOCK_NB);
    if( !lrc ){
      /* got the lock, unlock it */
      lrc = robust_flock(pFile->h, LOCK_UN);
      if ( lrc ) {
        int tErrno = errno;
        /* unlock failed with an error */
        lrc = SQLITE_IOERR_UNLOCK; 
        storeLastErrno(pFile, tErrno);
        rc = lrc;
      }
    } else {
      int tErrno = errno;
      reserved = 1;
      /* someone else might have it reserved */
      lrc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK); 
      if( IS_LOCK_ERROR(lrc) ){
        storeLastErrno(pFile, tErrno);
        rc = lrc;
      }
    }
  }
  OSTRACE(("TEST WR-LOCK %d %d %d (flock)\n", pFile->h, rc, reserved));

#ifdef SQLITE_IGNORE_FLOCK_LOCK_ERRORS
  if( (rc & 0xff) == SQLITE_IOERR ){
    rc = SQLITE_OK;
    reserved=1;
  }
#endif /* SQLITE_IGNORE_FLOCK_LOCK_ERRORS */
  *pResOut = reserved;
  return rc;
}

/*
** Lock the file with the lock specified by parameter eFileLock - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK
**     (3) PENDING_LOCK
**     (4) EXCLUSIVE_LOCK
**
** Sometimes when requesting one lock state, additional lock states
** are inserted in between.  The locking might fail on one of the later
** transitions leaving the lock state different from what it started but
** still short of its goal.  The following chart shows the allowed
** transitions and the inserted intermediate states:
**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** flock() only really support EXCLUSIVE locks.  We track intermediate
** lock states in the sqlite3_file structure, but all locks SHARED or
** above are really EXCLUSIVE locks and exclude all other processes from
** access the file.
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int flockLock(sqlite3_file *id, int eFileLock) {
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;

  assert( pFile );

  /* if we already have a lock, it is exclusive.  
  ** Just adjust level and punt on outta here. */
  if (pFile->eFileLock > NO_LOCK) {
    pFile->eFileLock = eFileLock;
    return SQLITE_OK;
  }
  
  /* grab an exclusive lock */
  
  if (robust_flock(pFile->h, LOCK_EX | LOCK_NB)) {
    int tErrno = errno;
    /* didn't get, must be busy */
    rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_LOCK);
    if( IS_LOCK_ERROR(rc) ){
      storeLastErrno(pFile, tErrno);
    }
  } else {
    /* got it, set the type and return ok */
    pFile->eFileLock = eFileLock;
  }
  OSTRACE(("LOCK    %d %s %s (flock)\n", pFile->h, azFileLock(eFileLock), 
           rc==SQLITE_OK ? "ok" : "failed"));
#ifdef SQLITE_IGNORE_FLOCK_LOCK_ERRORS
  if( (rc & 0xff) == SQLITE_IOERR ){
    rc = SQLITE_BUSY;
  }
#endif /* SQLITE_IGNORE_FLOCK_LOCK_ERRORS */
  return rc;
}


/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int flockUnlock(sqlite3_file *id, int eFileLock) {
  unixFile *pFile = (unixFile*)id;
  
  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d pid=%d (flock)\n", pFile->h, eFileLock,
           pFile->eFileLock, osGetpid(0)));
  assert( eFileLock<=SHARED_LOCK );
  
  /* no-op if possible */
  if( pFile->eFileLock==eFileLock ){
    return SQLITE_OK;
  }
  
  /* shared can just be set because we always have an exclusive */
  if (eFileLock==SHARED_LOCK) {
    pFile->eFileLock = eFileLock;
    return SQLITE_OK;
  }
  
  /* no, really, unlock. */
  if( robust_flock(pFile->h, LOCK_UN) ){
#ifdef SQLITE_IGNORE_FLOCK_LOCK_ERRORS
    return SQLITE_OK;
#endif /* SQLITE_IGNORE_FLOCK_LOCK_ERRORS */
    return SQLITE_IOERR_UNLOCK;
  }else{
    pFile->eFileLock = NO_LOCK;
    return SQLITE_OK;
  }
}

/*
** Close a file.
*/
static int flockClose(sqlite3_file *id) {
  assert( id!=0 );
  flockUnlock(id, NO_LOCK);
  return closeUnixFile(id);
}

#endif /* SQLITE_ENABLE_LOCKING_STYLE && !OS_VXWORK */

/******************* End of the flock lock implementation *********************
******************************************************************************/

/******************************************************************************
************************ Begin Named Semaphore Locking ************************
**
** Named semaphore locking is only supported on VxWorks.
**
** Semaphore locking is like dot-lock and flock in that it really only
** supports EXCLUSIVE locking.  Only a single process can read or write
** the database file at a time.  This reduces potential concurrency, but
** makes the lock implementation much easier.
*/
#if OS_VXWORKS

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, set *pResOut
** to a non-zero value otherwise *pResOut is set to zero.  The return value
** is set to SQLITE_OK unless an I/O error occurs during lock checking.
*/
static int semXCheckReservedLock(sqlite3_file *id, int *pResOut) {
  int rc = SQLITE_OK;
  int reserved = 0;
  unixFile *pFile = (unixFile*)id;

  SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );
  
  assert( pFile );

  /* Check if a thread in this process holds such a lock */
  if( pFile->eFileLock>SHARED_LOCK ){
    reserved = 1;
  }
  
  /* Otherwise see if some other process holds it. */
  if( !reserved ){
    sem_t *pSem = pFile->pInode->pSem;

    if( sem_trywait(pSem)==-1 ){
      int tErrno = errno;
      if( EAGAIN != tErrno ){
        rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_CHECKRESERVEDLOCK);
        storeLastErrno(pFile, tErrno);
      } else {
        /* someone else has the lock when we are in NO_LOCK */
        reserved = (pFile->eFileLock < SHARED_LOCK);
      }
    }else{
      /* we could have it if we want it */
      sem_post(pSem);
    }
  }
  OSTRACE(("TEST WR-LOCK %d %d %d (sem)\n", pFile->h, rc, reserved));

  *pResOut = reserved;
  return rc;
}

/*
** Lock the file with the lock specified by parameter eFileLock - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK
**     (3) PENDING_LOCK
**     (4) EXCLUSIVE_LOCK
**
** Sometimes when requesting one lock state, additional lock states
** are inserted in between.  The locking might fail on one of the later
** transitions leaving the lock state different from what it started but
** still short of its goal.  The following chart shows the allowed
** transitions and the inserted intermediate states:
**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** Semaphore locks only really support EXCLUSIVE locks.  We track intermediate
** lock states in the sqlite3_file structure, but all locks SHARED or
** above are really EXCLUSIVE locks and exclude all other processes from
** access the file.
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int semXLock(sqlite3_file *id, int eFileLock) {
  unixFile *pFile = (unixFile*)id;
  sem_t *pSem = pFile->pInode->pSem;
  int rc = SQLITE_OK;

  /* if we already have a lock, it is exclusive.  
  ** Just adjust level and punt on outta here. */
  if (pFile->eFileLock > NO_LOCK) {
    pFile->eFileLock = eFileLock;
    rc = SQLITE_OK;
    goto sem_end_lock;
  }
  
  /* lock semaphore now but bail out when already locked. */
  if( sem_trywait(pSem)==-1 ){
    rc = SQLITE_BUSY;
    goto sem_end_lock;
  }

  /* got it, set the type and return ok */
  pFile->eFileLock = eFileLock;

 sem_end_lock:
  return rc;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int semXUnlock(sqlite3_file *id, int eFileLock) {
  unixFile *pFile = (unixFile*)id;
  sem_t *pSem = pFile->pInode->pSem;

  assert( pFile );
  assert( pSem );
  OSTRACE(("UNLOCK  %d %d was %d pid=%d (sem)\n", pFile->h, eFileLock,
           pFile->eFileLock, osGetpid(0)));
  assert( eFileLock<=SHARED_LOCK );
  
  /* no-op if possible */
  if( pFile->eFileLock==eFileLock ){
    return SQLITE_OK;
  }
  
  /* shared can just be set because we always have an exclusive */
  if (eFileLock==SHARED_LOCK) {
    pFile->eFileLock = eFileLock;
    return SQLITE_OK;
  }
  
  /* no, really unlock. */
  if ( sem_post(pSem)==-1 ) {
    int rc, tErrno = errno;
    rc = sqliteErrorFromPosixError(tErrno, SQLITE_IOERR_UNLOCK);
    if( IS_LOCK_ERROR(rc) ){
      storeLastErrno(pFile, tErrno);
    }
    return rc; 
  }
  pFile->eFileLock = NO_LOCK;
  return SQLITE_OK;
}

/*
 ** Close a file.
 */
static int semXClose(sqlite3_file *id) {
  if( id ){
    unixFile *pFile = (unixFile*)id;
    semXUnlock(id, NO_LOCK);
    assert( pFile );
    unixEnterMutex();
    releaseInodeInfo(pFile);
    unixLeaveMutex();
    closeUnixFile(id);
  }
  return SQLITE_OK;
}

#endif /* OS_VXWORKS */
/*
** Named semaphore locking is only available on VxWorks.
**
*************** End of the named semaphore lock implementation ****************
******************************************************************************/


/******************************************************************************
*************************** Begin AFP Locking *********************************
**
** AFP is the Apple Filing Protocol.  AFP is a network filesystem found
** on Apple Macintosh computers - both OS9 and OSX.
**
** Third-party implementations of AFP are available.  But this code here
** only works on OSX.
*/

#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
/*
** The afpLockingContext structure contains all afp lock specific state
*/
typedef struct afpLockingContext afpLockingContext;
struct afpLockingContext {
  int reserved;
  const char *dbPath;             /* Name of the open file */
};

struct ByteRangeLockPB2
{
  unsigned long long offset;        /* offset to first byte to lock */
  unsigned long long length;        /* nbr of bytes to lock */
  unsigned long long retRangeStart; /* nbr of 1st byte locked if successful */
  unsigned char unLockFlag;         /* 1 = unlock, 0 = lock */
  unsigned char startEndFlag;       /* 1=rel to end of fork, 0=rel to start */
  int fd;                           /* file desc to assoc this lock with */
};

#define afpfsByteRangeLock2FSCTL        _IOWR('z', 23, struct ByteRangeLockPB2)

/*
** This is a utility for setting or clearing a bit-range lock on an
** AFP filesystem.
** 
** Return SQLITE_OK on success, SQLITE_BUSY on failure.
*/
static int afpSetLock(
  const char *path,              /* Name of the file to be locked or unlocked */
  unixFile *pFile,               /* Open file descriptor on path */
  unsigned long long offset,     /* First byte to be locked */
  unsigned long long length,     /* Number of bytes to lock */
  int setLockFlag                /* True to set lock.  False to clear lock */
){
  struct ByteRangeLockPB2 pb;
  int err;
  
  pb.unLockFlag = setLockFlag ? 0 : 1;
  pb.startEndFlag = 0;
  pb.offset = offset;
  pb.length = length; 
  pb.fd = pFile->h;
  
  OSTRACE(("AFPSETLOCK [%s] for %d%s in range %llx:%llx\n", 
    (setLockFlag?"ON":"OFF"), pFile->h, (pb.fd==-1?"[testval-1]":""),
    offset, length));
  err = fsctl(path, afpfsByteRangeLock2FSCTL, &pb, 0);
  if ( err==-1 ) {
    int rc;
    int tErrno = errno;
    OSTRACE(("AFPSETLOCK failed to fsctl() '%s' %d %s\n",
             path, tErrno, strerror(tErrno)));
#ifdef SQLITE_IGNORE_AFP_LOCK_ERRORS
    rc = SQLITE_BUSY;
#else
    rc = sqliteErrorFromPosixError(tErrno,
                    setLockFlag ? SQLITE_IOERR_LOCK : SQLITE_IOERR_UNLOCK);
#endif /* SQLITE_IGNORE_AFP_LOCK_ERRORS */
    if( IS_LOCK_ERROR(rc) ){
      storeLastErrno(pFile, tErrno);
    }
    return rc;
  } else {
    return SQLITE_OK;
  }
}

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, set *pResOut
** to a non-zero value otherwise *pResOut is set to zero.  The return value
** is set to SQLITE_OK unless an I/O error occurs during lock checking.
*/
static int afpCheckReservedLock(sqlite3_file *id, int *pResOut){
  int rc = SQLITE_OK;
  int reserved = 0;
  unixFile *pFile = (unixFile*)id;
  afpLockingContext *context;
  
  SimulateIOError( return SQLITE_IOERR_CHECKRESERVEDLOCK; );
  
  assert( pFile );
  context = (afpLockingContext *) pFile->lockingContext;
  if( context->reserved ){
    *pResOut = 1;
    return SQLITE_OK;
  }
  unixEnterMutex(); /* Because pFile->pInode is shared across threads */
  
  /* Check if a thread in this process holds such a lock */
  if( pFile->pInode->eFileLock>SHARED_LOCK ){
    reserved = 1;
  }
  
  /* Otherwise see if some other process holds it.
   */
  if( !reserved ){
    /* lock the RESERVED byte */
    int lrc = afpSetLock(context->dbPath, pFile, RESERVED_BYTE, 1,1);  
    if( SQLITE_OK==lrc ){
      /* if we succeeded in taking the reserved lock, unlock it to restore
      ** the original state */
      lrc = afpSetLock(context->dbPath, pFile, RESERVED_BYTE, 1, 0);
    } else {
      /* if we failed to get the lock then someone else must have it */
      reserved = 1;
    }
    if( IS_LOCK_ERROR(lrc) ){
      rc=lrc;
    }
  }
  
  unixLeaveMutex();
  OSTRACE(("TEST WR-LOCK %d %d %d (afp)\n", pFile->h, rc, reserved));
  
  *pResOut = reserved;
  return rc;
}

/*
** Lock the file with the lock specified by parameter eFileLock - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK
**     (3) PENDING_LOCK
**     (4) EXCLUSIVE_LOCK
**
** Sometimes when requesting one lock state, additional lock states
** are inserted in between.  The locking might fail on one of the later
** transitions leaving the lock state different from what it started but
** still short of its goal.  The following chart shows the allowed
** transitions and the inserted intermediate states:
**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int afpLock(sqlite3_file *id, int eFileLock){
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode = pFile->pInode;
  afpLockingContext *context = (afpLockingContext *) pFile->lockingContext;
  
  assert( pFile );
  OSTRACE(("LOCK    %d %s was %s(%s,%d) pid=%d (afp)\n", pFile->h,
           azFileLock(eFileLock), azFileLock(pFile->eFileLock),
           azFileLock(pInode->eFileLock), pInode->nShared , osGetpid(0)));

  /* If there is already a lock of this type or more restrictive on the
  ** unixFile, do nothing. Don't use the afp_end_lock: exit path, as
  ** unixEnterMutex() hasn't been called yet.
  */
  if( pFile->eFileLock>=eFileLock ){
    OSTRACE(("LOCK    %d %s ok (already held) (afp)\n", pFile->h,
           azFileLock(eFileLock)));
    return SQLITE_OK;
  }

  /* Make sure the locking sequence is correct
  **  (1) We never move from unlocked to anything higher than shared lock.
  **  (2) SQLite never explicitly requests a pendig lock.
  **  (3) A shared lock is always held when a reserve lock is requested.
  */
  assert( pFile->eFileLock!=NO_LOCK || eFileLock==SHARED_LOCK );
  assert( eFileLock!=PENDING_LOCK );
  assert( eFileLock!=RESERVED_LOCK || pFile->eFileLock==SHARED_LOCK );
  
  /* This mutex is needed because pFile->pInode is shared across threads
  */
  unixEnterMutex();
  pInode = pFile->pInode;

  /* If some thread using this PID has a lock via a different unixFile*
  ** handle that precludes the requested lock, return BUSY.
  */
  if( (pFile->eFileLock!=pInode->eFileLock && 
       (pInode->eFileLock>=PENDING_LOCK || eFileLock>SHARED_LOCK))
     ){
    rc = SQLITE_BUSY;
    goto afp_end_lock;
  }
  
  /* If a SHARED lock is requested, and some thread using this PID already
  ** has a SHARED or RESERVED lock, then increment reference counts and
  ** return SQLITE_OK.
  */
  if( eFileLock==SHARED_LOCK && 
     (pInode->eFileLock==SHARED_LOCK || pInode->eFileLock==RESERVED_LOCK) ){
    assert( eFileLock==SHARED_LOCK );
    assert( pFile->eFileLock==0 );
    assert( pInode->nShared>0 );
    pFile->eFileLock = SHARED_LOCK;
    pInode->nShared++;
    pInode->nLock++;
    goto afp_end_lock;
  }
    
  /* A PENDING lock is needed before acquiring a SHARED lock and before
  ** acquiring an EXCLUSIVE lock.  For the SHARED lock, the PENDING will
  ** be released.
  */
  if( eFileLock==SHARED_LOCK 
      || (eFileLock==EXCLUSIVE_LOCK && pFile->eFileLock<PENDING_LOCK)
  ){
    int failed;
    failed = afpSetLock(context->dbPath, pFile, PENDING_BYTE, 1, 1);
    if (failed) {
      rc = failed;
      goto afp_end_lock;
    }
  }
  
  /* If control gets to this point, then actually go ahead and make
  ** operating system calls for the specified lock.
  */
  if( eFileLock==SHARED_LOCK ){
    int lrc1, lrc2, lrc1Errno = 0;
    long lk, mask;
    
    assert( pInode->nShared==0 );
    assert( pInode->eFileLock==0 );
        
    mask = (sizeof(long)==8) ? LARGEST_INT64 : 0x7fffffff;
    /* Now get the read-lock SHARED_LOCK */
    /* note that the quality of the randomness doesn't matter that much */
    lk = random(); 
    pInode->sharedByte = (lk & mask)%(SHARED_SIZE - 1);
    lrc1 = afpSetLock(context->dbPath, pFile, 
          SHARED_FIRST+pInode->sharedByte, 1, 1);
    if( IS_LOCK_ERROR(lrc1) ){
      lrc1Errno = pFile->lastErrno;
    }
    /* Drop the temporary PENDING lock */
    lrc2 = afpSetLock(context->dbPath, pFile, PENDING_BYTE, 1, 0);
    
    if( IS_LOCK_ERROR(lrc1) ) {
      storeLastErrno(pFile, lrc1Errno);
      rc = lrc1;
      goto afp_end_lock;
    } else if( IS_LOCK_ERROR(lrc2) ){
      rc = lrc2;
      goto afp_end_lock;
    } else if( lrc1 != SQLITE_OK ) {
      rc = lrc1;
    } else {
      pFile->eFileLock = SHARED_LOCK;
      pInode->nLock++;
      pInode->nShared = 1;
    }
  }else if( eFileLock==EXCLUSIVE_LOCK && pInode->nShared>1 ){
    /* We are trying for an exclusive lock but another thread in this
     ** same process is still holding a shared lock. */
    rc = SQLITE_BUSY;
  }else{
    /* The request was for a RESERVED or EXCLUSIVE lock.  It is
    ** assumed that there is a SHARED or greater lock on the file
    ** already.
    */
    int failed = 0;
    assert( 0!=pFile->eFileLock );
    if (eFileLock >= RESERVED_LOCK && pFile->eFileLock < RESERVED_LOCK) {
        /* Acquire a RESERVED lock */
        failed = afpSetLock(context->dbPath, pFile, RESERVED_BYTE, 1,1);
      if( !failed ){
        context->reserved = 1;
      }
    }
    if (!failed && eFileLock == EXCLUSIVE_LOCK) {
      /* Acquire an EXCLUSIVE lock */
        
      /* Remove the shared lock before trying the range.  we'll need to 
      ** reestablish the shared lock if we can't get the  afpUnlock
      */
      if( !(failed = afpSetLock(context->dbPath, pFile, SHARED_FIRST +
                         pInode->sharedByte, 1, 0)) ){
        int failed2 = SQLITE_OK;
        /* now attemmpt to get the exclusive lock range */
        failed = afpSetLock(context->dbPath, pFile, SHARED_FIRST, 
                               SHARED_SIZE, 1);
        if( failed && (failed2 = afpSetLock(context->dbPath, pFile, 
                       SHARED_FIRST + pInode->sharedByte, 1, 1)) ){
          /* Can't reestablish the shared lock.  Sqlite can't deal, this is
          ** a critical I/O error
          */
          rc = ((failed & 0xff) == SQLITE_IOERR) ? failed2 : 
               SQLITE_IOERR_LOCK;
          goto afp_end_lock;
        } 
      }else{
        rc = failed; 
      }
    }
    if( failed ){
      rc = failed;
    }
  }
  
  if( rc==SQLITE_OK ){
    pFile->eFileLock = eFileLock;
    pInode->eFileLock = eFileLock;
  }else if( eFileLock==EXCLUSIVE_LOCK ){
    pFile->eFileLock = PENDING_LOCK;
    pInode->eFileLock = PENDING_LOCK;
  }
  
afp_end_lock:
  unixLeaveMutex();
  OSTRACE(("LOCK    %d %s %s (afp)\n", pFile->h, azFileLock(eFileLock), 
         rc==SQLITE_OK ? "ok" : "failed"));
  return rc;
}

/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int afpUnlock(sqlite3_file *id, int eFileLock) {
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  unixInodeInfo *pInode;
  afpLockingContext *context = (afpLockingContext *) pFile->lockingContext;
  int skipShared = 0;
#ifdef SQLITE_TEST
  int h = pFile->h;
#endif

  assert( pFile );
  OSTRACE(("UNLOCK  %d %d was %d(%d,%d) pid=%d (afp)\n", pFile->h, eFileLock,
           pFile->eFileLock, pFile->pInode->eFileLock, pFile->pInode->nShared,
           osGetpid(0)));

  assert( eFileLock<=SHARED_LOCK );
  if( pFile->eFileLock<=eFileLock ){
    return SQLITE_OK;
  }
  unixEnterMutex();
  pInode = pFile->pInode;
  assert( pInode->nShared!=0 );
  if( pFile->eFileLock>SHARED_LOCK ){
    assert( pInode->eFileLock==pFile->eFileLock );
    SimulateIOErrorBenign(1);
    SimulateIOError( h=(-1) )
    SimulateIOErrorBenign(0);
    
#ifdef SQLITE_DEBUG
    /* When reducing a lock such that other processes can start
    ** reading the database file again, make sure that the
    ** transaction counter was updated if any part of the database
    ** file changed.  If the transaction counter is not updated,
    ** other connections to the same file might not realize that
    ** the file has changed and hence might not know to flush their
    ** cache.  The use of a stale cache can lead to database corruption.
    */
    assert( pFile->inNormalWrite==0
           || pFile->dbUpdate==0
           || pFile->transCntrChng==1 );
    pFile->inNormalWrite = 0;
#endif
    
    if( pFile->eFileLock==EXCLUSIVE_LOCK ){
      rc = afpSetLock(context->dbPath, pFile, SHARED_FIRST, SHARED_SIZE, 0);
      if( rc==SQLITE_OK && (eFileLock==SHARED_LOCK || pInode->nShared>1) ){
        /* only re-establish the shared lock if necessary */
        int sharedLockByte = SHARED_FIRST+pInode->sharedByte;
        rc = afpSetLock(context->dbPath, pFile, sharedLockByte, 1, 1);
      } else {
        skipShared = 1;
      }
    }
    if( rc==SQLITE_OK && pFile->eFileLock>=PENDING_LOCK ){
      rc = afpSetLock(context->dbPath, pFile, PENDING_BYTE, 1, 0);
    } 
    if( rc==SQLITE_OK && pFile->eFileLock>=RESERVED_LOCK && context->reserved ){
      rc = afpSetLock(context->dbPath, pFile, RESERVED_BYTE, 1, 0);
      if( !rc ){ 
        context->reserved = 0; 
      }
    }
    if( rc==SQLITE_OK && (eFileLock==SHARED_LOCK || pInode->nShared>1)){
      pInode->eFileLock = SHARED_LOCK;
    }
  }
  if( rc==SQLITE_OK && eFileLock==NO_LOCK ){

    /* Decrement the shared lock counter.  Release the lock using an
    ** OS call only when all threads in this same process have released
    ** the lock.
    */
    unsigned long long sharedLockByte = SHARED_FIRST+pInode->sharedByte;
    pInode->nShared--;
    if( pInode->nShared==0 ){
      SimulateIOErrorBenign(1);
      SimulateIOError( h=(-1) )
      SimulateIOErrorBenign(0);
      if( !skipShared ){
        rc = afpSetLock(context->dbPath, pFile, sharedLockByte, 1, 0);
      }
      if( !rc ){
        pInode->eFileLock = NO_LOCK;
        pFile->eFileLock = NO_LOCK;
      }
    }
    if( rc==SQLITE_OK ){
      pInode->nLock--;
      assert( pInode->nLock>=0 );
      if( pInode->nLock==0 ){
        closePendingFds(pFile);
      }
    }
  }
  
  unixLeaveMutex();
  if( rc==SQLITE_OK ) pFile->eFileLock = eFileLock;
  return rc;
}

/*
** Close a file & cleanup AFP specific locking context 
*/
static int afpClose(sqlite3_file *id) {
  int rc = SQLITE_OK;
  unixFile *pFile = (unixFile*)id;
  assert( id!=0 );
  afpUnlock(id, NO_LOCK);
  unixEnterMutex();
  if( pFile->pInode && pFile->pInode->nLock ){
    /* If there are outstanding locks, do not actually close the file just
    ** yet because that would clear those locks.  Instead, add the file
    ** descriptor to pInode->aPending.  It will be automatically closed when
    ** the last lock is cleared.
    */
    setPendingFd(pFile);
  }
  releaseInodeInfo(pFile);
  sqlite3_free(pFile->lockingContext);
  rc = closeUnixFile(id);
  unixLeaveMutex();
  return rc;
}

#endif /* defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE */
/*
** The code above is the AFP lock implementation.  The code is specific
** to MacOSX and does not work on other unix platforms.  No alternative
** is available.  If you don't compile for a mac, then the "unix-afp"
** VFS is not available.
**
********************* End of the AFP lock implementation **********************
******************************************************************************/

/******************************************************************************
*************************** Begin NFS Locking ********************************/

#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
/*
 ** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
 ** must be either NO_LOCK or SHARED_LOCK.
 **
 ** If the locking level of the file descriptor is already at or below
 ** the requested locking level, this routine is a no-op.
 */
static int nfsUnlock(sqlite3_file *id, int eFileLock){
  return posixUnlock(id, eFileLock, 1);
}

#endif /* defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE */
/*
** The code above is the NFS lock implementation.  The code is specific
** to MacOSX and does not work on other unix platforms.  No alternative
** is available.  
**
********************* End of the NFS lock implementation **********************
******************************************************************************/

/******************************************************************************
**************** Non-locking sqlite3_file methods *****************************
**
** The next division contains implementations for all methods of the 
** sqlite3_file object other than the locking methods.  The locking
** methods were defined in divisions above (one locking method per
** division).  Those methods that are common to all locking modes
** are gather together into this division.
*/

/*
** Seek to the offset passed as the second argument, then read cnt 
** bytes into pBuf. Return the number of bytes actually read.
**
** NB:  If you define USE_PREAD or USE_PREAD64, then it might also
** be necessary to define _XOPEN_SOURCE to be 500.  This varies from
** one system to another.  Since SQLite does not define USE_PREAD
** in any form by default, we will not attempt to define _XOPEN_SOURCE.
** See tickets #2741 and #2681.
**
** To avoid stomping the errno value on a failed read the lastErrno value
** is set before returning.
*/
static int seekAndRead(unixFile *id, sqlite3_int64 offset, void *pBuf, int cnt){
  int got;
  int prior = 0;
#if (!defined(USE_PREAD) && !defined(USE_PREAD64))
  i64 newOffset;
#endif
  TIMER_START;
  assert( cnt==(cnt&0x1ffff) );
  assert( id->h>2 );
  do{
#if defined(USE_PREAD)
    got = osPread(id->h, pBuf, cnt, offset);
    SimulateIOError( got = -1 );
#elif defined(USE_PREAD64)
    got = osPread64(id->h, pBuf, cnt, offset);
    SimulateIOError( got = -1 );
#else
    newOffset = lseek(id->h, offset, SEEK_SET);
    SimulateIOError( newOffset = -1 );
    if( newOffset<0 ){
      storeLastErrno((unixFile*)id, errno);
      return -1;
    }
    got = osRead(id->h, pBuf, cnt);
#endif
    if( got==cnt ) break;
    if( got<0 ){
      if( errno==EINTR ){ got = 1; continue; }
      prior = 0;
      storeLastErrno((unixFile*)id,  errno);
      break;
    }else if( got>0 ){
      cnt -= got;
      offset += got;
      prior += got;
      pBuf = (void*)(got + (char*)pBuf);
    }
  }while( got>0 );
  TIMER_END;
  OSTRACE(("READ    %-3d %5d %7lld %llu\n",
            id->h, got+prior, offset-prior, TIMER_ELAPSED));
  return got+prior;
}

/*
** Read data from a file into a buffer.  Return SQLITE_OK if all
** bytes were read successfully and SQLITE_IOERR if anything goes
** wrong.
*/
static int unixRead(
  sqlite3_file *id, 
  void *pBuf, 
  int amt,
  sqlite3_int64 offset
){
  unixFile *pFile = (unixFile *)id;
  int got;
  assert( id );
  assert( offset>=0 );
  assert( amt>0 );

  /* If this is a database file (not a journal, master-journal or temp
  ** file), the bytes in the locking range should never be read or written. */
#if 0
  assert( pFile->pPreallocatedUnused==0
       || offset>=PENDING_BYTE+512
       || offset+amt<=PENDING_BYTE 
  );
#endif

#if SQLITE_MAX_MMAP_SIZE>0
  /* Deal with as much of this read request as possible by transfering
  ** data from the memory mapping using memcpy().  */
  if( offset<pFile->mmapSize ){
    if( offset+amt <= pFile->mmapSize ){
      memcpy(pBuf, &((u8 *)(pFile->pMapRegion))[offset], amt);
      return SQLITE_OK;
    }else{
      int nCopy = pFile->mmapSize - offset;
      memcpy(pBuf, &((u8 *)(pFile->pMapRegion))[offset], nCopy);
      pBuf = &((u8 *)pBuf)[nCopy];
      amt -= nCopy;
      offset += nCopy;
    }
  }
#endif

  got = seekAndRead(pFile, offset, pBuf, amt);
  if( got==amt ){
    return SQLITE_OK;
  }else if( got<0 ){
    /* lastErrno set by seekAndRead */
    return SQLITE_IOERR_READ;
  }else{
    storeLastErrno(pFile, 0);   /* not a system error */
    /* Unread parts of the buffer must be zero-filled */
    memset(&((char*)pBuf)[got], 0, amt-got);
    return SQLITE_IOERR_SHORT_READ;
  }
}

/*
** Attempt to seek the file-descriptor passed as the first argument to
** absolute offset iOff, then attempt to write nBuf bytes of data from
** pBuf to it. If an error occurs, return -1 and set *piErrno. Otherwise, 
** return the actual number of bytes written (which may be less than
** nBuf).
*/
static int seekAndWriteFd(
  int fd,                         /* File descriptor to write to */
  i64 iOff,                       /* File offset to begin writing at */
  const void *pBuf,               /* Copy data from this buffer to the file */
  int nBuf,                       /* Size of buffer pBuf in bytes */
  int *piErrno                    /* OUT: Error number if error occurs */
){
  int rc = 0;                     /* Value returned by system call */

  assert( nBuf==(nBuf&0x1ffff) );
  assert( fd>2 );
  assert( piErrno!=0 );
  nBuf &= 0x1ffff;
  TIMER_START;

#if defined(USE_PREAD)
  do{ rc = (int)osPwrite(fd, pBuf, nBuf, iOff); }while( rc<0 && errno==EINTR );
#elif defined(USE_PREAD64)
  do{ rc = (int)osPwrite64(fd, pBuf, nBuf, iOff);}while( rc<0 && errno==EINTR);
#else
  do{
    i64 iSeek = lseek(fd, iOff, SEEK_SET);
    SimulateIOError( iSeek = -1 );
    if( iSeek<0 ){
      rc = -1;
      break;
    }
    rc = osWrite(fd, pBuf, nBuf);
  }while( rc<0 && errno==EINTR );
#endif

  TIMER_END;
  OSTRACE(("WRITE   %-3d %5d %7lld %llu\n", fd, rc, iOff, TIMER_ELAPSED));

  if( rc<0 ) *piErrno = errno;
  return rc;
}


/*
** Seek to the offset in id->offset then read cnt bytes into pBuf.
** Return the number of bytes actually read.  Update the offset.
**
** To avoid stomping the errno value on a failed write the lastErrno value
** is set before returning.
*/
static int seekAndWrite(unixFile *id, i64 offset, const void *pBuf, int cnt){
  return seekAndWriteFd(id->h, offset, pBuf, cnt, &id->lastErrno);
}


/*
** Write data from a buffer into a file.  Return SQLITE_OK on success
** or some other error code on failure.
*/
static int unixWrite(
  sqlite3_file *id, 
  const void *pBuf, 
  int amt,
  sqlite3_int64 offset 
){
  unixFile *pFile = (unixFile*)id;
  int wrote = 0;
  assert( id );
  assert( amt>0 );

  /* If this is a database file (not a journal, master-journal or temp
  ** file), the bytes in the locking range should never be read or written. */
#if 0
  assert( pFile->pPreallocatedUnused==0
       || offset>=PENDING_BYTE+512
       || offset+amt<=PENDING_BYTE 
  );
#endif

#ifdef SQLITE_DEBUG
  /* If we are doing a normal write to a database file (as opposed to
  ** doing a hot-journal rollback or a write to some file other than a
  ** normal database file) then record the fact that the database
  ** has changed.  If the transaction counter is modified, record that
  ** fact too.
  */
  if( pFile->inNormalWrite ){
    pFile->dbUpdate = 1;  /* The database has been modified */
    if( offset<=24 && offset+amt>=27 ){
      int rc;
      char oldCntr[4];
      SimulateIOErrorBenign(1);
      rc = seekAndRead(pFile, 24, oldCntr, 4);
      SimulateIOErrorBenign(0);
      if( rc!=4 || memcmp(oldCntr, &((char*)pBuf)[24-offset], 4)!=0 ){
        pFile->transCntrChng = 1;  /* The transaction counter has changed */
      }
    }
  }
#endif

#if defined(SQLITE_MMAP_READWRITE) && SQLITE_MAX_MMAP_SIZE>0
  /* Deal with as much of this write request as possible by transfering
  ** data from the memory mapping using memcpy().  */
  if( offset<pFile->mmapSize ){
    if( offset+amt <= pFile->mmapSize ){
      memcpy(&((u8 *)(pFile->pMapRegion))[offset], pBuf, amt);
      return SQLITE_OK;
    }else{
      int nCopy = pFile->mmapSize - offset;
      memcpy(&((u8 *)(pFile->pMapRegion))[offset], pBuf, nCopy);
      pBuf = &((u8 *)pBuf)[nCopy];
      amt -= nCopy;
      offset += nCopy;
    }
  }
#endif
 
  while( (wrote = seekAndWrite(pFile, offset, pBuf, amt))<amt && wrote>0 ){
    amt -= wrote;
    offset += wrote;
    pBuf = &((char*)pBuf)[wrote];
  }
  SimulateIOError(( wrote=(-1), amt=1 ));
  SimulateDiskfullError(( wrote=0, amt=1 ));

  if( amt>wrote ){
    if( wrote<0 && pFile->lastErrno!=ENOSPC ){
      /* lastErrno set by seekAndWrite */
      return SQLITE_IOERR_WRITE;
    }else{
      storeLastErrno(pFile, 0); /* not a system error */
      return SQLITE_FULL;
    }
  }

  return SQLITE_OK;
}

#ifdef SQLITE_TEST
/*
** Count the number of fullsyncs and normal syncs.  This is used to test
** that syncs and fullsyncs are occurring at the right times.
*/
int sqlite3_sync_count = 0;
int sqlite3_fullsync_count = 0;
#endif

/*
** We do not trust systems to provide a working fdatasync().  Some do.
** Others do no.  To be safe, we will stick with the (slightly slower)
** fsync(). If you know that your system does support fdatasync() correctly,
** then simply compile with -Dfdatasync=fdatasync or -DHAVE_FDATASYNC
*/
#if !defined(fdatasync) && !HAVE_FDATASYNC
# define fdatasync fsync
#endif

/*
** Define HAVE_FULLFSYNC to 0 or 1 depending on whether or not
** the F_FULLFSYNC macro is defined.  F_FULLFSYNC is currently
** only available on Mac OS X.  But that could change.
*/
#ifdef F_FULLFSYNC
# define HAVE_FULLFSYNC 1
#else
# define HAVE_FULLFSYNC 0
#endif


/*
** The fsync() system call does not work as advertised on many
** unix systems.  The following procedure is an attempt to make
** it work better.
**
** The SQLITE_NO_SYNC macro disables all fsync()s.  This is useful
** for testing when we want to run through the test suite quickly.
** You are strongly advised *not* to deploy with SQLITE_NO_SYNC
** enabled, however, since with SQLITE_NO_SYNC enabled, an OS crash
** or power failure will likely corrupt the database file.
**
** SQLite sets the dataOnly flag if the size of the file is unchanged.
** The idea behind dataOnly is that it should only write the file content
** to disk, not the inode.  We only set dataOnly if the file size is 
** unchanged since the file size is part of the inode.  However, 
** Ted Ts'o tells us that fdatasync() will also write the inode if the
** file size has changed.  The only real difference between fdatasync()
** and fsync(), Ted tells us, is that fdatasync() will not flush the
** inode if the mtime or owner or other inode attributes have changed.
** We only care about the file size, not the other file attributes, so
** as far as SQLite is concerned, an fdatasync() is always adequate.
** So, we always use fdatasync() if it is available, regardless of
** the value of the dataOnly flag.
*/
static int full_fsync(int fd, int fullSync, int dataOnly){
  int rc;

  /* The following "ifdef/elif/else/" block has the same structure as
  ** the one below. It is replicated here solely to avoid cluttering 
  ** up the real code with the UNUSED_PARAMETER() macros.
  */
#ifdef SQLITE_NO_SYNC
  UNUSED_PARAMETER(fd);
  UNUSED_PARAMETER(fullSync);
  UNUSED_PARAMETER(dataOnly);
#elif HAVE_FULLFSYNC
  UNUSED_PARAMETER(dataOnly);
#else
  UNUSED_PARAMETER(fullSync);
  UNUSED_PARAMETER(dataOnly);
#endif

  /* Record the number of times that we do a normal fsync() and 
  ** FULLSYNC.  This is used during testing to verify that this procedure
  ** gets called with the correct arguments.
  */
#ifdef SQLITE_TEST
  if( fullSync ) sqlite3_fullsync_count++;
  sqlite3_sync_count++;
#endif

  /* If we compiled with the SQLITE_NO_SYNC flag, then syncing is a
  ** no-op.  But go ahead and call fstat() to validate the file
  ** descriptor as we need a method to provoke a failure during
  ** coverate testing.
  */
#ifdef SQLITE_NO_SYNC
  {
    struct stat buf;
    rc = osFstat(fd, &buf);
  }
#elif HAVE_FULLFSYNC
  if( fullSync ){
    rc = osFcntl(fd, F_FULLFSYNC, 0);
  }else{
    rc = 1;
  }
  /* If the FULLFSYNC failed, fall back to attempting an fsync().
  ** It shouldn't be possible for fullfsync to fail on the local 
  ** file system (on OSX), so failure indicates that FULLFSYNC
  ** isn't supported for this file system. So, attempt an fsync 
  ** and (for now) ignore the overhead of a superfluous fcntl call.  
  ** It'd be better to detect fullfsync support once and avoid 
  ** the fcntl call every time sync is called.
  */
  if( rc ) rc = fsync(fd);

#elif defined(__APPLE__)
  /* fdatasync() on HFS+ doesn't yet flush the file size if it changed correctly
  ** so currently we default to the macro that redefines fdatasync to fsync
  */
  rc = fsync(fd);
#else 
  rc = fdatasync(fd);
#if OS_VXWORKS
  if( rc==-1 && errno==ENOTSUP ){
    rc = fsync(fd);
  }
#endif /* OS_VXWORKS */
#endif /* ifdef SQLITE_NO_SYNC elif HAVE_FULLFSYNC */

  if( OS_VXWORKS && rc!= -1 ){
    rc = 0;
  }
  return rc;
}

/*
** Open a file descriptor to the directory containing file zFilename.
** If successful, *pFd is set to the opened file descriptor and
** SQLITE_OK is returned. If an error occurs, either SQLITE_NOMEM
** or SQLITE_CANTOPEN is returned and *pFd is set to an undefined
** value.
**
** The directory file descriptor is used for only one thing - to
** fsync() a directory to make sure file creation and deletion events
** are flushed to disk.  Such fsyncs are not needed on newer
** journaling filesystems, but are required on older filesystems.
**
** This routine can be overridden using the xSetSysCall interface.
** The ability to override this routine was added in support of the
** chromium sandbox.  Opening a directory is a security risk (we are
** told) so making it overrideable allows the chromium sandbox to
** replace this routine with a harmless no-op.  To make this routine
** a no-op, replace it with a stub that returns SQLITE_OK but leaves
** *pFd set to a negative number.
**
** If SQLITE_OK is returned, the caller is responsible for closing
** the file descriptor *pFd using close().
*/
static int openDirectory(const char *zFilename, int *pFd){
  int ii;
  int fd = -1;
  char zDirname[MAX_PATHNAME+1];

  sqlite3_snprintf(MAX_PATHNAME, zDirname, "%s", zFilename);
  for(ii=(int)strlen(zDirname); ii>0 && zDirname[ii]!='/'; ii--);
  if( ii>0 ){
    zDirname[ii] = '\0';
  }else{
    if( zDirname[0]!='/' ) zDirname[0] = '.';
    zDirname[1] = 0;
  }
  fd = robust_open(zDirname, O_RDONLY|O_BINARY, 0);
  if( fd>=0 ){
    OSTRACE(("OPENDIR %-3d %s\n", fd, zDirname));
  }
  *pFd = fd;
  if( fd>=0 ) return SQLITE_OK;
  return unixLogError(SQLITE_CANTOPEN_BKPT, "openDirectory", zDirname);
}

/*
** Make sure all writes to a particular file are committed to disk.
**
** If dataOnly==0 then both the file itself and its metadata (file
** size, access time, etc) are synced.  If dataOnly!=0 then only the
** file data is synced.
**
** Under Unix, also make sure that the directory entry for the file
** has been created by fsync-ing the directory that contains the file.
** If we do not do this and we encounter a power failure, the directory
** entry for the journal might not exist after we reboot.  The next
** SQLite to access the file will not know that the journal exists (because
** the directory entry for the journal was never created) and the transaction
** will not roll back - possibly leading to database corruption.
*/
static int unixSync(sqlite3_file *id, int flags){
  int rc;
  unixFile *pFile = (unixFile*)id;

  int isDataOnly = (flags&SQLITE_SYNC_DATAONLY);
  int isFullsync = (flags&0x0F)==SQLITE_SYNC_FULL;

  /* Check that one of SQLITE_SYNC_NORMAL or FULL was passed */
  assert((flags&0x0F)==SQLITE_SYNC_NORMAL
      || (flags&0x0F)==SQLITE_SYNC_FULL
  );

  /* Unix cannot, but some systems may return SQLITE_FULL from here. This
  ** line is to test that doing so does not cause any problems.
  */
  SimulateDiskfullError( return SQLITE_FULL );

  assert( pFile );
  OSTRACE(("SYNC    %-3d\n", pFile->h));
  rc = full_fsync(pFile->h, isFullsync, isDataOnly);
  SimulateIOError( rc=1 );
  if( rc ){
    storeLastErrno(pFile, errno);
    return unixLogError(SQLITE_IOERR_FSYNC, "full_fsync", pFile->zPath);
  }

  /* Also fsync the directory containing the file if the DIRSYNC flag
  ** is set.  This is a one-time occurrence.  Many systems (examples: AIX)
  ** are unable to fsync a directory, so ignore errors on the fsync.
  */
  if( pFile->ctrlFlags & UNIXFILE_DIRSYNC ){
    int dirfd;
    OSTRACE(("DIRSYNC %s (have_fullfsync=%d fullsync=%d)\n", pFile->zPath,
            HAVE_FULLFSYNC, isFullsync));
    rc = osOpenDirectory(pFile->zPath, &dirfd);
    if( rc==SQLITE_OK ){
      full_fsync(dirfd, 0, 0);
      robust_close(pFile, dirfd, __LINE__);
    }else{
      assert( rc==SQLITE_CANTOPEN );
      rc = SQLITE_OK;
    }
    pFile->ctrlFlags &= ~UNIXFILE_DIRSYNC;
  }
  return rc;
}

/*
** Truncate an open file to a specified size
*/
static int unixTruncate(sqlite3_file *id, i64 nByte){
  unixFile *pFile = (unixFile *)id;
  int rc;
  assert( pFile );
  SimulateIOError( return SQLITE_IOERR_TRUNCATE );

  /* If the user has configured a chunk-size for this file, truncate the
  ** file so that it consists of an integer number of chunks (i.e. the
  ** actual file size after the operation may be larger than the requested
  ** size).
  */
  if( pFile->szChunk>0 ){
    nByte = ((nByte + pFile->szChunk - 1)/pFile->szChunk) * pFile->szChunk;
  }

  rc = robust_ftruncate(pFile->h, nByte);
  if( rc ){
    storeLastErrno(pFile, errno);
    return unixLogError(SQLITE_IOERR_TRUNCATE, "ftruncate", pFile->zPath);
  }else{
#ifdef SQLITE_DEBUG
    /* If we are doing a normal write to a database file (as opposed to
    ** doing a hot-journal rollback or a write to some file other than a
    ** normal database file) and we truncate the file to zero length,
    ** that effectively updates the change counter.  This might happen
    ** when restoring a database using the backup API from a zero-length
    ** source.
    */
    if( pFile->inNormalWrite && nByte==0 ){
      pFile->transCntrChng = 1;
    }
#endif

#if SQLITE_MAX_MMAP_SIZE>0
    /* If the file was just truncated to a size smaller than the currently
    ** mapped region, reduce the effective mapping size as well. SQLite will
    ** use read() and write() to access data beyond this point from now on.  
    */
    if( nByte<pFile->mmapSize ){
      pFile->mmapSize = nByte;
    }
#endif

    return SQLITE_OK;
  }
}

/*
** Determine the current size of a file in bytes
*/
static int unixFileSize(sqlite3_file *id, i64 *pSize){
  int rc;
  struct stat buf;
  assert( id );
  rc = osFstat(((unixFile*)id)->h, &buf);
  SimulateIOError( rc=1 );
  if( rc!=0 ){
    storeLastErrno((unixFile*)id, errno);
    return SQLITE_IOERR_FSTAT;
  }
  *pSize = buf.st_size;

  /* When opening a zero-size database, the findInodeInfo() procedure
  ** writes a single byte into that file in order to work around a bug
  ** in the OS-X msdos filesystem.  In order to avoid problems with upper
  ** layers, we need to report this file size as zero even though it is
  ** really 1.   Ticket #3260.
  */
  if( *pSize==1 ) *pSize = 0;


  return SQLITE_OK;
}

#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
/*
** Handler for proxy-locking file-control verbs.  Defined below in the
** proxying locking division.
*/
static int proxyFileControl(sqlite3_file*,int,void*);
#endif

/* 
** This function is called to handle the SQLITE_FCNTL_SIZE_HINT 
** file-control operation.  Enlarge the database to nBytes in size
** (rounded up to the next chunk-size).  If the database is already
** nBytes or larger, this routine is a no-op.
*/
static int fcntlSizeHint(unixFile *pFile, i64 nByte){
  if( pFile->szChunk>0 ){
    i64 nSize;                    /* Required file size */
    struct stat buf;              /* Used to hold return values of fstat() */
   
    if( osFstat(pFile->h, &buf) ){
      return SQLITE_IOERR_FSTAT;
    }

    nSize = ((nByte+pFile->szChunk-1) / pFile->szChunk) * pFile->szChunk;
    if( nSize>(i64)buf.st_size ){

#if defined(HAVE_POSIX_FALLOCATE) && HAVE_POSIX_FALLOCATE
      /* The code below is handling the return value of osFallocate() 
      ** correctly. posix_fallocate() is defined to "returns zero on success, 
      ** or an error number on  failure". See the manpage for details. */
      int err;
      do{
        err = osFallocate(pFile->h, buf.st_size, nSize-buf.st_size);
      }while( err==EINTR );
      if( err && err!=EINVAL ) return SQLITE_IOERR_WRITE;
#else
      /* If the OS does not have posix_fallocate(), fake it. Write a 
      ** single byte to the last byte in each block that falls entirely
      ** within the extended region. Then, if required, a single byte
      ** at offset (nSize-1), to set the size of the file correctly.
      ** This is a similar technique to that used by glibc on systems
      ** that do not have a real fallocate() call.
      */
      int nBlk = buf.st_blksize;  /* File-system block size */
      int nWrite = 0;             /* Number of bytes written by seekAndWrite */
      i64 iWrite;                 /* Next offset to write to */

      iWrite = (buf.st_size/nBlk)*nBlk + nBlk - 1;
      assert( iWrite>=buf.st_size );
      assert( ((iWrite+1)%nBlk)==0 );
      for(/*no-op*/; iWrite<nSize+nBlk-1; iWrite+=nBlk ){
        if( iWrite>=nSize ) iWrite = nSize - 1;
        nWrite = seekAndWrite(pFile, iWrite, "", 1);
        if( nWrite!=1 ) return SQLITE_IOERR_WRITE;
      }
#endif
    }
  }

#if SQLITE_MAX_MMAP_SIZE>0
  if( pFile->mmapSizeMax>0 && nByte>pFile->mmapSize ){
    int rc;
    if( pFile->szChunk<=0 ){
      if( robust_ftruncate(pFile->h, nByte) ){
        storeLastErrno(pFile, errno);
        return unixLogError(SQLITE_IOERR_TRUNCATE, "ftruncate", pFile->zPath);
      }
    }

    rc = unixMapfile(pFile, nByte);
    return rc;
  }
#endif

  return SQLITE_OK;
}

/*
** If *pArg is initially negative then this is a query.  Set *pArg to
** 1 or 0 depending on whether or not bit mask of pFile->ctrlFlags is set.
**
** If *pArg is 0 or 1, then clear or set the mask bit of pFile->ctrlFlags.
*/
static void unixModeBit(unixFile *pFile, unsigned char mask, int *pArg){
  if( *pArg<0 ){
    *pArg = (pFile->ctrlFlags & mask)!=0;
  }else if( (*pArg)==0 ){
    pFile->ctrlFlags &= ~mask;
  }else{
    pFile->ctrlFlags |= mask;
  }
}

/* Forward declaration */
static int unixGetTempname(int nBuf, char *zBuf);

/*
** Information and control of an open file handle.
*/
static int unixFileControl(sqlite3_file *id, int op, void *pArg){
  unixFile *pFile = (unixFile*)id;
  switch( op ){
#if defined(__linux__) && defined(SQLITE_ENABLE_BATCH_ATOMIC_WRITE)
    case SQLITE_FCNTL_BEGIN_ATOMIC_WRITE: {
      int rc = osIoctl(pFile->h, F2FS_IOC_START_ATOMIC_WRITE);
      return rc ? SQLITE_IOERR_BEGIN_ATOMIC : SQLITE_OK;
    }
    case SQLITE_FCNTL_COMMIT_ATOMIC_WRITE: {
      int rc = osIoctl(pFile->h, F2FS_IOC_COMMIT_ATOMIC_WRITE);
      return rc ? SQLITE_IOERR_COMMIT_ATOMIC : SQLITE_OK;
    }
    case SQLITE_FCNTL_ROLLBACK_ATOMIC_WRITE: {
      int rc = osIoctl(pFile->h, F2FS_IOC_ABORT_VOLATILE_WRITE);
      return rc ? SQLITE_IOERR_ROLLBACK_ATOMIC : SQLITE_OK;
    }
#endif /* __linux__ && SQLITE_ENABLE_BATCH_ATOMIC_WRITE */

    case SQLITE_FCNTL_LOCKSTATE: {
      *(int*)pArg = pFile->eFileLock;
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_LAST_ERRNO: {
      *(int*)pArg = pFile->lastErrno;
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_CHUNK_SIZE: {
      pFile->szChunk = *(int *)pArg;
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_SIZE_HINT: {
      int rc;
      SimulateIOErrorBenign(1);
      rc = fcntlSizeHint(pFile, *(i64 *)pArg);
      SimulateIOErrorBenign(0);
      return rc;
    }
    case SQLITE_FCNTL_PERSIST_WAL: {
      unixModeBit(pFile, UNIXFILE_PERSIST_WAL, (int*)pArg);
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_POWERSAFE_OVERWRITE: {
      unixModeBit(pFile, UNIXFILE_PSOW, (int*)pArg);
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_VFSNAME: {
      *(char**)pArg = sqlite3_mprintf("%s", pFile->pVfs->zName);
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_TEMPFILENAME: {
      char *zTFile = sqlite3_malloc64( pFile->pVfs->mxPathname );
      if( zTFile ){
        unixGetTempname(pFile->pVfs->mxPathname, zTFile);
        *(char**)pArg = zTFile;
      }
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_HAS_MOVED: {
      *(int*)pArg = fileHasMoved(pFile);
      return SQLITE_OK;
    }
#ifdef SQLITE_ENABLE_SETLK_TIMEOUT
    case SQLITE_FCNTL_LOCK_TIMEOUT: {
      pFile->iBusyTimeout = *(int*)pArg;
      return SQLITE_OK;
    }
#endif
#if SQLITE_MAX_MMAP_SIZE>0
    case SQLITE_FCNTL_MMAP_SIZE: {
      i64 newLimit = *(i64*)pArg;
      int rc = SQLITE_OK;
      if( newLimit>sqlite3GlobalConfig.mxMmap ){
        newLimit = sqlite3GlobalConfig.mxMmap;
      }

      /* The value of newLimit may be eventually cast to (size_t) and passed
      ** to mmap(). Restrict its value to 2GB if (size_t) is not at least a
      ** 64-bit type. */
      if( newLimit>0 && sizeof(size_t)<8 ){
        newLimit = (newLimit & 0x7FFFFFFF);
      }

      *(i64*)pArg = pFile->mmapSizeMax;
      if( newLimit>=0 && newLimit!=pFile->mmapSizeMax && pFile->nFetchOut==0 ){
        pFile->mmapSizeMax = newLimit;
        if( pFile->mmapSize>0 ){
          unixUnmapfile(pFile);
          rc = unixMapfile(pFile, -1);
        }
      }
      return rc;
    }
#endif
#ifdef SQLITE_DEBUG
    /* The pager calls this method to signal that it has done
    ** a rollback and that the database is therefore unchanged and
    ** it hence it is OK for the transaction change counter to be
    ** unchanged.
    */
    case SQLITE_FCNTL_DB_UNCHANGED: {
      ((unixFile*)id)->dbUpdate = 0;
      return SQLITE_OK;
    }
#endif
#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
    case SQLITE_FCNTL_SET_LOCKPROXYFILE:
    case SQLITE_FCNTL_GET_LOCKPROXYFILE: {
      return proxyFileControl(id,op,pArg);
    }
#endif /* SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__) */
  }
  return SQLITE_NOTFOUND;
}

/*
** If pFd->sectorSize is non-zero when this function is called, it is a
** no-op. Otherwise, the values of pFd->sectorSize and 
** pFd->deviceCharacteristics are set according to the file-system 
** characteristics. 
**
** There are two versions of this function. One for QNX and one for all
** other systems.
*/
#ifndef __QNXNTO__
static void setDeviceCharacteristics(unixFile *pFd){
  assert( pFd->deviceCharacteristics==0 || pFd->sectorSize!=0 );
  if( pFd->sectorSize==0 ){
#if defined(__linux__) && defined(SQLITE_ENABLE_BATCH_ATOMIC_WRITE)
    int res;
    u32 f = 0;

    /* Check for support for F2FS atomic batch writes. */
    res = osIoctl(pFd->h, F2FS_IOC_GET_FEATURES, &f);
    if( res==0 && (f & F2FS_FEATURE_ATOMIC_WRITE) ){
      pFd->deviceCharacteristics = SQLITE_IOCAP_BATCH_ATOMIC;
    }
#endif /* __linux__ && SQLITE_ENABLE_BATCH_ATOMIC_WRITE */

    /* Set the POWERSAFE_OVERWRITE flag if requested. */
    if( pFd->ctrlFlags & UNIXFILE_PSOW ){
      pFd->deviceCharacteristics |= SQLITE_IOCAP_POWERSAFE_OVERWRITE;
    }

    pFd->sectorSize = SQLITE_DEFAULT_SECTOR_SIZE;
  }
}
#else
#include <sys/dcmd_blk.h>
#include <sys/statvfs.h>
static void setDeviceCharacteristics(unixFile *pFile){
  if( pFile->sectorSize == 0 ){
    struct statvfs fsInfo;
       
    /* Set defaults for non-supported filesystems */
    pFile->sectorSize = SQLITE_DEFAULT_SECTOR_SIZE;
    pFile->deviceCharacteristics = 0;
    if( fstatvfs(pFile->h, &fsInfo) == -1 ) {
      return;
    }

    if( !strcmp(fsInfo.f_basetype, "tmp") ) {
      pFile->sectorSize = fsInfo.f_bsize;
      pFile->deviceCharacteristics =
        SQLITE_IOCAP_ATOMIC4K |       /* All ram filesystem writes are atomic */
        SQLITE_IOCAP_SAFE_APPEND |    /* growing the file does not occur until
                                      ** the write succeeds */
        SQLITE_IOCAP_SEQUENTIAL |     /* The ram filesystem has no write behind
                                      ** so it is ordered */
        0;
    }else if( strstr(fsInfo.f_basetype, "etfs") ){
      pFile->sectorSize = fsInfo.f_bsize;
      pFile->deviceCharacteristics =
        /* etfs cluster size writes are atomic */
        (pFile->sectorSize / 512 * SQLITE_IOCAP_ATOMIC512) |
        SQLITE_IOCAP_SAFE_APPEND |    /* growing the file does not occur until
                                      ** the write succeeds */
        SQLITE_IOCAP_SEQUENTIAL |     /* The ram filesystem has no write behind
                                      ** so it is ordered */
        0;
    }else if( !strcmp(fsInfo.f_basetype, "qnx6") ){
      pFile->sectorSize = fsInfo.f_bsize;
      pFile->deviceCharacteristics =
        SQLITE_IOCAP_ATOMIC |         /* All filesystem writes are atomic */
        SQLITE_IOCAP_SAFE_APPEND |    /* growing the file does not occur until
                                      ** the write succeeds */
        SQLITE_IOCAP_SEQUENTIAL |     /* The ram filesystem has no write behind
                                      ** so it is ordered */
        0;
    }else if( !strcmp(fsInfo.f_basetype, "qnx4") ){
      pFile->sectorSize = fsInfo.f_bsize;
      pFile->deviceCharacteristics =
        /* full bitset of atomics from max sector size and smaller */
        ((pFile->sectorSize / 512 * SQLITE_IOCAP_ATOMIC512) << 1) - 2 |
        SQLITE_IOCAP_SEQUENTIAL |     /* The ram filesystem has no write behind
                                      ** so it is ordered */
        0;
    }else if( strstr(fsInfo.f_basetype, "dos") ){
      pFile->sectorSize = fsInfo.f_bsize;
      pFile->deviceCharacteristics =
        /* full bitset of atomics from max sector size and smaller */
        ((pFile->sectorSize / 512 * SQLITE_IOCAP_ATOMIC512) << 1) - 2 |
        SQLITE_IOCAP_SEQUENTIAL |     /* The ram filesystem has no write behind
                                      ** so it is ordered */
        0;
    }else{
      pFile->deviceCharacteristics =
        SQLITE_IOCAP_ATOMIC512 |      /* blocks are atomic */
        SQLITE_IOCAP_SAFE_APPEND |    /* growing the file does not occur until
                                      ** the write succeeds */
        0;
    }
  }
  /* Last chance verification.  If the sector size isn't a multiple of 512
  ** then it isn't valid.*/
  if( pFile->sectorSize % 512 != 0 ){
    pFile->deviceCharacteristics = 0;
    pFile->sectorSize = SQLITE_DEFAULT_SECTOR_SIZE;
  }
}
#endif

/*
** Return the sector size in bytes of the underlying block device for
** the specified file. This is almost always 512 bytes, but may be
** larger for some devices.
**
** SQLite code assumes this function cannot fail. It also assumes that
** if two files are created in the same file-system directory (i.e.
** a database and its journal file) that the sector size will be the
** same for both.
*/
static int unixSectorSize(sqlite3_file *id){
  unixFile *pFd = (unixFile*)id;
  setDeviceCharacteristics(pFd);
  return pFd->sectorSize;
}

/*
** Return the device characteristics for the file.
**
** This VFS is set up to return SQLITE_IOCAP_POWERSAFE_OVERWRITE by default.
** However, that choice is controversial since technically the underlying
** file system does not always provide powersafe overwrites.  (In other
** words, after a power-loss event, parts of the file that were never
** written might end up being altered.)  However, non-PSOW behavior is very,
** very rare.  And asserting PSOW makes a large reduction in the amount
** of required I/O for journaling, since a lot of padding is eliminated.
**  Hence, while POWERSAFE_OVERWRITE is on by default, there is a file-control
** available to turn it off and URI query parameter available to turn it off.
*/
static int unixDeviceCharacteristics(sqlite3_file *id){
  unixFile *pFd = (unixFile*)id;
  setDeviceCharacteristics(pFd);
  return pFd->deviceCharacteristics;
}

#if !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0

/*
** Return the system page size.
**
** This function should not be called directly by other code in this file. 
** Instead, it should be called via macro osGetpagesize().
*/
static int unixGetpagesize(void){
#if OS_VXWORKS
  return 1024;
#elif defined(_BSD_SOURCE)
  return getpagesize();
#else
  return (int)sysconf(_SC_PAGESIZE);
#endif
}

#endif /* !defined(SQLITE_OMIT_WAL) || SQLITE_MAX_MMAP_SIZE>0 */

#ifndef SQLITE_OMIT_WAL

/*
** Object used to represent an shared memory buffer.  
**
** When multiple threads all reference the same wal-index, each thread
** has its own unixShm object, but they all point to a single instance
** of this unixShmNode object.  In other words, each wal-index is opened
** only once per process.
**
** Each unixShmNode object is connected to a single unixInodeInfo object.
** We could coalesce this object into unixInodeInfo, but that would mean
** every open file that does not use shared memory (in other words, most
** open files) would have to carry around this extra information.  So
** the unixInodeInfo object contains a pointer to this unixShmNode object
** and the unixShmNode object is created only when needed.
**
** unixMutexHeld() must be true when creating or destroying
** this object or while reading or writing the following fields:
**
**      nRef
**
** The following fields are read-only after the object is created:
** 
**      fid
**      zFilename
**
** Either unixShmNode.mutex must be held or unixShmNode.nRef==0 and
** unixMutexHeld() is true when reading or writing any other field
** in this structure.
*/
struct unixShmNode {
  unixInodeInfo *pInode;     /* unixInodeInfo that owns this SHM node */
  sqlite3_mutex *mutex;      /* Mutex to access this object */
  char *zFilename;           /* Name of the mmapped file */
  int h;                     /* Open file descriptor */
  int szRegion;              /* Size of shared-memory regions */
  u16 nRegion;               /* Size of array apRegion */
  u8 isReadonly;             /* True if read-only */
  u8 isUnlocked;             /* True if no DMS lock held */
  char **apRegion;           /* Array of mapped shared-memory regions */
  int nRef;                  /* Number of unixShm objects pointing to this */
  unixShm *pFirst;           /* All unixShm objects pointing to this */
#ifdef SQLITE_DEBUG
  u8 exclMask;               /* Mask of exclusive locks held */
  u8 sharedMask;             /* Mask of shared locks held */
  u8 nextShmId;              /* Next available unixShm.id value */
#endif
};

/*
** Structure used internally by this VFS to record the state of an
** open shared memory connection.
**
** The following fields are initialized when this object is created and
** are read-only thereafter:
**
**    unixShm.pFile
**    unixShm.id
**
** All other fields are read/write.  The unixShm.pFile->mutex must be held
** while accessing any read/write fields.
*/
struct unixShm {
  unixShmNode *pShmNode;     /* The underlying unixShmNode object */
  unixShm *pNext;            /* Next unixShm with the same unixShmNode */
  u8 hasMutex;               /* True if holding the unixShmNode mutex */
  u8 id;                     /* Id of this connection within its unixShmNode */
  u16 sharedMask;            /* Mask of shared locks held */
  u16 exclMask;              /* Mask of exclusive locks held */
};

/*
** Constants used for locking
*/
#define UNIX_SHM_BASE   ((22+SQLITE_SHM_NLOCK)*4)         /* first lock byte */
#define UNIX_SHM_DMS    (UNIX_SHM_BASE+SQLITE_SHM_NLOCK)  /* deadman switch */

/*
** Apply posix advisory locks for all bytes from ofst through ofst+n-1.
**
** Locks block if the mask is exactly UNIX_SHM_C and are non-blocking
** otherwise.
*/
static int unixShmSystemLock(
  unixFile *pFile,       /* Open connection to the WAL file */
  int lockType,          /* F_UNLCK, F_RDLCK, or F_WRLCK */
  int ofst,              /* First byte of the locking range */
  int n                  /* Number of bytes to lock */
){
  unixShmNode *pShmNode; /* Apply locks to this open shared-memory segment */
  struct flock f;        /* The posix advisory locking structure */
  int rc = SQLITE_OK;    /* Result code form fcntl() */

  /* Access to the unixShmNode object is serialized by the caller */
  pShmNode = pFile->pInode->pShmNode;
  assert( pShmNode->nRef==0 || sqlite3_mutex_held(pShmNode->mutex) );

  /* Shared locks never span more than one byte */
  assert( n==1 || lockType!=F_RDLCK );

  /* Locks are within range */
  assert( n>=1 && n<=SQLITE_SHM_NLOCK );

  if( pShmNode->h>=0 ){
    /* Initialize the locking parameters */
    f.l_type = lockType;
    f.l_whence = SEEK_SET;
    f.l_start = ofst;
    f.l_len = n;
    rc = osSetPosixAdvisoryLock(pShmNode->h, &f, pFile);
    rc = (rc!=(-1)) ? SQLITE_OK : SQLITE_BUSY;
  }

  /* Update the global lock state and do debug tracing */
#ifdef SQLITE_DEBUG
  { u16 mask;
  OSTRACE(("SHM-LOCK "));
  mask = ofst>31 ? 0xffff : (1<<(ofst+n)) - (1<<ofst);
  if( rc==SQLITE_OK ){
    if( lockType==F_UNLCK ){
      OSTRACE(("unlock %d ok", ofst));
      pShmNode->exclMask &= ~mask;
      pShmNode->sharedMask &= ~mask;
    }else if( lockType==F_RDLCK ){
      OSTRACE(("read-lock %d ok", ofst));
      pShmNode->exclMask &= ~mask;
      pShmNode->sharedMask |= mask;
    }else{
      assert( lockType==F_WRLCK );
      OSTRACE(("write-lock %d ok", ofst));
      pShmNode->exclMask |= mask;
      pShmNode->sharedMask &= ~mask;
    }
  }else{
    if( lockType==F_UNLCK ){
      OSTRACE(("unlock %d failed", ofst));
    }else if( lockType==F_RDLCK ){
      OSTRACE(("read-lock failed"));
    }else{
      assert( lockType==F_WRLCK );
      OSTRACE(("write-lock %d failed", ofst));
    }
  }
  OSTRACE((" - afterwards %03x,%03x\n",
           pShmNode->sharedMask, pShmNode->exclMask));
  }
#endif

  return rc;        
}

/*
** Return the minimum number of 32KB shm regions that should be mapped at
** a time, assuming that each mapping must be an integer multiple of the
** current system page-size.
**
** Usually, this is 1. The exception seems to be systems that are configured
** to use 64KB pages - in this case each mapping must cover at least two
** shm regions.
*/
static int unixShmRegionPerMap(void){
  int shmsz = 32*1024;            /* SHM region size */
  int pgsz = osGetpagesize();   /* System page size */
  assert( ((pgsz-1)&pgsz)==0 );   /* Page size must be a power of 2 */
  if( pgsz<shmsz ) return 1;
  return pgsz/shmsz;
}

/*
** Purge the unixShmNodeList list of all entries with unixShmNode.nRef==0.
**
** This is not a VFS shared-memory method; it is a utility function called
** by VFS shared-memory methods.
*/
static void unixShmPurge(unixFile *pFd){
  unixShmNode *p = pFd->pInode->pShmNode;
  assert( unixMutexHeld() );
  if( p && ALWAYS(p->nRef==0) ){
    int nShmPerMap = unixShmRegionPerMap();
    int i;
    assert( p->pInode==pFd->pInode );
    sqlite3_mutex_free(p->mutex);
    for(i=0; i<p->nRegion; i+=nShmPerMap){
      if( p->h>=0 ){
        osMunmap(p->apRegion[i], p->szRegion);
      }else{
        sqlite3_free(p->apRegion[i]);
      }
    }
    sqlite3_free(p->apRegion);
    if( p->h>=0 ){
      robust_close(pFd, p->h, __LINE__);
      p->h = -1;
    }
    p->pInode->pShmNode = 0;
    sqlite3_free(p);
  }
}

/*
** The DMS lock has not yet been taken on shm file pShmNode. Attempt to
** take it now. Return SQLITE_OK if successful, or an SQLite error
** code otherwise.
**
** If the DMS cannot be locked because this is a readonly_shm=1 
** connection and no other process already holds a lock, return
** SQLITE_READONLY_CANTINIT and set pShmNode->isUnlocked=1.
*/
static int unixLockSharedMemory(unixFile *pDbFd, unixShmNode *pShmNode){
  struct flock lock;
  int rc = SQLITE_OK;

  /* Use F_GETLK to determine the locks other processes are holding
  ** on the DMS byte. If it indicates that another process is holding
  ** a SHARED lock, then this process may also take a SHARED lock
  ** and proceed with opening the *-shm file. 
  **
  ** Or, if no other process is holding any lock, then this process
  ** is the first to open it. In this case take an EXCLUSIVE lock on the
  ** DMS byte and truncate the *-shm file to zero bytes in size. Then
  ** downgrade to a SHARED lock on the DMS byte.
  **
  ** If another process is holding an EXCLUSIVE lock on the DMS byte,
  ** return SQLITE_BUSY to the caller (it will try again). An earlier
  ** version of this code attempted the SHARED lock at this point. But
  ** this introduced a subtle race condition: if the process holding
  ** EXCLUSIVE failed just before truncating the *-shm file, then this
  ** process might open and use the *-shm file without truncating it.
  ** And if the *-shm file has been corrupted by a power failure or
  ** system crash, the database itself may also become corrupt.  */
  lock.l_whence = SEEK_SET;
  lock.l_start = UNIX_SHM_DMS;
  lock.l_len = 1;
  lock.l_type = F_WRLCK;
  if( osFcntl(pShmNode->h, F_GETLK, &lock)!=0 ) {
    rc = SQLITE_IOERR_LOCK;
  }else if( lock.l_type==F_UNLCK ){
    if( pShmNode->isReadonly ){
      pShmNode->isUnlocked = 1;
      rc = SQLITE_READONLY_CANTINIT;
    }else{
      rc = unixShmSystemLock(pDbFd, F_WRLCK, UNIX_SHM_DMS, 1);
      if( rc==SQLITE_OK && robust_ftruncate(pShmNode->h, 0) ){
        rc = unixLogError(SQLITE_IOERR_SHMOPEN,"ftruncate",pShmNode->zFilename);
      }
    }
  }else if( lock.l_type==F_WRLCK ){
    rc = SQLITE_BUSY;
  }

  if( rc==SQLITE_OK ){
    assert( lock.l_type==F_UNLCK || lock.l_type==F_RDLCK );
    rc = unixShmSystemLock(pDbFd, F_RDLCK, UNIX_SHM_DMS, 1);
  }
  return rc;
}

/*
** Open a shared-memory area associated with open database file pDbFd.  
** This particular implementation uses mmapped files.
**
** The file used to implement shared-memory is in the same directory
** as the open database file and has the same name as the open database
** file with the "-shm" suffix added.  For example, if the database file
** is "/home/user1/config.db" then the file that is created and mmapped
** for shared memory will be called "/home/user1/config.db-shm".  
**
** Another approach to is to use files in /dev/shm or /dev/tmp or an
** some other tmpfs mount. But if a file in a different directory
** from the database file is used, then differing access permissions
** or a chroot() might cause two different processes on the same
** database to end up using different files for shared memory - 
** meaning that their memory would not really be shared - resulting
** in database corruption.  Nevertheless, this tmpfs file usage
** can be enabled at compile-time using -DSQLITE_SHM_DIRECTORY="/dev/shm"
** or the equivalent.  The use of the SQLITE_SHM_DIRECTORY compile-time
** option results in an incompatible build of SQLite;  builds of SQLite
** that with differing SQLITE_SHM_DIRECTORY settings attempt to use the
** same database file at the same time, database corruption will likely
** result. The SQLITE_SHM_DIRECTORY compile-time option is considered
** "unsupported" and may go away in a future SQLite release.
**
** When opening a new shared-memory file, if no other instances of that
** file are currently open, in this process or in other processes, then
** the file must be truncated to zero length or have its header cleared.
**
** If the original database file (pDbFd) is using the "unix-excl" VFS
** that means that an exclusive lock is held on the database file and
** that no other processes are able to read or write the database.  In
** that case, we do not really need shared memory.  No shared memory
** file is created.  The shared memory will be simulated with heap memory.
*/
static int unixOpenSharedMemory(unixFile *pDbFd){
  struct unixShm *p = 0;          /* The connection to be opened */
  struct unixShmNode *pShmNode;   /* The underlying mmapped file */
  int rc = SQLITE_OK;             /* Result code */
  unixInodeInfo *pInode;          /* The inode of fd */
  char *zShm;             /* Name of the file used for SHM */
  int nShmFilename;               /* Size of the SHM filename in bytes */

  /* Allocate space for the new unixShm object. */
  p = sqlite3_malloc64( sizeof(*p) );
  if( p==0 ) return SQLITE_NOMEM_BKPT;
  memset(p, 0, sizeof(*p));
  assert( pDbFd->pShm==0 );

  /* Check to see if a unixShmNode object already exists. Reuse an existing
  ** one if present. Create a new one if necessary.
  */
  unixEnterMutex();
  pInode = pDbFd->pInode;
  pShmNode = pInode->pShmNode;
  if( pShmNode==0 ){
    struct stat sStat;                 /* fstat() info for database file */
#ifndef SQLITE_SHM_DIRECTORY
    const char *zBasePath = pDbFd->zPath;
#endif

    /* Call fstat() to figure out the permissions on the database file. If
    ** a new *-shm file is created, an attempt will be made to create it
    ** with the same permissions.
    */
    if( osFstat(pDbFd->h, &sStat) ){
      rc = SQLITE_IOERR_FSTAT;
      goto shm_open_err;
    }

#ifdef SQLITE_SHM_DIRECTORY
    nShmFilename = sizeof(SQLITE_SHM_DIRECTORY) + 31;
#else
    nShmFilename = 6 + (int)strlen(zBasePath);
#endif
    pShmNode = sqlite3_malloc64( sizeof(*pShmNode) + nShmFilename );
    if( pShmNode==0 ){
      rc = SQLITE_NOMEM_BKPT;
      goto shm_open_err;
    }
    memset(pShmNode, 0, sizeof(*pShmNode)+nShmFilename);
    zShm = pShmNode->zFilename = (char*)&pShmNode[1];
#ifdef SQLITE_SHM_DIRECTORY
    sqlite3_snprintf(nShmFilename, zShm, 
                     SQLITE_SHM_DIRECTORY "/sqlite-shm-%x-%x",
                     (u32)sStat.st_ino, (u32)sStat.st_dev);
#else
    sqlite3_snprintf(nShmFilename, zShm, "%s-shm", zBasePath);
    sqlite3FileSuffix3(pDbFd->zPath, zShm);
#endif
    pShmNode->h = -1;
    pDbFd->pInode->pShmNode = pShmNode;
    pShmNode->pInode = pDbFd->pInode;
    if( sqlite3GlobalConfig.bCoreMutex ){
      pShmNode->mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST);
      if( pShmNode->mutex==0 ){
        rc = SQLITE_NOMEM_BKPT;
        goto shm_open_err;
      }
    }

    if( pInode->bProcessLock==0 ){
      if( 0==sqlite3_uri_boolean(pDbFd->zPath, "readonly_shm", 0) ){
        pShmNode->h = robust_open(zShm, O_RDWR|O_CREAT, (sStat.st_mode&0777));
      }
      if( pShmNode->h<0 ){
        pShmNode->h = robust_open(zShm, O_RDONLY, (sStat.st_mode&0777));
        if( pShmNode->h<0 ){
          rc = unixLogError(SQLITE_CANTOPEN_BKPT, "open", zShm);
          goto shm_open_err;
        }
        pShmNode->isReadonly = 1;
      }

      /* If this process is running as root, make sure that the SHM file
      ** is owned by the same user that owns the original database.  Otherwise,
      ** the original owner will not be able to connect.
      */
      robustFchown(pShmNode->h, sStat.st_uid, sStat.st_gid);

      rc = unixLockSharedMemory(pDbFd, pShmNode);
      if( rc!=SQLITE_OK && rc!=SQLITE_READONLY_CANTINIT ) goto shm_open_err;
    }
  }

  /* Make the new connection a child of the unixShmNode */
  p->pShmNode = pShmNode;
#ifdef SQLITE_DEBUG
  p->id = pShmNode->nextShmId++;
#endif
  pShmNode->nRef++;
  pDbFd->pShm = p;
  unixLeaveMutex();

  /* The reference count on pShmNode has already been incremented under
  ** the cover of the unixEnterMutex() mutex and the pointer from the
  ** new (struct unixShm) object to the pShmNode has been set. All that is
  ** left to do is to link the new object into the linked list starting
  ** at pShmNode->pFirst. This must be done while holding the pShmNode->mutex 
  ** mutex.
  */
  sqlite3_mutex_enter(pShmNode->mutex);
  p->pNext = pShmNode->pFirst;
  pShmNode->pFirst = p;
  sqlite3_mutex_leave(pShmNode->mutex);
  return rc;

  /* Jump here on any error */
shm_open_err:
  unixShmPurge(pDbFd);       /* This call frees pShmNode if required */
  sqlite3_free(p);
  unixLeaveMutex();
  return rc;
}

/*
** This function is called to obtain a pointer to region iRegion of the 
** shared-memory associated with the database file fd. Shared-memory regions 
** are numbered starting from zero. Each shared-memory region is szRegion 
** bytes in size.
**
** If an error occurs, an error code is returned and *pp is set to NULL.
**
** Otherwise, if the bExtend parameter is 0 and the requested shared-memory
** region has not been allocated (by any client, including one running in a
** separate process), then *pp is set to NULL and SQLITE_OK returned. If 
** bExtend is non-zero and the requested shared-memory region has not yet 
** been allocated, it is allocated by this function.
**
** If the shared-memory region has already been allocated or is allocated by
** this call as described above, then it is mapped into this processes 
** address space (if it is not already), *pp is set to point to the mapped 
** memory and SQLITE_OK returned.
*/
static int unixShmMap(
  sqlite3_file *fd,               /* Handle open on database file */
  int iRegion,                    /* Region to retrieve */
  int szRegion,                   /* Size of regions */
  int bExtend,                    /* True to extend file if necessary */
  void volatile **pp              /* OUT: Mapped memory */
){
  unixFile *pDbFd = (unixFile*)fd;
  unixShm *p;
  unixShmNode *pShmNode;
  int rc = SQLITE_OK;
  int nShmPerMap = unixShmRegionPerMap();
  int nReqRegion;

  /* If the shared-memory file has not yet been opened, open it now. */
  if( pDbFd->pShm==0 ){
    rc = unixOpenSharedMemory(pDbFd);
    if( rc!=SQLITE_OK ) return rc;
  }

  p = pDbFd->pShm;
  pShmNode = p->pShmNode;
  sqlite3_mutex_enter(pShmNode->mutex);
  if( pShmNode->isUnlocked ){
    rc = unixLockSharedMemory(pDbFd, pShmNode);
    if( rc!=SQLITE_OK ) goto shmpage_out;
    pShmNode->isUnlocked = 0;
  }
  assert( szRegion==pShmNode->szRegion || pShmNode->nRegion==0 );
  assert( pShmNode->pInode==pDbFd->pInode );
  assert( pShmNode->h>=0 || pDbFd->pInode->bProcessLock==1 );
  assert( pShmNode->h<0 || pDbFd->pInode->bProcessLock==0 );

  /* Minimum number of regions required to be mapped. */
  nReqRegion = ((iRegion+nShmPerMap) / nShmPerMap) * nShmPerMap;

  if( pShmNode->nRegion<nReqRegion ){
    char **apNew;                      /* New apRegion[] array */
    int nByte = nReqRegion*szRegion;   /* Minimum required file size */
    struct stat sStat;                 /* Used by fstat() */

    pShmNode->szRegion = szRegion;

    if( pShmNode->h>=0 ){
      /* The requested region is not mapped into this processes address space.
      ** Check to see if it has been allocated (i.e. if the wal-index file is
      ** large enough to contain the requested region).
      */
      if( osFstat(pShmNode->h, &sStat) ){
        rc = SQLITE_IOERR_SHMSIZE;
        goto shmpage_out;
      }
  
      if( sStat.st_size<nByte ){
        /* The requested memory region does not exist. If bExtend is set to
        ** false, exit early. *pp will be set to NULL and SQLITE_OK returned.
        */
        if( !bExtend ){
          goto shmpage_out;
        }

        /* Alternatively, if bExtend is true, extend the file. Do this by
        ** writing a single byte to the end of each (OS) page being
        ** allocated or extended. Technically, we need only write to the
        ** last page in order to extend the file. But writing to all new
        ** pages forces the OS to allocate them immediately, which reduces
        ** the chances of SIGBUS while accessing the mapped region later on.
        */
        else{
          static const int pgsz = 4096;
          int iPg;

          /* Write to the last byte of each newly allocated or extended page */
          assert( (nByte % pgsz)==0 );
          for(iPg=(sStat.st_size/pgsz); iPg<(nByte/pgsz); iPg++){
            int x = 0;
            if( seekAndWriteFd(pShmNode->h, iPg*pgsz + pgsz-1, "", 1, &x)!=1 ){
              const char *zFile = pShmNode->zFilename;
              rc = unixLogError(SQLITE_IOERR_SHMSIZE, "write", zFile);
              goto shmpage_out;
            }
          }
        }
      }
    }

    /* Map the requested memory region into this processes address space. */
    apNew = (char **)sqlite3_realloc(
        pShmNode->apRegion, nReqRegion*sizeof(char *)
    );
    if( !apNew ){
      rc = SQLITE_IOERR_NOMEM_BKPT;
      goto shmpage_out;
    }
    pShmNode->apRegion = apNew;
    while( pShmNode->nRegion<nReqRegion ){
      int nMap = szRegion*nShmPerMap;
      int i;
      void *pMem;
      if( pShmNode->h>=0 ){
        pMem = osMmap(0, nMap,
            pShmNode->isReadonly ? PROT_READ : PROT_READ|PROT_WRITE, 
            MAP_SHARED, pShmNode->h, szRegion*(i64)pShmNode->nRegion
        );
        if( pMem==MAP_FAILED ){
          rc = unixLogError(SQLITE_IOERR_SHMMAP, "mmap", pShmNode->zFilename);
          goto shmpage_out;
        }
      }else{
        pMem = sqlite3_malloc64(szRegion);
        if( pMem==0 ){
          rc = SQLITE_NOMEM_BKPT;
          goto shmpage_out;
        }
        memset(pMem, 0, szRegion);
      }

      for(i=0; i<nShmPerMap; i++){
        pShmNode->apRegion[pShmNode->nRegion+i] = &((char*)pMem)[szRegion*i];
      }
      pShmNode->nRegion += nShmPerMap;
    }
  }

shmpage_out:
  if( pShmNode->nRegion>iRegion ){
    *pp = pShmNode->apRegion[iRegion];
  }else{
    *pp = 0;
  }
  if( pShmNode->isReadonly && rc==SQLITE_OK ) rc = SQLITE_READONLY;
  sqlite3_mutex_leave(pShmNode->mutex);
  return rc;
}

/*
** Change the lock state for a shared-memory segment.
**
** Note that the relationship between SHAREd and EXCLUSIVE locks is a little
** different here than in posix.  In xShmLock(), one can go from unlocked
** to shared and back or from unlocked to exclusive and back.  But one may
** not go from shared to exclusive or from exclusive to shared.
*/
static int unixShmLock(
  sqlite3_file *fd,          /* Database file holding the shared memory */
  int ofst,                  /* First lock to acquire or release */
  int n,                     /* Number of locks to acquire or release */
  int flags                  /* What to do with the lock */
){
  unixFile *pDbFd = (unixFile*)fd;      /* Connection holding shared memory */
  unixShm *p = pDbFd->pShm;             /* The shared memory being locked */
  unixShm *pX;                          /* For looping over all siblings */
  unixShmNode *pShmNode = p->pShmNode;  /* The underlying file iNode */
  int rc = SQLITE_OK;                   /* Result code */
  u16 mask;                             /* Mask of locks to take or release */

  assert( pShmNode==pDbFd->pInode->pShmNode );
  assert( pShmNode->pInode==pDbFd->pInode );
  assert( ofst>=0 && ofst+n<=SQLITE_SHM_NLOCK );
  assert( n>=1 );
  assert( flags==(SQLITE_SHM_LOCK | SQLITE_SHM_SHARED)
       || flags==(SQLITE_SHM_LOCK | SQLITE_SHM_EXCLUSIVE)
       || flags==(SQLITE_SHM_UNLOCK | SQLITE_SHM_SHARED)
       || flags==(SQLITE_SHM_UNLOCK | SQLITE_SHM_EXCLUSIVE) );
  assert( n==1 || (flags & SQLITE_SHM_EXCLUSIVE)!=0 );
  assert( pShmNode->h>=0 || pDbFd->pInode->bProcessLock==1 );
  assert( pShmNode->h<0 || pDbFd->pInode->bProcessLock==0 );

  mask = (1<<(ofst+n)) - (1<<ofst);
  assert( n>1 || mask==(1<<ofst) );
  sqlite3_mutex_enter(pShmNode->mutex);
  if( flags & SQLITE_SHM_UNLOCK ){
    u16 allMask = 0; /* Mask of locks held by siblings */

    /* See if any siblings hold this same lock */
    for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
      if( pX==p ) continue;
      assert( (pX->exclMask & (p->exclMask|p->sharedMask))==0 );
      allMask |= pX->sharedMask;
    }

    /* Unlock the system-level locks */
    if( (mask & allMask)==0 ){
      rc = unixShmSystemLock(pDbFd, F_UNLCK, ofst+UNIX_SHM_BASE, n);
    }else{
      rc = SQLITE_OK;
    }

    /* Undo the local locks */
    if( rc==SQLITE_OK ){
      p->exclMask &= ~mask;
      p->sharedMask &= ~mask;
    } 
  }else if( flags & SQLITE_SHM_SHARED ){
    u16 allShared = 0;  /* Union of locks held by connections other than "p" */

    /* Find out which shared locks are already held by sibling connections.
    ** If any sibling already holds an exclusive lock, go ahead and return
    ** SQLITE_BUSY.
    */
    for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
      if( (pX->exclMask & mask)!=0 ){
        rc = SQLITE_BUSY;
        break;
      }
      allShared |= pX->sharedMask;
    }

    /* Get shared locks at the system level, if necessary */
    if( rc==SQLITE_OK ){
      if( (allShared & mask)==0 ){
        rc = unixShmSystemLock(pDbFd, F_RDLCK, ofst+UNIX_SHM_BASE, n);
      }else{
        rc = SQLITE_OK;
      }
    }

    /* Get the local shared locks */
    if( rc==SQLITE_OK ){
      p->sharedMask |= mask;
    }
  }else{
    /* Make sure no sibling connections hold locks that will block this
    ** lock.  If any do, return SQLITE_BUSY right away.
    */
    for(pX=pShmNode->pFirst; pX; pX=pX->pNext){
      if( (pX->exclMask & mask)!=0 || (pX->sharedMask & mask)!=0 ){
        rc = SQLITE_BUSY;
        break;
      }
    }
  
    /* Get the exclusive locks at the system level.  Then if successful
    ** also mark the local connection as being locked.
    */
    if( rc==SQLITE_OK ){
      rc = unixShmSystemLock(pDbFd, F_WRLCK, ofst+UNIX_SHM_BASE, n);
      if( rc==SQLITE_OK ){
        assert( (p->sharedMask & mask)==0 );
        p->exclMask |= mask;
      }
    }
  }
  sqlite3_mutex_leave(pShmNode->mutex);
  OSTRACE(("SHM-LOCK shmid-%d, pid-%d got %03x,%03x\n",
           p->id, osGetpid(0), p->sharedMask, p->exclMask));
  return rc;
}

/*
** Implement a memory barrier or memory fence on shared memory.  
**
** All loads and stores begun before the barrier must complete before
** any load or store begun after the barrier.
*/
static void unixShmBarrier(
  sqlite3_file *fd                /* Database file holding the shared memory */
){
  UNUSED_PARAMETER(fd);
  sqlite3MemoryBarrier();         /* compiler-defined memory barrier */
  unixEnterMutex();               /* Also mutex, for redundancy */
  unixLeaveMutex();
}

/*
** Close a connection to shared-memory.  Delete the underlying 
** storage if deleteFlag is true.
**
** If there is no shared memory associated with the connection then this
** routine is a harmless no-op.
*/
static int unixShmUnmap(
  sqlite3_file *fd,               /* The underlying database file */
  int deleteFlag                  /* Delete shared-memory if true */
){
  unixShm *p;                     /* The connection to be closed */
  unixShmNode *pShmNode;          /* The underlying shared-memory file */
  unixShm **pp;                   /* For looping over sibling connections */
  unixFile *pDbFd;                /* The underlying database file */

  pDbFd = (unixFile*)fd;
  p = pDbFd->pShm;
  if( p==0 ) return SQLITE_OK;
  pShmNode = p->pShmNode;

  assert( pShmNode==pDbFd->pInode->pShmNode );
  assert( pShmNode->pInode==pDbFd->pInode );

  /* Remove connection p from the set of connections associated
  ** with pShmNode */
  sqlite3_mutex_enter(pShmNode->mutex);
  for(pp=&pShmNode->pFirst; (*pp)!=p; pp = &(*pp)->pNext){}
  *pp = p->pNext;

  /* Free the connection p */
  sqlite3_free(p);
  pDbFd->pShm = 0;
  sqlite3_mutex_leave(pShmNode->mutex);

  /* If pShmNode->nRef has reached 0, then close the underlying
  ** shared-memory file, too */
  unixEnterMutex();
  assert( pShmNode->nRef>0 );
  pShmNode->nRef--;
  if( pShmNode->nRef==0 ){
    if( deleteFlag && pShmNode->h>=0 ){
      osUnlink(pShmNode->zFilename);
    }
    unixShmPurge(pDbFd);
  }
  unixLeaveMutex();

  return SQLITE_OK;
}


#else
# define unixShmMap     0
# define unixShmLock    0
# define unixShmBarrier 0
# define unixShmUnmap   0
#endif /* #ifndef SQLITE_OMIT_WAL */

#if SQLITE_MAX_MMAP_SIZE>0
/*
** If it is currently memory mapped, unmap file pFd.
*/
static void unixUnmapfile(unixFile *pFd){
  assert( pFd->nFetchOut==0 );
  if( pFd->pMapRegion ){
    osMunmap(pFd->pMapRegion, pFd->mmapSizeActual);
    pFd->pMapRegion = 0;
    pFd->mmapSize = 0;
    pFd->mmapSizeActual = 0;
  }
}

/*
** Attempt to set the size of the memory mapping maintained by file 
** descriptor pFd to nNew bytes. Any existing mapping is discarded.
**
** If successful, this function sets the following variables:
**
**       unixFile.pMapRegion
**       unixFile.mmapSize
**       unixFile.mmapSizeActual
**
** If unsuccessful, an error message is logged via sqlite3_log() and
** the three variables above are zeroed. In this case SQLite should
** continue accessing the database using the xRead() and xWrite()
** methods.
*/
static void unixRemapfile(
  unixFile *pFd,                  /* File descriptor object */
  i64 nNew                        /* Required mapping size */
){
  const char *zErr = "mmap";
  int h = pFd->h;                      /* File descriptor open on db file */
  u8 *pOrig = (u8 *)pFd->pMapRegion;   /* Pointer to current file mapping */
  i64 nOrig = pFd->mmapSizeActual;     /* Size of pOrig region in bytes */
  u8 *pNew = 0;                        /* Location of new mapping */
  int flags = PROT_READ;               /* Flags to pass to mmap() */

  assert( pFd->nFetchOut==0 );
  assert( nNew>pFd->mmapSize );
  assert( nNew<=pFd->mmapSizeMax );
  assert( nNew>0 );
  assert( pFd->mmapSizeActual>=pFd->mmapSize );
  assert( MAP_FAILED!=0 );

#ifdef SQLITE_MMAP_READWRITE
  if( (pFd->ctrlFlags & UNIXFILE_RDONLY)==0 ) flags |= PROT_WRITE;
#endif

  if( pOrig ){
#if HAVE_MREMAP
    i64 nReuse = pFd->mmapSize;
#else
    const int szSyspage = osGetpagesize();
    i64 nReuse = (pFd->mmapSize & ~(szSyspage-1));
#endif
    u8 *pReq = &pOrig[nReuse];

    /* Unmap any pages of the existing mapping that cannot be reused. */
    if( nReuse!=nOrig ){
      osMunmap(pReq, nOrig-nReuse);
    }

#if HAVE_MREMAP
    pNew = osMremap(pOrig, nReuse, nNew, MREMAP_MAYMOVE);
    zErr = "mremap";
#else
    pNew = osMmap(pReq, nNew-nReuse, flags, MAP_SHARED, h, nReuse);
    if( pNew!=MAP_FAILED ){
      if( pNew!=pReq ){
        osMunmap(pNew, nNew - nReuse);
        pNew = 0;
      }else{
        pNew = pOrig;
      }
    }
#endif

    /* The attempt to extend the existing mapping failed. Free it. */
    if( pNew==MAP_FAILED || pNew==0 ){
      osMunmap(pOrig, nReuse);
    }
  }

  /* If pNew is still NULL, try to create an entirely new mapping. */
  if( pNew==0 ){
    pNew = osMmap(0, nNew, flags, MAP_SHARED, h, 0);
  }

  if( pNew==MAP_FAILED ){
    pNew = 0;
    nNew = 0;
    unixLogError(SQLITE_OK, zErr, pFd->zPath);

    /* If the mmap() above failed, assume that all subsequent mmap() calls
    ** will probably fail too. Fall back to using xRead/xWrite exclusively
    ** in this case.  */
    pFd->mmapSizeMax = 0;
  }
  pFd->pMapRegion = (void *)pNew;
  pFd->mmapSize = pFd->mmapSizeActual = nNew;
}

/*
** Memory map or remap the file opened by file-descriptor pFd (if the file
** is already mapped, the existing mapping is replaced by the new). Or, if 
** there already exists a mapping for this file, and there are still 
** outstanding xFetch() references to it, this function is a no-op.
**
** If parameter nByte is non-negative, then it is the requested size of 
** the mapping to create. Otherwise, if nByte is less than zero, then the 
** requested size is the size of the file on disk. The actual size of the
** created mapping is either the requested size or the value configured 
** using SQLITE_FCNTL_MMAP_LIMIT, whichever is smaller.
**
** SQLITE_OK is returned if no error occurs (even if the mapping is not
** recreated as a result of outstanding references) or an SQLite error
** code otherwise.
*/
static int unixMapfile(unixFile *pFd, i64 nMap){
  assert( nMap>=0 || pFd->nFetchOut==0 );
  assert( nMap>0 || (pFd->mmapSize==0 && pFd->pMapRegion==0) );
  if( pFd->nFetchOut>0 ) return SQLITE_OK;

  if( nMap<0 ){
    struct stat statbuf;          /* Low-level file information */
    if( osFstat(pFd->h, &statbuf) ){
      return SQLITE_IOERR_FSTAT;
    }
    nMap = statbuf.st_size;
  }
  if( nMap>pFd->mmapSizeMax ){
    nMap = pFd->mmapSizeMax;
  }

  assert( nMap>0 || (pFd->mmapSize==0 && pFd->pMapRegion==0) );
  if( nMap!=pFd->mmapSize ){
    unixRemapfile(pFd, nMap);
  }

  return SQLITE_OK;
}
#endif /* SQLITE_MAX_MMAP_SIZE>0 */

/*
** If possible, return a pointer to a mapping of file fd starting at offset
** iOff. The mapping must be valid for at least nAmt bytes.
**
** If such a pointer can be obtained, store it in *pp and return SQLITE_OK.
** Or, if one cannot but no error occurs, set *pp to 0 and return SQLITE_OK.
** Finally, if an error does occur, return an SQLite error code. The final
** value of *pp is undefined in this case.
**
** If this function does return a pointer, the caller must eventually 
** release the reference by calling unixUnfetch().
*/
static int unixFetch(sqlite3_file *fd, i64 iOff, int nAmt, void **pp){
#if SQLITE_MAX_MMAP_SIZE>0
  unixFile *pFd = (unixFile *)fd;   /* The underlying database file */
#endif
  *pp = 0;

#if SQLITE_MAX_MMAP_SIZE>0
  if( pFd->mmapSizeMax>0 ){
    if( pFd->pMapRegion==0 ){
      int rc = unixMapfile(pFd, -1);
      if( rc!=SQLITE_OK ) return rc;
    }
    if( pFd->mmapSize >= iOff+nAmt ){
      *pp = &((u8 *)pFd->pMapRegion)[iOff];
      pFd->nFetchOut++;
    }
  }
#endif
  return SQLITE_OK;
}

/*
** If the third argument is non-NULL, then this function releases a 
** reference obtained by an earlier call to unixFetch(). The second
** argument passed to this function must be the same as the corresponding
** argument that was passed to the unixFetch() invocation. 
**
** Or, if the third argument is NULL, then this function is being called 
** to inform the VFS layer that, according to POSIX, any existing mapping 
** may now be invalid and should be unmapped.
*/
static int unixUnfetch(sqlite3_file *fd, i64 iOff, void *p){
#if SQLITE_MAX_MMAP_SIZE>0
  unixFile *pFd = (unixFile *)fd;   /* The underlying database file */
  UNUSED_PARAMETER(iOff);

  /* If p==0 (unmap the entire file) then there must be no outstanding 
  ** xFetch references. Or, if p!=0 (meaning it is an xFetch reference),
  ** then there must be at least one outstanding.  */
  assert( (p==0)==(pFd->nFetchOut==0) );

  /* If p!=0, it must match the iOff value. */
  assert( p==0 || p==&((u8 *)pFd->pMapRegion)[iOff] );

  if( p ){
    pFd->nFetchOut--;
  }else{
    unixUnmapfile(pFd);
  }

  assert( pFd->nFetchOut>=0 );
#else
  UNUSED_PARAMETER(fd);
  UNUSED_PARAMETER(p);
  UNUSED_PARAMETER(iOff);
#endif
  return SQLITE_OK;
}

/*
** Here ends the implementation of all sqlite3_file methods.
**
********************** End sqlite3_file Methods *******************************
******************************************************************************/

/*
** This division contains definitions of sqlite3_io_methods objects that
** implement various file locking strategies.  It also contains definitions
** of "finder" functions.  A finder-function is used to locate the appropriate
** sqlite3_io_methods object for a particular database file.  The pAppData
** field of the sqlite3_vfs VFS objects are initialized to be pointers to
** the correct finder-function for that VFS.
**
** Most finder functions return a pointer to a fixed sqlite3_io_methods
** object.  The only interesting finder-function is autolockIoFinder, which
** looks at the filesystem type and tries to guess the best locking
** strategy from that.
**
** For finder-function F, two objects are created:
**
**    (1) The real finder-function named "FImpt()".
**
**    (2) A constant pointer to this function named just "F".
**
**
** A pointer to the F pointer is used as the pAppData value for VFS
** objects.  We have to do this instead of letting pAppData point
** directly at the finder-function since C90 rules prevent a void*
** from be cast into a function pointer.
**
**
** Each instance of this macro generates two objects:
**
**   *  A constant sqlite3_io_methods object call METHOD that has locking
**      methods CLOSE, LOCK, UNLOCK, CKRESLOCK.
**
**   *  An I/O method finder function called FINDER that returns a pointer
**      to the METHOD object in the previous bullet.
*/
#define IOMETHODS(FINDER,METHOD,VERSION,CLOSE,LOCK,UNLOCK,CKLOCK,SHMMAP)     \
static const sqlite3_io_methods METHOD = {                                   \
   VERSION,                    /* iVersion */                                \
   CLOSE,                      /* xClose */                                  \
   unixRead,                   /* xRead */                                   \
   unixWrite,                  /* xWrite */                                  \
   unixTruncate,               /* xTruncate */                               \
   unixSync,                   /* xSync */                                   \
   unixFileSize,               /* xFileSize */                               \
   LOCK,                       /* xLock */                                   \
   UNLOCK,                     /* xUnlock */                                 \
   CKLOCK,                     /* xCheckReservedLock */                      \
   unixFileControl,            /* xFileControl */                            \
   unixSectorSize,             /* xSectorSize */                             \
   unixDeviceCharacteristics,  /* xDeviceCapabilities */                     \
   SHMMAP,                     /* xShmMap */                                 \
   unixShmLock,                /* xShmLock */                                \
   unixShmBarrier,             /* xShmBarrier */                             \
   unixShmUnmap,               /* xShmUnmap */                               \
   unixFetch,                  /* xFetch */                                  \
   unixUnfetch,                /* xUnfetch */                                \
};                                                                           \
static const sqlite3_io_methods *FINDER##Impl(const char *z, unixFile *p){   \
  UNUSED_PARAMETER(z); UNUSED_PARAMETER(p);                                  \
  return &METHOD;                                                            \
}                                                                            \
static const sqlite3_io_methods *(*const FINDER)(const char*,unixFile *p)    \
    = FINDER##Impl;

/*
** Here are all of the sqlite3_io_methods objects for each of the
** locking strategies.  Functions that return pointers to these methods
** are also created.
*/
IOMETHODS(
  posixIoFinder,            /* Finder function name */
  posixIoMethods,           /* sqlite3_io_methods object name */
  3,                        /* shared memory and mmap are enabled */
  unixClose,                /* xClose method */
  unixLock,                 /* xLock method */
  unixUnlock,               /* xUnlock method */
  unixCheckReservedLock,    /* xCheckReservedLock method */
  unixShmMap                /* xShmMap method */
)
IOMETHODS(
  nolockIoFinder,           /* Finder function name */
  nolockIoMethods,          /* sqlite3_io_methods object name */
  3,                        /* shared memory is disabled */
  nolockClose,              /* xClose method */
  nolockLock,               /* xLock method */
  nolockUnlock,             /* xUnlock method */
  nolockCheckReservedLock,  /* xCheckReservedLock method */
  0                         /* xShmMap method */
)
IOMETHODS(
  dotlockIoFinder,          /* Finder function name */
  dotlockIoMethods,         /* sqlite3_io_methods object name */
  1,                        /* shared memory is disabled */
  dotlockClose,             /* xClose method */
  dotlockLock,              /* xLock method */
  dotlockUnlock,            /* xUnlock method */
  dotlockCheckReservedLock, /* xCheckReservedLock method */
  0                         /* xShmMap method */
)

#if SQLITE_ENABLE_LOCKING_STYLE
IOMETHODS(
  flockIoFinder,            /* Finder function name */
  flockIoMethods,           /* sqlite3_io_methods object name */
  1,                        /* shared memory is disabled */
  flockClose,               /* xClose method */
  flockLock,                /* xLock method */
  flockUnlock,              /* xUnlock method */
  flockCheckReservedLock,   /* xCheckReservedLock method */
  0                         /* xShmMap method */
)
#endif

#if OS_VXWORKS
IOMETHODS(
  semIoFinder,              /* Finder function name */
  semIoMethods,             /* sqlite3_io_methods object name */
  1,                        /* shared memory is disabled */
  semXClose,                /* xClose method */
  semXLock,                 /* xLock method */
  semXUnlock,               /* xUnlock method */
  semXCheckReservedLock,    /* xCheckReservedLock method */
  0                         /* xShmMap method */
)
#endif

#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
IOMETHODS(
  afpIoFinder,              /* Finder function name */
  afpIoMethods,             /* sqlite3_io_methods object name */
  1,                        /* shared memory is disabled */
  afpClose,                 /* xClose method */
  afpLock,                  /* xLock method */
  afpUnlock,                /* xUnlock method */
  afpCheckReservedLock,     /* xCheckReservedLock method */
  0                         /* xShmMap method */
)
#endif

/*
** The proxy locking method is a "super-method" in the sense that it
** opens secondary file descriptors for the conch and lock files and
** it uses proxy, dot-file, AFP, and flock() locking methods on those
** secondary files.  For this reason, the division that implements
** proxy locking is located much further down in the file.  But we need
** to go ahead and define the sqlite3_io_methods and finder function
** for proxy locking here.  So we forward declare the I/O methods.
*/
#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
static int proxyClose(sqlite3_file*);
static int proxyLock(sqlite3_file*, int);
static int proxyUnlock(sqlite3_file*, int);
static int proxyCheckReservedLock(sqlite3_file*, int*);
IOMETHODS(
  proxyIoFinder,            /* Finder function name */
  proxyIoMethods,           /* sqlite3_io_methods object name */
  1,                        /* shared memory is disabled */
  proxyClose,               /* xClose method */
  proxyLock,                /* xLock method */
  proxyUnlock,              /* xUnlock method */
  proxyCheckReservedLock,   /* xCheckReservedLock method */
  0                         /* xShmMap method */
)
#endif

/* nfs lockd on OSX 10.3+ doesn't clear write locks when a read lock is set */
#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
IOMETHODS(
  nfsIoFinder,               /* Finder function name */
  nfsIoMethods,              /* sqlite3_io_methods object name */
  1,                         /* shared memory is disabled */
  unixClose,                 /* xClose method */
  unixLock,                  /* xLock method */
  nfsUnlock,                 /* xUnlock method */
  unixCheckReservedLock,     /* xCheckReservedLock method */
  0                          /* xShmMap method */
)
#endif

#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
/* 
** This "finder" function attempts to determine the best locking strategy 
** for the database file "filePath".  It then returns the sqlite3_io_methods
** object that implements that strategy.
**
** This is for MacOSX only.
*/
static const sqlite3_io_methods *autolockIoFinderImpl(
  const char *filePath,    /* name of the database file */
  unixFile *pNew           /* open file object for the database file */
){
  static const struct Mapping {
    const char *zFilesystem;              /* Filesystem type name */
    const sqlite3_io_methods *pMethods;   /* Appropriate locking method */
  } aMap[] = {
    { "hfs",    &posixIoMethods },
    { "ufs",    &posixIoMethods },
    { "afpfs",  &afpIoMethods },
    { "smbfs",  &afpIoMethods },
    { "webdav", &nolockIoMethods },
    { 0, 0 }
  };
  int i;
  struct statfs fsInfo;
  struct flock lockInfo;

  if( !filePath ){
    /* If filePath==NULL that means we are dealing with a transient file
    ** that does not need to be locked. */
    return &nolockIoMethods;
  }
  if( statfs(filePath, &fsInfo) != -1 ){
    if( fsInfo.f_flags & MNT_RDONLY ){
      return &nolockIoMethods;
    }
    for(i=0; aMap[i].zFilesystem; i++){
      if( strcmp(fsInfo.f_fstypename, aMap[i].zFilesystem)==0 ){
        return aMap[i].pMethods;
      }
    }
  }

  /* Default case. Handles, amongst others, "nfs".
  ** Test byte-range lock using fcntl(). If the call succeeds, 
  ** assume that the file-system supports POSIX style locks. 
  */
  lockInfo.l_len = 1;
  lockInfo.l_start = 0;
  lockInfo.l_whence = SEEK_SET;
  lockInfo.l_type = F_RDLCK;
  if( osFcntl(pNew->h, F_GETLK, &lockInfo)!=-1 ) {
    if( strcmp(fsInfo.f_fstypename, "nfs")==0 ){
      return &nfsIoMethods;
    } else {
      return &posixIoMethods;
    }
  }else{
    return &dotlockIoMethods;
  }
}
static const sqlite3_io_methods 
  *(*const autolockIoFinder)(const char*,unixFile*) = autolockIoFinderImpl;

#endif /* defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE */

#if OS_VXWORKS
/*
** This "finder" function for VxWorks checks to see if posix advisory
** locking works.  If it does, then that is what is used.  If it does not
** work, then fallback to named semaphore locking.
*/
static const sqlite3_io_methods *vxworksIoFinderImpl(
  const char *filePath,    /* name of the database file */
  unixFile *pNew           /* the open file object */
){
  struct flock lockInfo;

  if( !filePath ){
    /* If filePath==NULL that means we are dealing with a transient file
    ** that does not need to be locked. */
    return &nolockIoMethods;
  }

  /* Test if fcntl() is supported and use POSIX style locks.
  ** Otherwise fall back to the named semaphore method.
  */
  lockInfo.l_len = 1;
  lockInfo.l_start = 0;
  lockInfo.l_whence = SEEK_SET;
  lockInfo.l_type = F_RDLCK;
  if( osFcntl(pNew->h, F_GETLK, &lockInfo)!=-1 ) {
    return &posixIoMethods;
  }else{
    return &semIoMethods;
  }
}
static const sqlite3_io_methods 
  *(*const vxworksIoFinder)(const char*,unixFile*) = vxworksIoFinderImpl;

#endif /* OS_VXWORKS */

/*
** An abstract type for a pointer to an IO method finder function:
*/
typedef const sqlite3_io_methods *(*finder_type)(const char*,unixFile*);


/****************************************************************************
**************************** sqlite3_vfs methods ****************************
**
** This division contains the implementation of methods on the
** sqlite3_vfs object.
*/

/*
** Initialize the contents of the unixFile structure pointed to by pId.
*/
static int fillInUnixFile(
  sqlite3_vfs *pVfs,      /* Pointer to vfs object */
  int h,                  /* Open file descriptor of file being opened */
  sqlite3_file *pId,      /* Write to the unixFile structure here */
  const char *zFilename,  /* Name of the file being opened */
  int ctrlFlags           /* Zero or more UNIXFILE_* values */
){
  const sqlite3_io_methods *pLockingStyle;
  unixFile *pNew = (unixFile *)pId;
  int rc = SQLITE_OK;

  assert( pNew->pInode==NULL );

  /* No locking occurs in temporary files */
  assert( zFilename!=0 || (ctrlFlags & UNIXFILE_NOLOCK)!=0 );

  OSTRACE(("OPEN    %-3d %s\n", h, zFilename));
  pNew->h = h;
  pNew->pVfs = pVfs;
  pNew->zPath = zFilename;
  pNew->ctrlFlags = (u8)ctrlFlags;
#if SQLITE_MAX_MMAP_SIZE>0
  pNew->mmapSizeMax = sqlite3GlobalConfig.szMmap;
#endif
  if( sqlite3_uri_boolean(((ctrlFlags & UNIXFILE_URI) ? zFilename : 0),
                           "psow", SQLITE_POWERSAFE_OVERWRITE) ){
    pNew->ctrlFlags |= UNIXFILE_PSOW;
  }
  if( strcmp(pVfs->zName,"unix-excl")==0 ){
    pNew->ctrlFlags |= UNIXFILE_EXCL;
  }

#if OS_VXWORKS
  pNew->pId = vxworksFindFileId(zFilename);
  if( pNew->pId==0 ){
    ctrlFlags |= UNIXFILE_NOLOCK;
    rc = SQLITE_NOMEM_BKPT;
  }
#endif

  if( ctrlFlags & UNIXFILE_NOLOCK ){
    pLockingStyle = &nolockIoMethods;
  }else{
    pLockingStyle = (**(finder_type*)pVfs->pAppData)(zFilename, pNew);
#if SQLITE_ENABLE_LOCKING_STYLE
    /* Cache zFilename in the locking context (AFP and dotlock override) for
    ** proxyLock activation is possible (remote proxy is based on db name)
    ** zFilename remains valid until file is closed, to support */
    pNew->lockingContext = (void*)zFilename;
#endif
  }

  if( pLockingStyle == &posixIoMethods
#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE
    || pLockingStyle == &nfsIoMethods
#endif
  ){
    unixEnterMutex();
    rc = findInodeInfo(pNew, &pNew->pInode);
    if( rc!=SQLITE_OK ){
      /* If an error occurred in findInodeInfo(), close the file descriptor
      ** immediately, before releasing the mutex. findInodeInfo() may fail
      ** in two scenarios:
      **
      **   (a) A call to fstat() failed.
      **   (b) A malloc failed.
      **
      ** Scenario (b) may only occur if the process is holding no other
      ** file descriptors open on the same file. If there were other file
      ** descriptors on this file, then no malloc would be required by
      ** findInodeInfo(). If this is the case, it is quite safe to close
      ** handle h - as it is guaranteed that no posix locks will be released
      ** by doing so.
      **
      ** If scenario (a) caused the error then things are not so safe. The
      ** implicit assumption here is that if fstat() fails, things are in
      ** such bad shape that dropping a lock or two doesn't matter much.
      */
      robust_close(pNew, h, __LINE__);
      h = -1;
    }
    unixLeaveMutex();
  }

#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
  else if( pLockingStyle == &afpIoMethods ){
    /* AFP locking uses the file path so it needs to be included in
    ** the afpLockingContext.
    */
    afpLockingContext *pCtx;
    pNew->lockingContext = pCtx = sqlite3_malloc64( sizeof(*pCtx) );
    if( pCtx==0 ){
      rc = SQLITE_NOMEM_BKPT;
    }else{
      /* NB: zFilename exists and remains valid until the file is closed
      ** according to requirement F11141.  So we do not need to make a
      ** copy of the filename. */
      pCtx->dbPath = zFilename;
      pCtx->reserved = 0;
      srandomdev();
      unixEnterMutex();
      rc = findInodeInfo(pNew, &pNew->pInode);
      if( rc!=SQLITE_OK ){
        sqlite3_free(pNew->lockingContext);
        robust_close(pNew, h, __LINE__);
        h = -1;
      }
      unixLeaveMutex();        
    }
  }
#endif

  else if( pLockingStyle == &dotlockIoMethods ){
    /* Dotfile locking uses the file path so it needs to be included in
    ** the dotlockLockingContext 
    */
    char *zLockFile;
    int nFilename;
    assert( zFilename!=0 );
    nFilename = (int)strlen(zFilename) + 6;
    zLockFile = (char *)sqlite3_malloc64(nFilename);
    if( zLockFile==0 ){
      rc = SQLITE_NOMEM_BKPT;
    }else{
      sqlite3_snprintf(nFilename, zLockFile, "%s" DOTLOCK_SUFFIX, zFilename);
    }
    pNew->lockingContext = zLockFile;
  }

#if OS_VXWORKS
  else if( pLockingStyle == &semIoMethods ){
    /* Named semaphore locking uses the file path so it needs to be
    ** included in the semLockingContext
    */
    unixEnterMutex();
    rc = findInodeInfo(pNew, &pNew->pInode);
    if( (rc==SQLITE_OK) && (pNew->pInode->pSem==NULL) ){
      char *zSemName = pNew->pInode->aSemName;
      int n;
      sqlite3_snprintf(MAX_PATHNAME, zSemName, "/%s.sem",
                       pNew->pId->zCanonicalName);
      for( n=1; zSemName[n]; n++ )
        if( zSemName[n]=='/' ) zSemName[n] = '_';
      pNew->pInode->pSem = sem_open(zSemName, O_CREAT, 0666, 1);
      if( pNew->pInode->pSem == SEM_FAILED ){
        rc = SQLITE_NOMEM_BKPT;
        pNew->pInode->aSemName[0] = '\0';
      }
    }
    unixLeaveMutex();
  }
#endif
  
  storeLastErrno(pNew, 0);
#if OS_VXWORKS
  if( rc!=SQLITE_OK ){
    if( h>=0 ) robust_close(pNew, h, __LINE__);
    h = -1;
    osUnlink(zFilename);
    pNew->ctrlFlags |= UNIXFILE_DELETE;
  }
#endif
  if( rc!=SQLITE_OK ){
    if( h>=0 ) robust_close(pNew, h, __LINE__);
  }else{
    pNew->pMethod = pLockingStyle;
    OpenCounter(+1);
    verifyDbFile(pNew);
  }
  return rc;
}

/*
** Return the name of a directory in which to put temporary files.
** If no suitable temporary file directory can be found, return NULL.
*/
static const char *unixTempFileDir(void){
  static const char *azDirs[] = {
     0,
     0,
     "/var/tmp",
     "/usr/tmp",
     "/tmp",
     "."
  };
  unsigned int i = 0;
  struct stat buf;
  const char *zDir = sqlite3_temp_directory;

  if( !azDirs[0] ) azDirs[0] = getenv("SQLITE_TMPDIR");
  if( !azDirs[1] ) azDirs[1] = getenv("TMPDIR");
  while(1){
    if( zDir!=0
     && osStat(zDir, &buf)==0
     && S_ISDIR(buf.st_mode)
     && osAccess(zDir, 03)==0
    ){
      return zDir;
    }
    if( i>=sizeof(azDirs)/sizeof(azDirs[0]) ) break;
    zDir = azDirs[i++];
  }
  return 0;
}

/*
** Create a temporary file name in zBuf.  zBuf must be allocated
** by the calling process and must be big enough to hold at least
** pVfs->mxPathname bytes.
*/
static int unixGetTempname(int nBuf, char *zBuf){
  const char *zDir;
  int iLimit = 0;

  /* It's odd to simulate an io-error here, but really this is just
  ** using the io-error infrastructure to test that SQLite handles this
  ** function failing. 
  */
  zBuf[0] = 0;
  SimulateIOError( return SQLITE_IOERR );

  zDir = unixTempFileDir();
  if( zDir==0 ) return SQLITE_IOERR_GETTEMPPATH;
  do{
    u64 r;
    sqlite3_randomness(sizeof(r), &r);
    assert( nBuf>2 );
    zBuf[nBuf-2] = 0;
    sqlite3_snprintf(nBuf, zBuf, "%s/"SQLITE_TEMP_FILE_PREFIX"%llx%c",
                     zDir, r, 0);
    if( zBuf[nBuf-2]!=0 || (iLimit++)>10 ) return SQLITE_ERROR;
  }while( osAccess(zBuf,0)==0 );
  return SQLITE_OK;
}

#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
/*
** Routine to transform a unixFile into a proxy-locking unixFile.
** Implementation in the proxy-lock division, but used by unixOpen()
** if SQLITE_PREFER_PROXY_LOCKING is defined.
*/
static int proxyTransformUnixFile(unixFile*, const char*);
#endif

/*
** Search for an unused file descriptor that was opened on the database 
** file (not a journal or master-journal file) identified by pathname
** zPath with SQLITE_OPEN_XXX flags matching those passed as the second
** argument to this function.
**
** Such a file descriptor may exist if a database connection was closed
** but the associated file descriptor could not be closed because some
** other file descriptor open on the same file is holding a file-lock.
** Refer to comments in the unixClose() function and the lengthy comment
** describing "Posix Advisory Locking" at the start of this file for 
** further details. Also, ticket #4018.
**
** If a suitable file descriptor is found, then it is returned. If no
** such file descriptor is located, -1 is returned.
*/
static UnixUnusedFd *findReusableFd(const char *zPath, int flags){
  UnixUnusedFd *pUnused = 0;

  /* Do not search for an unused file descriptor on vxworks. Not because
  ** vxworks would not benefit from the change (it might, we're not sure),
  ** but because no way to test it is currently available. It is better 
  ** not to risk breaking vxworks support for the sake of such an obscure 
  ** feature.  */
#if !OS_VXWORKS
  struct stat sStat;                   /* Results of stat() call */

  unixEnterMutex();

  /* A stat() call may fail for various reasons. If this happens, it is
  ** almost certain that an open() call on the same path will also fail.
  ** For this reason, if an error occurs in the stat() call here, it is
  ** ignored and -1 is returned. The caller will try to open a new file
  ** descriptor on the same path, fail, and return an error to SQLite.
  **
  ** Even if a subsequent open() call does succeed, the consequences of
  ** not searching for a reusable file descriptor are not dire.  */
  if( nUnusedFd>0 && 0==osStat(zPath, &sStat) ){
    unixInodeInfo *pInode;

    pInode = inodeList;
    while( pInode && (pInode->fileId.dev!=sStat.st_dev
                     || pInode->fileId.ino!=(u64)sStat.st_ino) ){
       pInode = pInode->pNext;
    }
    if( pInode ){
      UnixUnusedFd **pp;
      for(pp=&pInode->pUnused; *pp && (*pp)->flags!=flags; pp=&((*pp)->pNext));
      pUnused = *pp;
      if( pUnused ){
        nUnusedFd--;
        *pp = pUnused->pNext;
      }
    }
  }
  unixLeaveMutex();
#endif    /* if !OS_VXWORKS */
  return pUnused;
}

/*
** Find the mode, uid and gid of file zFile. 
*/
static int getFileMode(
  const char *zFile,              /* File name */
  mode_t *pMode,                  /* OUT: Permissions of zFile */
  uid_t *pUid,                    /* OUT: uid of zFile. */
  gid_t *pGid                     /* OUT: gid of zFile. */
){
  struct stat sStat;              /* Output of stat() on database file */
  int rc = SQLITE_OK;
  if( 0==osStat(zFile, &sStat) ){
    *pMode = sStat.st_mode & 0777;
    *pUid = sStat.st_uid;
    *pGid = sStat.st_gid;
  }else{
    rc = SQLITE_IOERR_FSTAT;
  }
  return rc;
}

/*
** This function is called by unixOpen() to determine the unix permissions
** to create new files with. If no error occurs, then SQLITE_OK is returned
** and a value suitable for passing as the third argument to open(2) is
** written to *pMode. If an IO error occurs, an SQLite error code is 
** returned and the value of *pMode is not modified.
**
** In most cases, this routine sets *pMode to 0, which will become
** an indication to robust_open() to create the file using
** SQLITE_DEFAULT_FILE_PERMISSIONS adjusted by the umask.
** But if the file being opened is a WAL or regular journal file, then 
** this function queries the file-system for the permissions on the 
** corresponding database file and sets *pMode to this value. Whenever 
** possible, WAL and journal files are created using the same permissions 
** as the associated database file.
**
** If the SQLITE_ENABLE_8_3_NAMES option is enabled, then the
** original filename is unavailable.  But 8_3_NAMES is only used for
** FAT filesystems and permissions do not matter there, so just use
** the default permissions.
*/
static int findCreateFileMode(
  const char *zPath,              /* Path of file (possibly) being created */
  int flags,                      /* Flags passed as 4th argument to xOpen() */
  mode_t *pMode,                  /* OUT: Permissions to open file with */
  uid_t *pUid,                    /* OUT: uid to set on the file */
  gid_t *pGid                     /* OUT: gid to set on the file */
){
  int rc = SQLITE_OK;             /* Return Code */
  *pMode = 0;
  *pUid = 0;
  *pGid = 0;
  if( flags & (SQLITE_OPEN_WAL|SQLITE_OPEN_MAIN_JOURNAL) ){
    char zDb[MAX_PATHNAME+1];     /* Database file path */
    int nDb;                      /* Number of valid bytes in zDb */

    /* zPath is a path to a WAL or journal file. The following block derives
    ** the path to the associated database file from zPath. This block handles
    ** the following naming conventions:
    **
    **   "<path to db>-journal"
    **   "<path to db>-wal"
    **   "<path to db>-journalNN"
    **   "<path to db>-walNN"
    **
    ** where NN is a decimal number. The NN naming schemes are 
    ** used by the test_multiplex.c module.
    */
    nDb = sqlite3Strlen30(zPath) - 1; 
    while( zPath[nDb]!='-' ){
      /* In normal operation, the journal file name will always contain
      ** a '-' character.  However in 8+3 filename mode, or if a corrupt
      ** rollback journal specifies a master journal with a goofy name, then
      ** the '-' might be missing. */
      if( nDb==0 || zPath[nDb]=='.' ) return SQLITE_OK;
      nDb--;
    }
    memcpy(zDb, zPath, nDb);
    zDb[nDb] = '\0';

    rc = getFileMode(zDb, pMode, pUid, pGid);
  }else if( flags & SQLITE_OPEN_DELETEONCLOSE ){
    *pMode = 0600;
  }else if( flags & SQLITE_OPEN_URI ){
    /* If this is a main database file and the file was opened using a URI
    ** filename, check for the "modeof" parameter. If present, interpret
    ** its value as a filename and try to copy the mode, uid and gid from
    ** that file.  */
    const char *z = sqlite3_uri_parameter(zPath, "modeof");
    if( z ){
      rc = getFileMode(z, pMode, pUid, pGid);
    }
  }
  return rc;
}

/*
** Open the file zPath.
** 
** Previously, the SQLite OS layer used three functions in place of this
** one:
**
**     sqlite3OsOpenReadWrite();
**     sqlite3OsOpenReadOnly();
**     sqlite3OsOpenExclusive();
**
** These calls correspond to the following combinations of flags:
**
**     ReadWrite() ->     (READWRITE | CREATE)
**     ReadOnly()  ->     (READONLY) 
**     OpenExclusive() -> (READWRITE | CREATE | EXCLUSIVE)
**
** The old OpenExclusive() accepted a boolean argument - "delFlag". If
** true, the file was configured to be automatically deleted when the
** file handle closed. To achieve the same effect using this new 
** interface, add the DELETEONCLOSE flag to those specified above for 
** OpenExclusive().
*/
static int unixOpen(
  sqlite3_vfs *pVfs,           /* The VFS for which this is the xOpen method */
  const char *zPath,           /* Pathname of file to be opened */
  sqlite3_file *pFile,         /* The file descriptor to be filled in */
  int flags,                   /* Input flags to control the opening */
  int *pOutFlags               /* Output flags returned to SQLite core */
){
  unixFile *p = (unixFile *)pFile;
  int fd = -1;                   /* File descriptor returned by open() */
  int openFlags = 0;             /* Flags to pass to open() */
  int eType = flags&0xFFFFFF00;  /* Type of file to open */
  int noLock;                    /* True to omit locking primitives */
  int rc = SQLITE_OK;            /* Function Return Code */
  int ctrlFlags = 0;             /* UNIXFILE_* flags */

  int isExclusive  = (flags & SQLITE_OPEN_EXCLUSIVE);
  int isDelete     = (flags & SQLITE_OPEN_DELETEONCLOSE);
  int isCreate     = (flags & SQLITE_OPEN_CREATE);
  int isReadonly   = (flags & SQLITE_OPEN_READONLY);
  int isReadWrite  = (flags & SQLITE_OPEN_READWRITE);
#if SQLITE_ENABLE_LOCKING_STYLE
  int isAutoProxy  = (flags & SQLITE_OPEN_AUTOPROXY);
#endif
#if defined(__APPLE__) || SQLITE_ENABLE_LOCKING_STYLE
  struct statfs fsInfo;
#endif

  /* If creating a master or main-file journal, this function will open
  ** a file-descriptor on the directory too. The first time unixSync()
  ** is called the directory file descriptor will be fsync()ed and close()d.
  */
  int isNewJrnl = (isCreate && (
        eType==SQLITE_OPEN_MASTER_JOURNAL 
     || eType==SQLITE_OPEN_MAIN_JOURNAL 
     || eType==SQLITE_OPEN_WAL
  ));

  /* If argument zPath is a NULL pointer, this function is required to open
  ** a temporary file. Use this buffer to store the file name in.
  */
  char zTmpname[MAX_PATHNAME+2];
  const char *zName = zPath;

  /* Check the following statements are true: 
  **
  **   (a) Exactly one of the READWRITE and READONLY flags must be set, and 
  **   (b) if CREATE is set, then READWRITE must also be set, and
  **   (c) if EXCLUSIVE is set, then CREATE must also be set.
  **   (d) if DELETEONCLOSE is set, then CREATE must also be set.
  */
  assert((isReadonly==0 || isReadWrite==0) && (isReadWrite || isReadonly));
  assert(isCreate==0 || isReadWrite);
  assert(isExclusive==0 || isCreate);
  assert(isDelete==0 || isCreate);

  /* The main DB, main journal, WAL file and master journal are never 
  ** automatically deleted. Nor are they ever temporary files.  */
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_MAIN_DB );
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_MAIN_JOURNAL );
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_MASTER_JOURNAL );
  assert( (!isDelete && zName) || eType!=SQLITE_OPEN_WAL );

  /* Assert that the upper layer has set one of the "file-type" flags. */
  assert( eType==SQLITE_OPEN_MAIN_DB      || eType==SQLITE_OPEN_TEMP_DB 
       || eType==SQLITE_OPEN_MAIN_JOURNAL || eType==SQLITE_OPEN_TEMP_JOURNAL 
       || eType==SQLITE_OPEN_SUBJOURNAL   || eType==SQLITE_OPEN_MASTER_JOURNAL 
       || eType==SQLITE_OPEN_TRANSIENT_DB || eType==SQLITE_OPEN_WAL
  );

  /* Detect a pid change and reset the PRNG.  There is a race condition
  ** here such that two or more threads all trying to open databases at
  ** the same instant might all reset the PRNG.  But multiple resets
  ** are harmless.
  */
  if( randomnessPid!=osGetpid(0) ){
    randomnessPid = osGetpid(0);
    sqlite3_randomness(0,0);
  }
  memset(p, 0, sizeof(unixFile));

  if( eType==SQLITE_OPEN_MAIN_DB ){
    UnixUnusedFd *pUnused;
    pUnused = findReusableFd(zName, flags);
    if( pUnused ){
      fd = pUnused->fd;
    }else{
      pUnused = sqlite3_malloc64(sizeof(*pUnused));
      if( !pUnused ){
        return SQLITE_NOMEM_BKPT;
      }
    }
    p->pPreallocatedUnused = pUnused;

    /* Database filenames are double-zero terminated if they are not
    ** URIs with parameters.  Hence, they can always be passed into
    ** sqlite3_uri_parameter(). */
    assert( (flags & SQLITE_OPEN_URI) || zName[strlen(zName)+1]==0 );

  }else if( !zName ){
    /* If zName is NULL, the upper layer is requesting a temp file. */
    assert(isDelete && !isNewJrnl);
    rc = unixGetTempname(pVfs->mxPathname, zTmpname);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    zName = zTmpname;

    /* Generated temporary filenames are always double-zero terminated
    ** for use by sqlite3_uri_parameter(). */
    assert( zName[strlen(zName)+1]==0 );
  }

  /* Determine the value of the flags parameter passed to POSIX function
  ** open(). These must be calculated even if open() is not called, as
  ** they may be stored as part of the file handle and used by the 
  ** 'conch file' locking functions later on.  */
  if( isReadonly )  openFlags |= O_RDONLY;
  if( isReadWrite ) openFlags |= O_RDWR;
  if( isCreate )    openFlags |= O_CREAT;
  if( isExclusive ) openFlags |= (O_EXCL|O_NOFOLLOW);
  openFlags |= (O_LARGEFILE|O_BINARY);

  if( fd<0 ){
    mode_t openMode;              /* Permissions to create file with */
    uid_t uid;                    /* Userid for the file */
    gid_t gid;                    /* Groupid for the file */
    rc = findCreateFileMode(zName, flags, &openMode, &uid, &gid);
    if( rc!=SQLITE_OK ){
      assert( !p->pPreallocatedUnused );
      assert( eType==SQLITE_OPEN_WAL || eType==SQLITE_OPEN_MAIN_JOURNAL );
      return rc;
    }
    fd = robust_open(zName, openFlags, openMode);
    OSTRACE(("OPENX   %-3d %s 0%o\n", fd, zName, openFlags));
    assert( !isExclusive || (openFlags & O_CREAT)!=0 );
    if( fd<0 ){
      if( isNewJrnl && errno==EACCES && osAccess(zName, F_OK) ){
        /* If unable to create a journal because the directory is not
        ** writable, change the error code to indicate that. */
        rc = SQLITE_READONLY_DIRECTORY;
      }else if( errno!=EISDIR && isReadWrite ){
        /* Failed to open the file for read/write access. Try read-only. */
        flags &= ~(SQLITE_OPEN_READWRITE|SQLITE_OPEN_CREATE);
        openFlags &= ~(O_RDWR|O_CREAT);
        flags |= SQLITE_OPEN_READONLY;
        openFlags |= O_RDONLY;
        isReadonly = 1;
        fd = robust_open(zName, openFlags, openMode);
      }
    }
    if( fd<0 ){
      int rc2 = unixLogError(SQLITE_CANTOPEN_BKPT, "open", zName);
      if( rc==SQLITE_OK ) rc = rc2;
      goto open_finished;
    }

    /* If this process is running as root and if creating a new rollback
    ** journal or WAL file, set the ownership of the journal or WAL to be
    ** the same as the original database.
    */
    if( flags & (SQLITE_OPEN_WAL|SQLITE_OPEN_MAIN_JOURNAL) ){
      robustFchown(fd, uid, gid);
    }
  }
  assert( fd>=0 );
  if( pOutFlags ){
    *pOutFlags = flags;
  }

  if( p->pPreallocatedUnused ){
    p->pPreallocatedUnused->fd = fd;
    p->pPreallocatedUnused->flags = flags;
  }

  if( isDelete ){
#if OS_VXWORKS
    zPath = zName;
#elif defined(SQLITE_UNLINK_AFTER_CLOSE)
    zPath = sqlite3_mprintf("%s", zName);
    if( zPath==0 ){
      robust_close(p, fd, __LINE__);
      return SQLITE_NOMEM_BKPT;
    }
#else
    osUnlink(zName);
#endif
  }
#if SQLITE_ENABLE_LOCKING_STYLE
  else{
    p->openFlags = openFlags;
  }
#endif
  
#if defined(__APPLE__) || SQLITE_ENABLE_LOCKING_STYLE
  if( fstatfs(fd, &fsInfo) == -1 ){
    storeLastErrno(p, errno);
    robust_close(p, fd, __LINE__);
    return SQLITE_IOERR_ACCESS;
  }
  if (0 == strncmp("msdos", fsInfo.f_fstypename, 5)) {
    ((unixFile*)pFile)->fsFlags |= SQLITE_FSFLAGS_IS_MSDOS;
  }
  if (0 == strncmp("exfat", fsInfo.f_fstypename, 5)) {
    ((unixFile*)pFile)->fsFlags |= SQLITE_FSFLAGS_IS_MSDOS;
  }
#endif

  /* Set up appropriate ctrlFlags */
  if( isDelete )                ctrlFlags |= UNIXFILE_DELETE;
  if( isReadonly )              ctrlFlags |= UNIXFILE_RDONLY;
  noLock = eType!=SQLITE_OPEN_MAIN_DB;
  if( noLock )                  ctrlFlags |= UNIXFILE_NOLOCK;
  if( isNewJrnl )               ctrlFlags |= UNIXFILE_DIRSYNC;
  if( flags & SQLITE_OPEN_URI ) ctrlFlags |= UNIXFILE_URI;

#if SQLITE_ENABLE_LOCKING_STYLE
#if SQLITE_PREFER_PROXY_LOCKING
  isAutoProxy = 1;
#endif
  if( isAutoProxy && (zPath!=NULL) && (!noLock) && pVfs->xOpen ){
    char *envforce = getenv("SQLITE_FORCE_PROXY_LOCKING");
    int useProxy = 0;

    /* SQLITE_FORCE_PROXY_LOCKING==1 means force always use proxy, 0 means 
    ** never use proxy, NULL means use proxy for non-local files only.  */
    if( envforce!=NULL ){
      useProxy = atoi(envforce)>0;
    }else{
      useProxy = !(fsInfo.f_flags&MNT_LOCAL);
    }
    if( useProxy ){
      rc = fillInUnixFile(pVfs, fd, pFile, zPath, ctrlFlags);
      if( rc==SQLITE_OK ){
        rc = proxyTransformUnixFile((unixFile*)pFile, ":auto:");
        if( rc!=SQLITE_OK ){
          /* Use unixClose to clean up the resources added in fillInUnixFile 
          ** and clear all the structure's references.  Specifically, 
          ** pFile->pMethods will be NULL so sqlite3OsClose will be a no-op 
          */
          unixClose(pFile);
          return rc;
        }
      }
      goto open_finished;
    }
  }
#endif
  
  assert( zPath==0 || zPath[0]=='/' 
      || eType==SQLITE_OPEN_MASTER_JOURNAL || eType==SQLITE_OPEN_MAIN_JOURNAL 
  );
  rc = fillInUnixFile(pVfs, fd, pFile, zPath, ctrlFlags);

open_finished:
  if( rc!=SQLITE_OK ){
    sqlite3_free(p->pPreallocatedUnused);
  }
  return rc;
}


/*
** Delete the file at zPath. If the dirSync argument is true, fsync()
** the directory after deleting the file.
*/
static int unixDelete(
  sqlite3_vfs *NotUsed,     /* VFS containing this as the xDelete method */
  const char *zPath,        /* Name of file to be deleted */
  int dirSync               /* If true, fsync() directory after deleting file */
){
  int rc = SQLITE_OK;
  UNUSED_PARAMETER(NotUsed);
  SimulateIOError(return SQLITE_IOERR_DELETE);
  if( osUnlink(zPath)==(-1) ){
    if( errno==ENOENT
#if OS_VXWORKS
        || osAccess(zPath,0)!=0
#endif
    ){
      rc = SQLITE_IOERR_DELETE_NOENT;
    }else{
      rc = unixLogError(SQLITE_IOERR_DELETE, "unlink", zPath);
    }
    return rc;
  }
#ifndef SQLITE_DISABLE_DIRSYNC
  if( (dirSync & 1)!=0 ){
    int fd;
    rc = osOpenDirectory(zPath, &fd);
    if( rc==SQLITE_OK ){
      if( full_fsync(fd,0,0) ){
        rc = unixLogError(SQLITE_IOERR_DIR_FSYNC, "fsync", zPath);
      }
      robust_close(0, fd, __LINE__);
    }else{
      assert( rc==SQLITE_CANTOPEN );
      rc = SQLITE_OK;
    }
  }
#endif
  return rc;
}

/*
** Test the existence of or access permissions of file zPath. The
** test performed depends on the value of flags:
**
**     SQLITE_ACCESS_EXISTS: Return 1 if the file exists
**     SQLITE_ACCESS_READWRITE: Return 1 if the file is read and writable.
**     SQLITE_ACCESS_READONLY: Return 1 if the file is readable.
**
** Otherwise return 0.
*/
static int unixAccess(
  sqlite3_vfs *NotUsed,   /* The VFS containing this xAccess method */
  const char *zPath,      /* Path of the file to examine */
  int flags,              /* What do we want to learn about the zPath file? */
  int *pResOut            /* Write result boolean here */
){
  UNUSED_PARAMETER(NotUsed);
  SimulateIOError( return SQLITE_IOERR_ACCESS; );
  assert( pResOut!=0 );

  /* The spec says there are three possible values for flags.  But only
  ** two of them are actually used */
  assert( flags==SQLITE_ACCESS_EXISTS || flags==SQLITE_ACCESS_READWRITE );

  if( flags==SQLITE_ACCESS_EXISTS ){
    struct stat buf;
    *pResOut = (0==osStat(zPath, &buf) && buf.st_size>0);
  }else{
    *pResOut = osAccess(zPath, W_OK|R_OK)==0;
  }
  return SQLITE_OK;
}

/*
**
*/
static int mkFullPathname(
  const char *zPath,              /* Input path */
  char *zOut,                     /* Output buffer */
  int nOut                        /* Allocated size of buffer zOut */
){
  int nPath = sqlite3Strlen30(zPath);
  int iOff = 0;
  if( zPath[0]!='/' ){
    if( osGetcwd(zOut, nOut-2)==0 ){
      return unixLogError(SQLITE_CANTOPEN_BKPT, "getcwd", zPath);
    }
    iOff = sqlite3Strlen30(zOut);
    zOut[iOff++] = '/';
  }
  if( (iOff+nPath+1)>nOut ){
    /* SQLite assumes that xFullPathname() nul-terminates the output buffer
    ** even if it returns an error.  */
    zOut[iOff] = '\0';
    return SQLITE_CANTOPEN_BKPT;
  }
  sqlite3_snprintf(nOut-iOff, &zOut[iOff], "%s", zPath);
  return SQLITE_OK;
}

/*
** Turn a relative pathname into a full pathname. The relative path
** is stored as a nul-terminated string in the buffer pointed to by
** zPath. 
**
** zOut points to a buffer of at least sqlite3_vfs.mxPathname bytes 
** (in this case, MAX_PATHNAME bytes). The full-path is written to
** this buffer before returning.
*/
static int unixFullPathname(
  sqlite3_vfs *pVfs,            /* Pointer to vfs object */
  const char *zPath,            /* Possibly relative input path */
  int nOut,                     /* Size of output buffer in bytes */
  char *zOut                    /* Output buffer */
){
#if !defined(HAVE_READLINK) || !defined(HAVE_LSTAT)
  return mkFullPathname(zPath, zOut, nOut);
#else
  int rc = SQLITE_OK;
  int nByte;
  int nLink = 1;                /* Number of symbolic links followed so far */
  const char *zIn = zPath;      /* Input path for each iteration of loop */
  char *zDel = 0;

  assert( pVfs->mxPathname==MAX_PATHNAME );
  UNUSED_PARAMETER(pVfs);

  /* It's odd to simulate an io-error here, but really this is just
  ** using the io-error infrastructure to test that SQLite handles this
  ** function failing. This function could fail if, for example, the
  ** current working directory has been unlinked.
  */
  SimulateIOError( return SQLITE_ERROR );

  do {

    /* Call stat() on path zIn. Set bLink to true if the path is a symbolic
    ** link, or false otherwise.  */
    int bLink = 0;
    struct stat buf;
    if( osLstat(zIn, &buf)!=0 ){
      if( errno!=ENOENT ){
        rc = unixLogError(SQLITE_CANTOPEN_BKPT, "lstat", zIn);
      }
    }else{
      bLink = S_ISLNK(buf.st_mode);
    }

    if( bLink ){
      if( zDel==0 ){
        zDel = sqlite3_malloc(nOut);
        if( zDel==0 ) rc = SQLITE_NOMEM_BKPT;
      }else if( ++nLink>SQLITE_MAX_SYMLINKS ){
        rc = SQLITE_CANTOPEN_BKPT;
      }

      if( rc==SQLITE_OK ){
        nByte = osReadlink(zIn, zDel, nOut-1);
        if( nByte<0 ){
          rc = unixLogError(SQLITE_CANTOPEN_BKPT, "readlink", zIn);
        }else{
          if( zDel[0]!='/' ){
            int n;
            for(n = sqlite3Strlen30(zIn); n>0 && zIn[n-1]!='/'; n--);
            if( nByte+n+1>nOut ){
              rc = SQLITE_CANTOPEN_BKPT;
            }else{
              memmove(&zDel[n], zDel, nByte+1);
              memcpy(zDel, zIn, n);
              nByte += n;
            }
          }
          zDel[nByte] = '\0';
        }
      }

      zIn = zDel;
    }

    assert( rc!=SQLITE_OK || zIn!=zOut || zIn[0]=='/' );
    if( rc==SQLITE_OK && zIn!=zOut ){
      rc = mkFullPathname(zIn, zOut, nOut);
    }
    if( bLink==0 ) break;
    zIn = zOut;
  }while( rc==SQLITE_OK );

  sqlite3_free(zDel);
  return rc;
#endif   /* HAVE_READLINK && HAVE_LSTAT */
}


#ifndef SQLITE_OMIT_LOAD_EXTENSION
/*
** Interfaces for opening a shared library, finding entry points
** within the shared library, and closing the shared library.
*/
#include <dlfcn.h>
static void *unixDlOpen(sqlite3_vfs *NotUsed, const char *zFilename){
  UNUSED_PARAMETER(NotUsed);
  return dlopen(zFilename, RTLD_NOW | RTLD_GLOBAL);
}

/*
** SQLite calls this function immediately after a call to unixDlSym() or
** unixDlOpen() fails (returns a null pointer). If a more detailed error
** message is available, it is written to zBufOut. If no error message
** is available, zBufOut is left unmodified and SQLite uses a default
** error message.
*/
static void unixDlError(sqlite3_vfs *NotUsed, int nBuf, char *zBufOut){
  const char *zErr;
  UNUSED_PARAMETER(NotUsed);
  unixEnterMutex();
  zErr = dlerror();
  if( zErr ){
    sqlite3_snprintf(nBuf, zBufOut, "%s", zErr);
  }
  unixLeaveMutex();
}
static void (*unixDlSym(sqlite3_vfs *NotUsed, void *p, const char*zSym))(void){
  /* 
  ** GCC with -pedantic-errors says that C90 does not allow a void* to be
  ** cast into a pointer to a function.  And yet the library dlsym() routine
  ** returns a void* which is really a pointer to a function.  So how do we
  ** use dlsym() with -pedantic-errors?
  **
  ** Variable x below is defined to be a pointer to a function taking
  ** parameters void* and const char* and returning a pointer to a function.
  ** We initialize x by assigning it a pointer to the dlsym() function.
  ** (That assignment requires a cast.)  Then we call the function that
  ** x points to.  
  **
  ** This work-around is unlikely to work correctly on any system where
  ** you really cannot cast a function pointer into void*.  But then, on the
  ** other hand, dlsym() will not work on such a system either, so we have
  ** not really lost anything.
  */
  void (*(*x)(void*,const char*))(void);
  UNUSED_PARAMETER(NotUsed);
  x = (void(*(*)(void*,const char*))(void))dlsym;
  return (*x)(p, zSym);
}
static void unixDlClose(sqlite3_vfs *NotUsed, void *pHandle){
  UNUSED_PARAMETER(NotUsed);
  dlclose(pHandle);
}
#else /* if SQLITE_OMIT_LOAD_EXTENSION is defined: */
  #define unixDlOpen  0
  #define unixDlError 0
  #define unixDlSym   0
  #define unixDlClose 0
#endif

/*
** Write nBuf bytes of random data to the supplied buffer zBuf.
*/
static int unixRandomness(sqlite3_vfs *NotUsed, int nBuf, char *zBuf){
  UNUSED_PARAMETER(NotUsed);
  assert((size_t)nBuf>=(sizeof(time_t)+sizeof(int)));

  /* We have to initialize zBuf to prevent valgrind from reporting
  ** errors.  The reports issued by valgrind are incorrect - we would
  ** prefer that the randomness be increased by making use of the
  ** uninitialized space in zBuf - but valgrind errors tend to worry
  ** some users.  Rather than argue, it seems easier just to initialize
  ** the whole array and silence valgrind, even if that means less randomness
  ** in the random seed.
  **
  ** When testing, initializing zBuf[] to zero is all we do.  That means
  ** that we always use the same random number sequence.  This makes the
  ** tests repeatable.
  */
  memset(zBuf, 0, nBuf);
  randomnessPid = osGetpid(0);  
#if !defined(SQLITE_TEST) && !defined(SQLITE_OMIT_RANDOMNESS)
  {
    int fd, got;
    fd = robust_open("/dev/urandom", O_RDONLY, 0);
    if( fd<0 ){
      time_t t;
      time(&t);
      memcpy(zBuf, &t, sizeof(t));
      memcpy(&zBuf[sizeof(t)], &randomnessPid, sizeof(randomnessPid));
      assert( sizeof(t)+sizeof(randomnessPid)<=(size_t)nBuf );
      nBuf = sizeof(t) + sizeof(randomnessPid);
    }else{
      do{ got = osRead(fd, zBuf, nBuf); }while( got<0 && errno==EINTR );
      robust_close(0, fd, __LINE__);
    }
  }
#endif
  return nBuf;
}


/*
** Sleep for a little while.  Return the amount of time slept.
** The argument is the number of microseconds we want to sleep.
** The return value is the number of microseconds of sleep actually
** requested from the underlying operating system, a number which
** might be greater than or equal to the argument, but not less
** than the argument.
*/
static int unixSleep(sqlite3_vfs *NotUsed, int microseconds){
#if OS_VXWORKS
  struct timespec sp;

  sp.tv_sec = microseconds / 1000000;
  sp.tv_nsec = (microseconds % 1000000) * 1000;
  nanosleep(&sp, NULL);
  UNUSED_PARAMETER(NotUsed);
  return microseconds;
#elif defined(HAVE_USLEEP) && HAVE_USLEEP
  usleep(microseconds);
  UNUSED_PARAMETER(NotUsed);
  return microseconds;
#else
  int seconds = (microseconds+999999)/1000000;
  sleep(seconds);
  UNUSED_PARAMETER(NotUsed);
  return seconds*1000000;
#endif
}

/*
** The following variable, if set to a non-zero value, is interpreted as
** the number of seconds since 1970 and is used to set the result of
** sqlite3OsCurrentTime() during testing.
*/
#ifdef SQLITE_TEST
int sqlite3_current_time = 0;  /* Fake system time in seconds since 1970. */
#endif

/*
** Find the current time (in Universal Coordinated Time).  Write into *piNow
** the current time and date as a Julian Day number times 86_400_000.  In
** other words, write into *piNow the number of milliseconds since the Julian
** epoch of noon in Greenwich on November 24, 4714 B.C according to the
** proleptic Gregorian calendar.
**
** On success, return SQLITE_OK.  Return SQLITE_ERROR if the time and date 
** cannot be found.
*/
static int unixCurrentTimeInt64(sqlite3_vfs *NotUsed, sqlite3_int64 *piNow){
  static const sqlite3_int64 unixEpoch = 24405875*(sqlite3_int64)8640000;
  int rc = SQLITE_OK;
#if defined(NO_GETTOD)
  time_t t;
  time(&t);
  *piNow = ((sqlite3_int64)t)*1000 + unixEpoch;
#elif OS_VXWORKS
  struct timespec sNow;
  clock_gettime(CLOCK_REALTIME, &sNow);
  *piNow = unixEpoch + 1000*(sqlite3_int64)sNow.tv_sec + sNow.tv_nsec/1000000;
#else
  struct timeval sNow;
  (void)gettimeofday(&sNow, 0);  /* Cannot fail given valid arguments */
  *piNow = unixEpoch + 1000*(sqlite3_int64)sNow.tv_sec + sNow.tv_usec/1000;
#endif

#ifdef SQLITE_TEST
  if( sqlite3_current_time ){
    *piNow = 1000*(sqlite3_int64)sqlite3_current_time + unixEpoch;
  }
#endif
  UNUSED_PARAMETER(NotUsed);
  return rc;
}

#ifndef SQLITE_OMIT_DEPRECATED
/*
** Find the current time (in Universal Coordinated Time).  Write the
** current time and date as a Julian Day number into *prNow and
** return 0.  Return 1 if the time and date cannot be found.
*/
static int unixCurrentTime(sqlite3_vfs *NotUsed, double *prNow){
  sqlite3_int64 i = 0;
  int rc;
  UNUSED_PARAMETER(NotUsed);
  rc = unixCurrentTimeInt64(0, &i);
  *prNow = i/86400000.0;
  return rc;
}
#else
# define unixCurrentTime 0
#endif

/*
** The xGetLastError() method is designed to return a better
** low-level error message when operating-system problems come up
** during SQLite operation.  Only the integer return code is currently
** used.
*/
static int unixGetLastError(sqlite3_vfs *NotUsed, int NotUsed2, char *NotUsed3){
  UNUSED_PARAMETER(NotUsed);
  UNUSED_PARAMETER(NotUsed2);
  UNUSED_PARAMETER(NotUsed3);
  return errno;
}


/*
************************ End of sqlite3_vfs methods ***************************
******************************************************************************/

/******************************************************************************
************************** Begin Proxy Locking ********************************
**
** Proxy locking is a "uber-locking-method" in this sense:  It uses the
** other locking methods on secondary lock files.  Proxy locking is a
** meta-layer over top of the primitive locking implemented above.  For
** this reason, the division that implements of proxy locking is deferred
** until late in the file (here) after all of the other I/O methods have
** been defined - so that the primitive locking methods are available
** as services to help with the implementation of proxy locking.
**
****
**
** The default locking schemes in SQLite use byte-range locks on the
** database file to coordinate safe, concurrent access by multiple readers
** and writers [http://sqlite.org/lockingv3.html].  The five file locking
** states (UNLOCKED, PENDING, SHARED, RESERVED, EXCLUSIVE) are implemented
** as POSIX read & write locks over fixed set of locations (via fsctl),
** on AFP and SMB only exclusive byte-range locks are available via fsctl
** with _IOWR('z', 23, struct ByteRangeLockPB2) to track the same 5 states.
** To simulate a F_RDLCK on the shared range, on AFP a randomly selected
** address in the shared range is taken for a SHARED lock, the entire
** shared range is taken for an EXCLUSIVE lock):
**
**      PENDING_BYTE        0x40000000
**      RESERVED_BYTE       0x40000001
**      SHARED_RANGE        0x40000002 -> 0x40000200
**
** This works well on the local file system, but shows a nearly 100x
** slowdown in read performance on AFP because the AFP client disables
** the read cache when byte-range locks are present.  Enabling the read
** cache exposes a cache coherency problem that is present on all OS X
** supported network file systems.  NFS and AFP both observe the
** close-to-open semantics for ensuring cache coherency
** [http://nfs.sourceforge.net/#faq_a8], which does not effectively
** address the requirements for concurrent database access by multiple
** readers and writers
** [http://www.nabble.com/SQLite-on-NFS-cache-coherency-td15655701.html].
**
** To address the performance and cache coherency issues, proxy file locking
** changes the way database access is controlled by limiting access to a
** single host at a time and moving file locks off of the database file
** and onto a proxy file on the local file system.  
**
**
** Using proxy locks
** -----------------
**
** C APIs
**
**  sqlite3_file_control(db, dbname, SQLITE_FCNTL_SET_LOCKPROXYFILE,
**                       <proxy_path> | ":auto:");
**  sqlite3_file_control(db, dbname, SQLITE_FCNTL_GET_LOCKPROXYFILE,
**                       &<proxy_path>);
**
**
** SQL pragmas
**
**  PRAGMA [database.]lock_proxy_file=<proxy_path> | :auto:
**  PRAGMA [database.]lock_proxy_file
**
** Specifying ":auto:" means that if there is a conch file with a matching
** host ID in it, the proxy path in the conch file will be used, otherwise
** a proxy path based on the user's temp dir
** (via confstr(_CS_DARWIN_USER_TEMP_DIR,...)) will be used and the
** actual proxy file name is generated from the name and path of the
** database file.  For example:
**
**       For database path "/Users/me/foo.db" 
**       The lock path will be "<tmpdir>/sqliteplocks/_Users_me_foo.db:auto:")
**
** Once a lock proxy is configured for a database connection, it can not
** be removed, however it may be switched to a different proxy path via
** the above APIs (assuming the conch file is not being held by another
** connection or process). 
**
**
** How proxy locking works
** -----------------------
**
** Proxy file locking relies primarily on two new supporting files: 
**
**   *  conch file to limit access to the database file to a single host
**      at a time
**
**   *  proxy file to act as a proxy for the advisory locks normally
**      taken on the database
**
** The conch file - to use a proxy file, sqlite must first "hold the conch"
** by taking an sqlite-style shared lock on the conch file, reading the
** contents and comparing the host's unique host ID (see below) and lock
** proxy path against the values stored in the conch.  The conch file is
** stored in the same directory as the database file and the file name
** is patterned after the database file name as ".<databasename>-conch".
** If the conch file does not exist, or its contents do not match the
** host ID and/or proxy path, then the lock is escalated to an exclusive
** lock and the conch file contents is updated with the host ID and proxy
** path and the lock is downgraded to a shared lock again.  If the conch
** is held by another process (with a shared lock), the exclusive lock
** will fail and SQLITE_BUSY is returned.
**
** The proxy file - a single-byte file used for all advisory file locks
** normally taken on the database file.   This allows for safe sharing
** of the database file for multiple readers and writers on the same
** host (the conch ensures that they all use the same local lock file).
**
** Requesting the lock proxy does not immediately take the conch, it is
** only taken when the first request to lock database file is made.  
** This matches the semantics of the traditional locking behavior, where
** opening a connection to a database file does not take a lock on it.
** The shared lock and an open file descriptor are maintained until 
** the connection to the database is closed. 
**
** The proxy file and the lock file are never deleted so they only need
** to be created the first time they are used.
**
** Configuration options
** ---------------------
**
**  SQLITE_PREFER_PROXY_LOCKING
**
**       Database files accessed on non-local file systems are
**       automatically configured for proxy locking, lock files are
**       named automatically using the same logic as
**       PRAGMA lock_proxy_file=":auto:"
**    
**  SQLITE_PROXY_DEBUG
**
**       Enables the logging of error messages during host id file
**       retrieval and creation
**
**  LOCKPROXYDIR
**
**       Overrides the default directory used for lock proxy files that
**       are named automatically via the ":auto:" setting
**
**  SQLITE_DEFAULT_PROXYDIR_PERMISSIONS
**
**       Permissions to use when creating a directory for storing the
**       lock proxy files, only used when LOCKPROXYDIR is not set.
**    
**    
** As mentioned above, when compiled with SQLITE_PREFER_PROXY_LOCKING,
** setting the environment variable SQLITE_FORCE_PROXY_LOCKING to 1 will
** force proxy locking to be used for every database file opened, and 0
** will force automatic proxy locking to be disabled for all database
** files (explicitly calling the SQLITE_FCNTL_SET_LOCKPROXYFILE pragma or
** sqlite_file_control API is not affected by SQLITE_FORCE_PROXY_LOCKING).
*/

/*
** Proxy locking is only available on MacOSX 
*/
#if defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE

/*
** The proxyLockingContext has the path and file structures for the remote 
** and local proxy files in it
*/
typedef struct proxyLockingContext proxyLockingContext;
struct proxyLockingContext {
  unixFile *conchFile;         /* Open conch file */
  char *conchFilePath;         /* Name of the conch file */
  unixFile *lockProxy;         /* Open proxy lock file */
  char *lockProxyPath;         /* Name of the proxy lock file */
  char *dbPath;                /* Name of the open file */
  int conchHeld;               /* 1 if the conch is held, -1 if lockless */
  int nFails;                  /* Number of conch taking failures */
  void *oldLockingContext;     /* Original lockingcontext to restore on close */
  sqlite3_io_methods const *pOldMethod;     /* Original I/O methods for close */
};

/* 
** The proxy lock file path for the database at dbPath is written into lPath, 
** which must point to valid, writable memory large enough for a maxLen length
** file path. 
*/
static int proxyGetLockPath(const char *dbPath, char *lPath, size_t maxLen){
  int len;
  int dbLen;
  int i;

#ifdef LOCKPROXYDIR
  len = strlcpy(lPath, LOCKPROXYDIR, maxLen);
#else
# ifdef _CS_DARWIN_USER_TEMP_DIR
  {
    if( !confstr(_CS_DARWIN_USER_TEMP_DIR, lPath, maxLen) ){
      OSTRACE(("GETLOCKPATH  failed %s errno=%d pid=%d\n",
               lPath, errno, osGetpid(0)));
      return SQLITE_IOERR_LOCK;
    }
    len = strlcat(lPath, "sqliteplocks", maxLen);    
  }
# else
  len = strlcpy(lPath, "/tmp/", maxLen);
# endif
#endif

  if( lPath[len-1]!='/' ){
    len = strlcat(lPath, "/", maxLen);
  }
  
  /* transform the db path to a unique cache name */
  dbLen = (int)strlen(dbPath);
  for( i=0; i<dbLen && (i+len+7)<(int)maxLen; i++){
    char c = dbPath[i];
    lPath[i+len] = (c=='/')?'_':c;
  }
  lPath[i+len]='\0';
  strlcat(lPath, ":auto:", maxLen);
  OSTRACE(("GETLOCKPATH  proxy lock path=%s pid=%d\n", lPath, osGetpid(0)));
  return SQLITE_OK;
}

/* 
 ** Creates the lock file and any missing directories in lockPath
 */
static int proxyCreateLockPath(const char *lockPath){
  int i, len;
  char buf[MAXPATHLEN];
  int start = 0;
  
  assert(lockPath!=NULL);
  /* try to create all the intermediate directories */
  len = (int)strlen(lockPath);
  buf[0] = lockPath[0];
  for( i=1; i<len; i++ ){
    if( lockPath[i] == '/' && (i - start > 0) ){
      /* only mkdir if leaf dir != "." or "/" or ".." */
      if( i-start>2 || (i-start==1 && buf[start] != '.' && buf[start] != '/') 
         || (i-start==2 && buf[start] != '.' && buf[start+1] != '.') ){
        buf[i]='\0';
        if( osMkdir(buf, SQLITE_DEFAULT_PROXYDIR_PERMISSIONS) ){
          int err=errno;
          if( err!=EEXIST ) {
            OSTRACE(("CREATELOCKPATH  FAILED creating %s, "
                     "'%s' proxy lock path=%s pid=%d\n",
                     buf, strerror(err), lockPath, osGetpid(0)));
            return err;
          }
        }
      }
      start=i+1;
    }
    buf[i] = lockPath[i];
  }
  OSTRACE(("CREATELOCKPATH  proxy lock path=%s pid=%d\n",lockPath,osGetpid(0)));
  return 0;
}

/*
** Create a new VFS file descriptor (stored in memory obtained from
** sqlite3_malloc) and open the file named "path" in the file descriptor.
**
** The caller is responsible not only for closing the file descriptor
** but also for freeing the memory associated with the file descriptor.
*/
static int proxyCreateUnixFile(
    const char *path,        /* path for the new unixFile */
    unixFile **ppFile,       /* unixFile created and returned by ref */
    int islockfile           /* if non zero missing dirs will be created */
) {
  int fd = -1;
  unixFile *pNew;
  int rc = SQLITE_OK;
  int openFlags = O_RDWR | O_CREAT;
  sqlite3_vfs dummyVfs;
  int terrno = 0;
  UnixUnusedFd *pUnused = NULL;

  /* 1. first try to open/create the file
  ** 2. if that fails, and this is a lock file (not-conch), try creating
  ** the parent directories and then try again.
  ** 3. if that fails, try to open the file read-only
  ** otherwise return BUSY (if lock file) or CANTOPEN for the conch file
  */
  pUnused = findReusableFd(path, openFlags);
  if( pUnused ){
    fd = pUnused->fd;
  }else{
    pUnused = sqlite3_malloc64(sizeof(*pUnused));
    if( !pUnused ){
      return SQLITE_NOMEM_BKPT;
    }
  }
  if( fd<0 ){
    fd = robust_open(path, openFlags, 0);
    terrno = errno;
    if( fd<0 && errno==ENOENT && islockfile ){
      if( proxyCreateLockPath(path) == SQLITE_OK ){
        fd = robust_open(path, openFlags, 0);
      }
    }
  }
  if( fd<0 ){
    openFlags = O_RDONLY;
    fd = robust_open(path, openFlags, 0);
    terrno = errno;
  }
  if( fd<0 ){
    if( islockfile ){
      return SQLITE_BUSY;
    }
    switch (terrno) {
      case EACCES:
        return SQLITE_PERM;
      case EIO: 
        return SQLITE_IOERR_LOCK; /* even though it is the conch */
      default:
        return SQLITE_CANTOPEN_BKPT;
    }
  }
  
  pNew = (unixFile *)sqlite3_malloc64(sizeof(*pNew));
  if( pNew==NULL ){
    rc = SQLITE_NOMEM_BKPT;
    goto end_create_proxy;
  }
  memset(pNew, 0, sizeof(unixFile));
  pNew->openFlags = openFlags;
  memset(&dummyVfs, 0, sizeof(dummyVfs));
  dummyVfs.pAppData = (void*)&autolockIoFinder;
  dummyVfs.zName = "dummy";
  pUnused->fd = fd;
  pUnused->flags = openFlags;
  pNew->pPreallocatedUnused = pUnused;
  
  rc = fillInUnixFile(&dummyVfs, fd, (sqlite3_file*)pNew, path, 0);
  if( rc==SQLITE_OK ){
    *ppFile = pNew;
    return SQLITE_OK;
  }
end_create_proxy:    
  robust_close(pNew, fd, __LINE__);
  sqlite3_free(pNew);
  sqlite3_free(pUnused);
  return rc;
}

#ifdef SQLITE_TEST
/* simulate multiple hosts by creating unique hostid file paths */
int sqlite3_hostid_num = 0;
#endif

#define PROXY_HOSTIDLEN    16  /* conch file host id length */

#ifdef HAVE_GETHOSTUUID
/* Not always defined in the headers as it ought to be */
extern int gethostuuid(uuid_t id, const struct timespec *wait);
#endif

/* get the host ID via gethostuuid(), pHostID must point to PROXY_HOSTIDLEN 
** bytes of writable memory.
*/
static int proxyGetHostID(unsigned char *pHostID, int *pError){
  assert(PROXY_HOSTIDLEN == sizeof(uuid_t));
  memset(pHostID, 0, PROXY_HOSTIDLEN);
#ifdef HAVE_GETHOSTUUID
  {
    struct timespec timeout = {1, 0}; /* 1 sec timeout */
    if( gethostuuid(pHostID, &timeout) ){
      int err = errno;
      if( pError ){
        *pError = err;
      }
      return SQLITE_IOERR;
    }
  }
#else
  UNUSED_PARAMETER(pError);
#endif
#ifdef SQLITE_TEST
  /* simulate multiple hosts by creating unique hostid file paths */
  if( sqlite3_hostid_num != 0){
    pHostID[0] = (char)(pHostID[0] + (char)(sqlite3_hostid_num & 0xFF));
  }
#endif
  
  return SQLITE_OK;
}

/* The conch file contains the header, host id and lock file path
 */
#define PROXY_CONCHVERSION 2   /* 1-byte header, 16-byte host id, path */
#define PROXY_HEADERLEN    1   /* conch file header length */
#define PROXY_PATHINDEX    (PROXY_HEADERLEN+PROXY_HOSTIDLEN)
#define PROXY_MAXCONCHLEN  (PROXY_HEADERLEN+PROXY_HOSTIDLEN+MAXPATHLEN)

/* 
** Takes an open conch file, copies the contents to a new path and then moves 
** it back.  The newly created file's file descriptor is assigned to the
** conch file structure and finally the original conch file descriptor is 
** closed.  Returns zero if successful.
*/
static int proxyBreakConchLock(unixFile *pFile, uuid_t myHostID){
  proxyLockingContext *pCtx = (proxyLockingContext *)pFile->lockingContext; 
  unixFile *conchFile = pCtx->conchFile;
  char tPath[MAXPATHLEN];
  char buf[PROXY_MAXCONCHLEN];
  char *cPath = pCtx->conchFilePath;
  size_t readLen = 0;
  size_t pathLen = 0;
  char errmsg[64] = "";
  int fd = -1;
  int rc = -1;
  UNUSED_PARAMETER(myHostID);

  /* create a new path by replace the trailing '-conch' with '-break' */
  pathLen = strlcpy(tPath, cPath, MAXPATHLEN);
  if( pathLen>MAXPATHLEN || pathLen<6 || 
     (strlcpy(&tPath[pathLen-5], "break", 6) != 5) ){
    sqlite3_snprintf(sizeof(errmsg),errmsg,"path error (len %d)",(int)pathLen);
    goto end_breaklock;
  }
  /* read the conch content */
  readLen = osPread(conchFile->h, buf, PROXY_MAXCONCHLEN, 0);
  if( readLen<PROXY_PATHINDEX ){
    sqlite3_snprintf(sizeof(errmsg),errmsg,"read error (len %d)",(int)readLen);
    goto end_breaklock;
  }
  /* write it out to the temporary break file */
  fd = robust_open(tPath, (O_RDWR|O_CREAT|O_EXCL), 0);
  if( fd<0 ){
    sqlite3_snprintf(sizeof(errmsg), errmsg, "create failed (%d)", errno);
    goto end_breaklock;
  }
  if( osPwrite(fd, buf, readLen, 0) != (ssize_t)readLen ){
    sqlite3_snprintf(sizeof(errmsg), errmsg, "write failed (%d)", errno);
    goto end_breaklock;
  }
  if( rename(tPath, cPath) ){
    sqlite3_snprintf(sizeof(errmsg), errmsg, "rename failed (%d)", errno);
    goto end_breaklock;
  }
  rc = 0;
  fprintf(stderr, "broke stale lock on %s\n", cPath);
  robust_close(pFile, conchFile->h, __LINE__);
  conchFile->h = fd;
  conchFile->openFlags = O_RDWR | O_CREAT;

end_breaklock:
  if( rc ){
    if( fd>=0 ){
      osUnlink(tPath);
      robust_close(pFile, fd, __LINE__);
    }
    fprintf(stderr, "failed to break stale lock on %s, %s\n", cPath, errmsg);
  }
  return rc;
}

/* Take the requested lock on the conch file and break a stale lock if the 
** host id matches.
*/
static int proxyConchLock(unixFile *pFile, uuid_t myHostID, int lockType){
  proxyLockingContext *pCtx = (proxyLockingContext *)pFile->lockingContext; 
  unixFile *conchFile = pCtx->conchFile;
  int rc = SQLITE_OK;
  int nTries = 0;
  struct timespec conchModTime;
  
  memset(&conchModTime, 0, sizeof(conchModTime));
  do {
    rc = conchFile->pMethod->xLock((sqlite3_file*)conchFile, lockType);
    nTries ++;
    if( rc==SQLITE_BUSY ){
      /* If the lock failed (busy):
       * 1st try: get the mod time of the conch, wait 0.5s and try again. 
       * 2nd try: fail if the mod time changed or host id is different, wait 
       *           10 sec and try again
       * 3rd try: break the lock unless the mod time has changed.
       */
      struct stat buf;
      if( osFstat(conchFile->h, &buf) ){
        storeLastErrno(pFile, errno);
        return SQLITE_IOERR_LOCK;
      }
      
      if( nTries==1 ){
        conchModTime = buf.st_mtimespec;
        usleep(500000); /* wait 0.5 sec and try the lock again*/
        continue;  
      }

      assert( nTries>1 );
      if( conchModTime.tv_sec != buf.st_mtimespec.tv_sec || 
         conchModTime.tv_nsec != buf.st_mtimespec.tv_nsec ){
        return SQLITE_BUSY;
      }
      
      if( nTries==2 ){  
        char tBuf[PROXY_MAXCONCHLEN];
        int len = osPread(conchFile->h, tBuf, PROXY_MAXCONCHLEN, 0);
        if( len<0 ){
          storeLastErrno(pFile, errno);
          return SQLITE_IOERR_LOCK;
        }
        if( len>PROXY_PATHINDEX && tBuf[0]==(char)PROXY_CONCHVERSION){
          /* don't break the lock if the host id doesn't match */
          if( 0!=memcmp(&tBuf[PROXY_HEADERLEN], myHostID, PROXY_HOSTIDLEN) ){
            return SQLITE_BUSY;
          }
        }else{
          /* don't break the lock on short read or a version mismatch */
          return SQLITE_BUSY;
        }
        usleep(10000000); /* wait 10 sec and try the lock again */
        continue; 
      }
      
      assert( nTries==3 );
      if( 0==proxyBreakConchLock(pFile, myHostID) ){
        rc = SQLITE_OK;
        if( lockType==EXCLUSIVE_LOCK ){
          rc = conchFile->pMethod->xLock((sqlite3_file*)conchFile, SHARED_LOCK);
        }
        if( !rc ){
          rc = conchFile->pMethod->xLock((sqlite3_file*)conchFile, lockType);
        }
      }
    }
  } while( rc==SQLITE_BUSY && nTries<3 );
  
  return rc;
}

/* Takes the conch by taking a shared lock and read the contents conch, if 
** lockPath is non-NULL, the host ID and lock file path must match.  A NULL 
** lockPath means that the lockPath in the conch file will be used if the 
** host IDs match, or a new lock path will be generated automatically 
** and written to the conch file.
*/
static int proxyTakeConch(unixFile *pFile){
  proxyLockingContext *pCtx = (proxyLockingContext *)pFile->lockingContext; 
  
  if( pCtx->conchHeld!=0 ){
    return SQLITE_OK;
  }else{
    unixFile *conchFile = pCtx->conchFile;
    uuid_t myHostID;
    int pError = 0;
    char readBuf[PROXY_MAXCONCHLEN];
    char lockPath[MAXPATHLEN];
    char *tempLockPath = NULL;
    int rc = SQLITE_OK;
    int createConch = 0;
    int hostIdMatch = 0;
    int readLen = 0;
    int tryOldLockPath = 0;
    int forceNewLockPath = 0;
    
    OSTRACE(("TAKECONCH  %d for %s pid=%d\n", conchFile->h,
             (pCtx->lockProxyPath ? pCtx->lockProxyPath : ":auto:"),
             osGetpid(0)));

    rc = proxyGetHostID(myHostID, &pError);
    if( (rc&0xff)==SQLITE_IOERR ){
      storeLastErrno(pFile, pError);
      goto end_takeconch;
    }
    rc = proxyConchLock(pFile, myHostID, SHARED_LOCK);
    if( rc!=SQLITE_OK ){
      goto end_takeconch;
    }
    /* read the existing conch file */
    readLen = seekAndRead((unixFile*)conchFile, 0, readBuf, PROXY_MAXCONCHLEN);
    if( readLen<0 ){
      /* I/O error: lastErrno set by seekAndRead */
      storeLastErrno(pFile, conchFile->lastErrno);
      rc = SQLITE_IOERR_READ;
      goto end_takeconch;
    }else if( readLen<=(PROXY_HEADERLEN+PROXY_HOSTIDLEN) || 
             readBuf[0]!=(char)PROXY_CONCHVERSION ){
      /* a short read or version format mismatch means we need to create a new 
      ** conch file. 
      */
      createConch = 1;
    }
    /* if the host id matches and the lock path already exists in the conch
    ** we'll try to use the path there, if we can't open that path, we'll 
    ** retry with a new auto-generated path 
    */
    do { /* in case we need to try again for an :auto: named lock file */

      if( !createConch && !forceNewLockPath ){
        hostIdMatch = !memcmp(&readBuf[PROXY_HEADERLEN], myHostID, 
                                  PROXY_HOSTIDLEN);
        /* if the conch has data compare the contents */
        if( !pCtx->lockProxyPath ){
          /* for auto-named local lock file, just check the host ID and we'll
           ** use the local lock file path that's already in there
           */
          if( hostIdMatch ){
            size_t pathLen = (readLen - PROXY_PATHINDEX);
            
            if( pathLen>=MAXPATHLEN ){
              pathLen=MAXPATHLEN-1;
            }
            memcpy(lockPath, &readBuf[PROXY_PATHINDEX], pathLen);
            lockPath[pathLen] = 0;
            tempLockPath = lockPath;
            tryOldLockPath = 1;
            /* create a copy of the lock path if the conch is taken */
            goto end_takeconch;
          }
        }else if( hostIdMatch
               && !strncmp(pCtx->lockProxyPath, &readBuf[PROXY_PATHINDEX],
                           readLen-PROXY_PATHINDEX)
        ){
          /* conch host and lock path match */
          goto end_takeconch; 
        }
      }
      
      /* if the conch isn't writable and doesn't match, we can't take it */
      if( (conchFile->openFlags&O_RDWR) == 0 ){
        rc = SQLITE_BUSY;
        goto end_takeconch;
      }
      
      /* either the conch didn't match or we need to create a new one */
      if( !pCtx->lockProxyPath ){
        proxyGetLockPath(pCtx->dbPath, lockPath, MAXPATHLEN);
        tempLockPath = lockPath;
        /* create a copy of the lock path _only_ if the conch is taken */
      }
      
      /* update conch with host and path (this will fail if other process
      ** has a shared lock already), if the host id matches, use the big
      ** stick.
      */
      futimes(conchFile->h, NULL);
      if( hostIdMatch && !createConch ){
        if( conchFile->pInode && conchFile->pInode->nShared>1 ){
          /* We are trying for an exclusive lock but another thread in this
           ** same process is still holding a shared lock. */
          rc = SQLITE_BUSY;
        } else {          
          rc = proxyConchLock(pFile, myHostID, EXCLUSIVE_LOCK);
        }
      }else{
        rc = proxyConchLock(pFile, myHostID, EXCLUSIVE_LOCK);
      }
      if( rc==SQLITE_OK ){
        char writeBuffer[PROXY_MAXCONCHLEN];
        int writeSize = 0;
        
        writeBuffer[0] = (char)PROXY_CONCHVERSION;
        memcpy(&writeBuffer[PROXY_HEADERLEN], myHostID, PROXY_HOSTIDLEN);
        if( pCtx->lockProxyPath!=NULL ){
          strlcpy(&writeBuffer[PROXY_PATHINDEX], pCtx->lockProxyPath,
                  MAXPATHLEN);
        }else{
          strlcpy(&writeBuffer[PROXY_PATHINDEX], tempLockPath, MAXPATHLEN);
        }
        writeSize = PROXY_PATHINDEX + strlen(&writeBuffer[PROXY_PATHINDEX]);
        robust_ftruncate(conchFile->h, writeSize);
        rc = unixWrite((sqlite3_file *)conchFile, writeBuffer, writeSize, 0);
        full_fsync(conchFile->h,0,0);
        /* If we created a new conch file (not just updated the contents of a 
         ** valid conch file), try to match the permissions of the database 
         */
        if( rc==SQLITE_OK && createConch ){
          struct stat buf;
          int err = osFstat(pFile->h, &buf);
          if( err==0 ){
            mode_t cmode = buf.st_mode&(S_IRUSR|S_IWUSR | S_IRGRP|S_IWGRP |
                                        S_IROTH|S_IWOTH);
            /* try to match the database file R/W permissions, ignore failure */
#ifndef SQLITE_PROXY_DEBUG
            osFchmod(conchFile->h, cmode);
#else
            do{
              rc = osFchmod(conchFile->h, cmode);
            }while( rc==(-1) && errno==EINTR );
            if( rc!=0 ){
              int code = errno;
              fprintf(stderr, "fchmod %o FAILED with %d %s\n",
                      cmode, code, strerror(code));
            } else {
              fprintf(stderr, "fchmod %o SUCCEDED\n",cmode);
            }
          }else{
            int code = errno;
            fprintf(stderr, "STAT FAILED[%d] with %d %s\n", 
                    err, code, strerror(code));
#endif
          }
        }
      }
      conchFile->pMethod->xUnlock((sqlite3_file*)conchFile, SHARED_LOCK);
      
    end_takeconch:
      OSTRACE(("TRANSPROXY: CLOSE  %d\n", pFile->h));
      if( rc==SQLITE_OK && pFile->openFlags ){
        int fd;
        if( pFile->h>=0 ){
          robust_close(pFile, pFile->h, __LINE__);
        }
        pFile->h = -1;
        fd = robust_open(pCtx->dbPath, pFile->openFlags, 0);
        OSTRACE(("TRANSPROXY: OPEN  %d\n", fd));
        if( fd>=0 ){
          pFile->h = fd;
        }else{
          rc=SQLITE_CANTOPEN_BKPT; /* SQLITE_BUSY? proxyTakeConch called
           during locking */
        }
      }
      if( rc==SQLITE_OK && !pCtx->lockProxy ){
        char *path = tempLockPath ? tempLockPath : pCtx->lockProxyPath;
        rc = proxyCreateUnixFile(path, &pCtx->lockProxy, 1);
        if( rc!=SQLITE_OK && rc!=SQLITE_NOMEM && tryOldLockPath ){
          /* we couldn't create the proxy lock file with the old lock file path
           ** so try again via auto-naming 
           */
          forceNewLockPath = 1;
          tryOldLockPath = 0;
          continue; /* go back to the do {} while start point, try again */
        }
      }
      if( rc==SQLITE_OK ){
        /* Need to make a copy of path if we extracted the value
         ** from the conch file or the path was allocated on the stack
         */
        if( tempLockPath ){
          pCtx->lockProxyPath = sqlite3DbStrDup(0, tempLockPath);
          if( !pCtx->lockProxyPath ){
            rc = SQLITE_NOMEM_BKPT;
          }
        }
      }
      if( rc==SQLITE_OK ){
        pCtx->conchHeld = 1;
        
        if( pCtx->lockProxy->pMethod == &afpIoMethods ){
          afpLockingContext *afpCtx;
          afpCtx = (afpLockingContext *)pCtx->lockProxy->lockingContext;
          afpCtx->dbPath = pCtx->lockProxyPath;
        }
      } else {
        conchFile->pMethod->xUnlock((sqlite3_file*)conchFile, NO_LOCK);
      }
      OSTRACE(("TAKECONCH  %d %s\n", conchFile->h,
               rc==SQLITE_OK?"ok":"failed"));
      return rc;
    } while (1); /* in case we need to retry the :auto: lock file - 
                 ** we should never get here except via the 'continue' call. */
  }
}

/*
** If pFile holds a lock on a conch file, then release that lock.
*/
static int proxyReleaseConch(unixFile *pFile){
  int rc = SQLITE_OK;         /* Subroutine return code */
  proxyLockingContext *pCtx;  /* The locking context for the proxy lock */
  unixFile *conchFile;        /* Name of the conch file */

  pCtx = (proxyLockingContext *)pFile->lockingContext;
  conchFile = pCtx->conchFile;
  OSTRACE(("RELEASECONCH  %d for %s pid=%d\n", conchFile->h,
           (pCtx->lockProxyPath ? pCtx->lockProxyPath : ":auto:"), 
           osGetpid(0)));
  if( pCtx->conchHeld>0 ){
    rc = conchFile->pMethod->xUnlock((sqlite3_file*)conchFile, NO_LOCK);
  }
  pCtx->conchHeld = 0;
  OSTRACE(("RELEASECONCH  %d %s\n", conchFile->h,
           (rc==SQLITE_OK ? "ok" : "failed")));
  return rc;
}

/*
** Given the name of a database file, compute the name of its conch file.
** Store the conch filename in memory obtained from sqlite3_malloc64().
** Make *pConchPath point to the new name.  Return SQLITE_OK on success
** or SQLITE_NOMEM if unable to obtain memory.
**
** The caller is responsible for ensuring that the allocated memory
** space is eventually freed.
**
** *pConchPath is set to NULL if a memory allocation error occurs.
*/
static int proxyCreateConchPathname(char *dbPath, char **pConchPath){
  int i;                        /* Loop counter */
  int len = (int)strlen(dbPath); /* Length of database filename - dbPath */
  char *conchPath;              /* buffer in which to construct conch name */

  /* Allocate space for the conch filename and initialize the name to
  ** the name of the original database file. */  
  *pConchPath = conchPath = (char *)sqlite3_malloc64(len + 8);
  if( conchPath==0 ){
    return SQLITE_NOMEM_BKPT;
  }
  memcpy(conchPath, dbPath, len+1);
  
  /* now insert a "." before the last / character */
  for( i=(len-1); i>=0; i-- ){
    if( conchPath[i]=='/' ){
      i++;
      break;
    }
  }
  conchPath[i]='.';
  while ( i<len ){
    conchPath[i+1]=dbPath[i];
    i++;
  }

  /* append the "-conch" suffix to the file */
  memcpy(&conchPath[i+1], "-conch", 7);
  assert( (int)strlen(conchPath) == len+7 );

  return SQLITE_OK;
}


/* Takes a fully configured proxy locking-style unix file and switches
** the local lock file path 
*/
static int switchLockProxyPath(unixFile *pFile, const char *path) {
  proxyLockingContext *pCtx = (proxyLockingContext*)pFile->lockingContext;
  char *oldPath = pCtx->lockProxyPath;
  int rc = SQLITE_OK;

  if( pFile->eFileLock!=NO_LOCK ){
    return SQLITE_BUSY;
  }  

  /* nothing to do if the path is NULL, :auto: or matches the existing path */
  if( !path || path[0]=='\0' || !strcmp(path, ":auto:") ||
    (oldPath && !strncmp(oldPath, path, MAXPATHLEN)) ){
    return SQLITE_OK;
  }else{
    unixFile *lockProxy = pCtx->lockProxy;
    pCtx->lockProxy=NULL;
    pCtx->conchHeld = 0;
    if( lockProxy!=NULL ){
      rc=lockProxy->pMethod->xClose((sqlite3_file *)lockProxy);
      if( rc ) return rc;
      sqlite3_free(lockProxy);
    }
    sqlite3_free(oldPath);
    pCtx->lockProxyPath = sqlite3DbStrDup(0, path);
  }
  
  return rc;
}

/*
** pFile is a file that has been opened by a prior xOpen call.  dbPath
** is a string buffer at least MAXPATHLEN+1 characters in size.
**
** This routine find the filename associated with pFile and writes it
** int dbPath.
*/
static int proxyGetDbPathForUnixFile(unixFile *pFile, char *dbPath){
#if defined(__APPLE__)
  if( pFile->pMethod == &afpIoMethods ){
    /* afp style keeps a reference to the db path in the filePath field 
    ** of the struct */
    assert( (int)strlen((char*)pFile->lockingContext)<=MAXPATHLEN );
    strlcpy(dbPath, ((afpLockingContext *)pFile->lockingContext)->dbPath,
            MAXPATHLEN);
  } else
#endif
  if( pFile->pMethod == &dotlockIoMethods ){
    /* dot lock style uses the locking context to store the dot lock
    ** file path */
    int len = strlen((char *)pFile->lockingContext) - strlen(DOTLOCK_SUFFIX);
    memcpy(dbPath, (char *)pFile->lockingContext, len + 1);
  }else{
    /* all other styles use the locking context to store the db file path */
    assert( strlen((char*)pFile->lockingContext)<=MAXPATHLEN );
    strlcpy(dbPath, (char *)pFile->lockingContext, MAXPATHLEN);
  }
  return SQLITE_OK;
}

/*
** Takes an already filled in unix file and alters it so all file locking 
** will be performed on the local proxy lock file.  The following fields
** are preserved in the locking context so that they can be restored and 
** the unix structure properly cleaned up at close time:
**  ->lockingContext
**  ->pMethod
*/
static int proxyTransformUnixFile(unixFile *pFile, const char *path) {
  proxyLockingContext *pCtx;
  char dbPath[MAXPATHLEN+1];       /* Name of the database file */
  char *lockPath=NULL;
  int rc = SQLITE_OK;
  
  if( pFile->eFileLock!=NO_LOCK ){
    return SQLITE_BUSY;
  }
  proxyGetDbPathForUnixFile(pFile, dbPath);
  if( !path || path[0]=='\0' || !strcmp(path, ":auto:") ){
    lockPath=NULL;
  }else{
    lockPath=(char *)path;
  }
  
  OSTRACE(("TRANSPROXY  %d for %s pid=%d\n", pFile->h,
           (lockPath ? lockPath : ":auto:"), osGetpid(0)));

  pCtx = sqlite3_malloc64( sizeof(*pCtx) );
  if( pCtx==0 ){
    return SQLITE_NOMEM_BKPT;
  }
  memset(pCtx, 0, sizeof(*pCtx));

  rc = proxyCreateConchPathname(dbPath, &pCtx->conchFilePath);
  if( rc==SQLITE_OK ){
    rc = proxyCreateUnixFile(pCtx->conchFilePath, &pCtx->conchFile, 0);
    if( rc==SQLITE_CANTOPEN && ((pFile->openFlags&O_RDWR) == 0) ){
      /* if (a) the open flags are not O_RDWR, (b) the conch isn't there, and
      ** (c) the file system is read-only, then enable no-locking access.
      ** Ugh, since O_RDONLY==0x0000 we test for !O_RDWR since unixOpen asserts
      ** that openFlags will have only one of O_RDONLY or O_RDWR.
      */
      struct statfs fsInfo;
      struct stat conchInfo;
      int goLockless = 0;

      if( osStat(pCtx->conchFilePath, &conchInfo) == -1 ) {
        int err = errno;
        if( (err==ENOENT) && (statfs(dbPath, &fsInfo) != -1) ){
          goLockless = (fsInfo.f_flags&MNT_RDONLY) == MNT_RDONLY;
        }
      }
      if( goLockless ){
        pCtx->conchHeld = -1; /* read only FS/ lockless */
        rc = SQLITE_OK;
      }
    }
  }  
  if( rc==SQLITE_OK && lockPath ){
    pCtx->lockProxyPath = sqlite3DbStrDup(0, lockPath);
  }

  if( rc==SQLITE_OK ){
    pCtx->dbPath = sqlite3DbStrDup(0, dbPath);
    if( pCtx->dbPath==NULL ){
      rc = SQLITE_NOMEM_BKPT;
    }
  }
  if( rc==SQLITE_OK ){
    /* all memory is allocated, proxys are created and assigned, 
    ** switch the locking context and pMethod then return.
    */
    pCtx->oldLockingContext = pFile->lockingContext;
    pFile->lockingContext = pCtx;
    pCtx->pOldMethod = pFile->pMethod;
    pFile->pMethod = &proxyIoMethods;
  }else{
    if( pCtx->conchFile ){ 
      pCtx->conchFile->pMethod->xClose((sqlite3_file *)pCtx->conchFile);
      sqlite3_free(pCtx->conchFile);
    }
    sqlite3DbFree(0, pCtx->lockProxyPath);
    sqlite3_free(pCtx->conchFilePath); 
    sqlite3_free(pCtx);
  }
  OSTRACE(("TRANSPROXY  %d %s\n", pFile->h,
           (rc==SQLITE_OK ? "ok" : "failed")));
  return rc;
}


/*
** This routine handles sqlite3_file_control() calls that are specific
** to proxy locking.
*/
static int proxyFileControl(sqlite3_file *id, int op, void *pArg){
  switch( op ){
    case SQLITE_FCNTL_GET_LOCKPROXYFILE: {
      unixFile *pFile = (unixFile*)id;
      if( pFile->pMethod == &proxyIoMethods ){
        proxyLockingContext *pCtx = (proxyLockingContext*)pFile->lockingContext;
        proxyTakeConch(pFile);
        if( pCtx->lockProxyPath ){
          *(const char **)pArg = pCtx->lockProxyPath;
        }else{
          *(const char **)pArg = ":auto: (not held)";
        }
      } else {
        *(const char **)pArg = NULL;
      }
      return SQLITE_OK;
    }
    case SQLITE_FCNTL_SET_LOCKPROXYFILE: {
      unixFile *pFile = (unixFile*)id;
      int rc = SQLITE_OK;
      int isProxyStyle = (pFile->pMethod == &proxyIoMethods);
      if( pArg==NULL || (const char *)pArg==0 ){
        if( isProxyStyle ){
          /* turn off proxy locking - not supported.  If support is added for
          ** switching proxy locking mode off then it will need to fail if
          ** the journal mode is WAL mode. 
          */
          rc = SQLITE_ERROR /*SQLITE_PROTOCOL? SQLITE_MISUSE?*/;
        }else{
          /* turn off proxy locking - already off - NOOP */
          rc = SQLITE_OK;
        }
      }else{
        const char *proxyPath = (const char *)pArg;
        if( isProxyStyle ){
          proxyLockingContext *pCtx = 
            (proxyLockingContext*)pFile->lockingContext;
          if( !strcmp(pArg, ":auto:") 
           || (pCtx->lockProxyPath &&
               !strncmp(pCtx->lockProxyPath, proxyPath, MAXPATHLEN))
          ){
            rc = SQLITE_OK;
          }else{
            rc = switchLockProxyPath(pFile, proxyPath);
          }
        }else{
          /* turn on proxy file locking */
          rc = proxyTransformUnixFile(pFile, proxyPath);
        }
      }
      return rc;
    }
    default: {
      assert( 0 );  /* The call assures that only valid opcodes are sent */
    }
  }
  /*NOTREACHED*/
  return SQLITE_ERROR;
}

/*
** Within this division (the proxying locking implementation) the procedures
** above this point are all utilities.  The lock-related methods of the
** proxy-locking sqlite3_io_method object follow.
*/


/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, set *pResOut
** to a non-zero value otherwise *pResOut is set to zero.  The return value
** is set to SQLITE_OK unless an I/O error occurs during lock checking.
*/
static int proxyCheckReservedLock(sqlite3_file *id, int *pResOut) {
  unixFile *pFile = (unixFile*)id;
  int rc = proxyTakeConch(pFile);
  if( rc==SQLITE_OK ){
    proxyLockingContext *pCtx = (proxyLockingContext *)pFile->lockingContext;
    if( pCtx->conchHeld>0 ){
      unixFile *proxy = pCtx->lockProxy;
      return proxy->pMethod->xCheckReservedLock((sqlite3_file*)proxy, pResOut);
    }else{ /* conchHeld < 0 is lockless */
      pResOut=0;
    }
  }
  return rc;
}

/*
** Lock the file with the lock specified by parameter eFileLock - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK
**     (3) PENDING_LOCK
**     (4) EXCLUSIVE_LOCK
**
** Sometimes when requesting one lock state, additional lock states
** are inserted in between.  The locking might fail on one of the later
** transitions leaving the lock state different from what it started but
** still short of its goal.  The following chart shows the allowed
** transitions and the inserted intermediate states:
**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
static int proxyLock(sqlite3_file *id, int eFileLock) {
  unixFile *pFile = (unixFile*)id;
  int rc = proxyTakeConch(pFile);
  if( rc==SQLITE_OK ){
    proxyLockingContext *pCtx = (proxyLockingContext *)pFile->lockingContext;
    if( pCtx->conchHeld>0 ){
      unixFile *proxy = pCtx->lockProxy;
      rc = proxy->pMethod->xLock((sqlite3_file*)proxy, eFileLock);
      pFile->eFileLock = proxy->eFileLock;
    }else{
      /* conchHeld < 0 is lockless */
    }
  }
  return rc;
}


/*
** Lower the locking level on file descriptor pFile to eFileLock.  eFileLock
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
*/
static int proxyUnlock(sqlite3_file *id, int eFileLock) {
  unixFile *pFile = (unixFile*)id;
  int rc = proxyTakeConch(pFile);
  if( rc==SQLITE_OK ){
    proxyLockingContext *pCtx = (proxyLockingContext *)pFile->lockingContext;
    if( pCtx->conchHeld>0 ){
      unixFile *proxy = pCtx->lockProxy;
      rc = proxy->pMethod->xUnlock((sqlite3_file*)proxy, eFileLock);
      pFile->eFileLock = proxy->eFileLock;
    }else{
      /* conchHeld < 0 is lockless */
    }
  }
  return rc;
}

/*
** Close a file that uses proxy locks.
*/
static int proxyClose(sqlite3_file *id) {
  if( ALWAYS(id) ){
    unixFile *pFile = (unixFile*)id;
    proxyLockingContext *pCtx = (proxyLockingContext *)pFile->lockingContext;
    unixFile *lockProxy = pCtx->lockProxy;
    unixFile *conchFile = pCtx->conchFile;
    int rc = SQLITE_OK;
    
    if( lockProxy ){
      rc = lockProxy->pMethod->xUnlock((sqlite3_file*)lockProxy, NO_LOCK);
      if( rc ) return rc;
      rc = lockProxy->pMethod->xClose((sqlite3_file*)lockProxy);
      if( rc ) return rc;
      sqlite3_free(lockProxy);
      pCtx->lockProxy = 0;
    }
    if( conchFile ){
      if( pCtx->conchHeld ){
        rc = proxyReleaseConch(pFile);
        if( rc ) return rc;
      }
      rc = conchFile->pMethod->xClose((sqlite3_file*)conchFile);
      if( rc ) return rc;
      sqlite3_free(conchFile);
    }
    sqlite3DbFree(0, pCtx->lockProxyPath);
    sqlite3_free(pCtx->conchFilePath);
    sqlite3DbFree(0, pCtx->dbPath);
    /* restore the original locking context and pMethod then close it */
    pFile->lockingContext = pCtx->oldLockingContext;
    pFile->pMethod = pCtx->pOldMethod;
    sqlite3_free(pCtx);
    return pFile->pMethod->xClose(id);
  }
  return SQLITE_OK;
}



#endif /* defined(__APPLE__) && SQLITE_ENABLE_LOCKING_STYLE */
/*
** The proxy locking style is intended for use with AFP filesystems.
** And since AFP is only supported on MacOSX, the proxy locking is also
** restricted to MacOSX.
** 
**
******************* End of the proxy lock implementation **********************
******************************************************************************/

/*
** Initialize the operating system interface.
**
** This routine registers all VFS implementations for unix-like operating
** systems.  This routine, and the sqlite3_os_end() routine that follows,
** should be the only routines in this file that are visible from other
** files.
**
** This routine is called once during SQLite initialization and by a
** single thread.  The memory allocation and mutex subsystems have not
** necessarily been initialized when this routine is called, and so they
** should not be used.
*/
int sqlite3_os_init(void){ 
  /* 
  ** The following macro defines an initializer for an sqlite3_vfs object.
  ** The name of the VFS is NAME.  The pAppData is a pointer to a pointer
  ** to the "finder" function.  (pAppData is a pointer to a pointer because
  ** silly C90 rules prohibit a void* from being cast to a function pointer
  ** and so we have to go through the intermediate pointer to avoid problems
  ** when compiling with -pedantic-errors on GCC.)
  **
  ** The FINDER parameter to this macro is the name of the pointer to the
  ** finder-function.  The finder-function returns a pointer to the
  ** sqlite_io_methods object that implements the desired locking
  ** behaviors.  See the division above that contains the IOMETHODS
  ** macro for addition information on finder-functions.
  **
  ** Most finders simply return a pointer to a fixed sqlite3_io_methods
  ** object.  But the "autolockIoFinder" available on MacOSX does a little
  ** more than that; it looks at the filesystem type that hosts the 
  ** database file and tries to choose an locking method appropriate for
  ** that filesystem time.
  */
  #define UNIXVFS(VFSNAME, FINDER) {                        \
    3,                    /* iVersion */                    \
    sizeof(unixFile),     /* szOsFile */                    \
    MAX_PATHNAME,         /* mxPathname */                  \
    0,                    /* pNext */                       \
    VFSNAME,              /* zName */                       \
    (void*)&FINDER,       /* pAppData */                    \
    unixOpen,             /* xOpen */                       \
    unixDelete,           /* xDelete */                     \
    unixAccess,           /* xAccess */                     \
    unixFullPathname,     /* xFullPathname */               \
    unixDlOpen,           /* xDlOpen */                     \
    unixDlError,          /* xDlError */                    \
    unixDlSym,            /* xDlSym */                      \
    unixDlClose,          /* xDlClose */                    \
    unixRandomness,       /* xRandomness */                 \
    unixSleep,            /* xSleep */                      \
    unixCurrentTime,      /* xCurrentTime */                \
    unixGetLastError,     /* xGetLastError */               \
    unixCurrentTimeInt64, /* xCurrentTimeInt64 */           \
    unixSetSystemCall,    /* xSetSystemCall */              \
    unixGetSystemCall,    /* xGetSystemCall */              \
    unixNextSystemCall,   /* xNextSystemCall */             \
  }

  /*
  ** All default VFSes for unix are contained in the following array.
  **
  ** Note that the sqlite3_vfs.pNext field of the VFS object is modified
  ** by the SQLite core when the VFS is registered.  So the following
  ** array cannot be const.
  */
  static sqlite3_vfs aVfs[] = {
#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
    UNIXVFS("unix",          autolockIoFinder ),
#elif OS_VXWORKS
    UNIXVFS("unix",          vxworksIoFinder ),
#else
    UNIXVFS("unix",          posixIoFinder ),
#endif
    UNIXVFS("unix-none",     nolockIoFinder ),
    UNIXVFS("unix-dotfile",  dotlockIoFinder ),
    UNIXVFS("unix-excl",     posixIoFinder ),
#if OS_VXWORKS
    UNIXVFS("unix-namedsem", semIoFinder ),
#endif
#if SQLITE_ENABLE_LOCKING_STYLE || OS_VXWORKS
    UNIXVFS("unix-posix",    posixIoFinder ),
#endif
#if SQLITE_ENABLE_LOCKING_STYLE
    UNIXVFS("unix-flock",    flockIoFinder ),
#endif
#if SQLITE_ENABLE_LOCKING_STYLE && defined(__APPLE__)
    UNIXVFS("unix-afp",      afpIoFinder ),
    UNIXVFS("unix-nfs",      nfsIoFinder ),
    UNIXVFS("unix-proxy",    proxyIoFinder ),
#endif
  };
  unsigned int i;          /* Loop counter */

  /* Double-check that the aSyscall[] array has been constructed
  ** correctly.  See ticket [bb3a86e890c8e96ab] */
  assert( ArraySize(aSyscall)==29 );

  /* Register all VFSes defined in the aVfs[] array */
  for(i=0; i<(sizeof(aVfs)/sizeof(sqlite3_vfs)); i++){
    sqlite3_vfs_register(&aVfs[i], i==0);
  }
  unixBigLock = sqlite3MutexAlloc(SQLITE_MUTEX_STATIC_VFS1);
  return SQLITE_OK; 
}

/*
** Shutdown the operating system interface.
**
** Some operating systems might need to do some cleanup in this routine,
** to release dynamically allocated objects.  But not on unix.
** This routine is a no-op for unix.
*/
int sqlite3_os_end(void){ 
  unixBigLock = 0;
  return SQLITE_OK; 
}
 
#endif /* SQLITE_OS_UNIX */