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
|
/**
* Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
* SPDX-License-Identifier: Apache-2.0.
*/
#include "aws/s3/private/s3_auto_ranged_get.h"
#include "aws/s3/private/s3_auto_ranged_put.h"
#include "aws/s3/private/s3_buffer_pool.h"
#include "aws/s3/private/s3_client_impl.h"
#include "aws/s3/private/s3_copy_object.h"
#include "aws/s3/private/s3_default_meta_request.h"
#include "aws/s3/private/s3_meta_request_impl.h"
#include "aws/s3/private/s3_parallel_input_stream.h"
#include "aws/s3/private/s3_request_messages.h"
#include "aws/s3/private/s3_util.h"
#include "aws/s3/private/s3express_credentials_provider_impl.h"
#include "aws/s3/s3express_credentials_provider.h"
#include <aws/auth/credentials.h>
#include <aws/common/assert.h>
#include <aws/common/atomics.h>
#include <aws/common/clock.h>
#include <aws/common/device_random.h>
#include <aws/common/json.h>
#include <aws/common/string.h>
#include <aws/common/system_info.h>
#include <aws/http/connection.h>
#include <aws/http/connection_manager.h>
#include <aws/http/proxy.h>
#include <aws/http/request_response.h>
#include <aws/io/channel_bootstrap.h>
#include <aws/io/event_loop.h>
#include <aws/io/host_resolver.h>
#include <aws/io/retry_strategy.h>
#include <aws/io/socket.h>
#include <aws/io/stream.h>
#include <aws/io/tls_channel_handler.h>
#include <aws/io/uri.h>
#include <inttypes.h>
#include <math.h>
#ifdef _MSC_VER
# pragma warning(disable : 4232) /* function pointer to dll symbol */
#endif /* _MSC_VER */
struct aws_s3_meta_request_work {
struct aws_linked_list_node node;
struct aws_s3_meta_request *meta_request;
};
static const enum aws_log_level s_log_level_client_stats = AWS_LL_INFO;
static const uint32_t s_max_requests_multiplier = 4;
/* TODO Provide analysis on origins of this value. */
static const double s_throughput_per_vip_gbps = 4.0;
/* Preferred amount of active connections per meta request type. */
const uint32_t g_num_conns_per_vip_meta_request_look_up[AWS_S3_META_REQUEST_TYPE_MAX] = {
10, /* AWS_S3_META_REQUEST_TYPE_DEFAULT */
10, /* AWS_S3_META_REQUEST_TYPE_GET_OBJECT */
10, /* AWS_S3_META_REQUEST_TYPE_PUT_OBJECT */
10 /* AWS_S3_META_REQUEST_TYPE_COPY_OBJECT */
};
/* Should be max of s_num_conns_per_vip_meta_request_look_up */
const uint32_t g_max_num_connections_per_vip = 10;
/**
* Default part size is 8 MiB to reach the best performance from the experiments we had.
* Default max part size is 5GiB as the server limit. Object size limit is 5TiB for now.
* max number of upload parts is 10000.
* TODO Provide more information on other values.
*/
static const size_t s_default_part_size = 8 * 1024 * 1024;
static const uint64_t s_default_max_part_size = 5368709120ULL;
static const double s_default_throughput_target_gbps = 10.0;
static const uint32_t s_default_max_retries = 5;
static size_t s_dns_host_address_ttl_seconds = 5 * 60;
/* Default time until a connection is declared dead, while handling a request but seeing no activity.
* 30 seconds mirrors the value currently used by the Java SDK. */
static const uint32_t s_default_throughput_failure_interval_seconds = 30;
/* Amount of time spent idling before trimming buffer. */
static const size_t s_buffer_pool_trim_time_offset_in_s = 5;
/* Called when ref count is 0. */
static void s_s3_client_start_destroy(void *user_data);
/* Called by s_s3_client_process_work_default when all shutdown criteria has been met. */
static void s_s3_client_finish_destroy_default(struct aws_s3_client *client);
/* Called when the body streaming elg shutdown has completed. */
static void s_s3_client_body_streaming_elg_shutdown(void *user_data);
static void s_s3_client_create_connection_for_request(struct aws_s3_client *client, struct aws_s3_request *request);
/* Callback which handles the HTTP connection retrieved by acquire_http_connection. */
static void s_s3_client_on_acquire_http_connection(
struct aws_http_connection *http_connection,
int error_code,
void *user_data);
static void s_s3_client_push_meta_request_synced(
struct aws_s3_client *client,
struct aws_s3_meta_request *meta_request);
/* Schedule task for processing work. (Calls the corresponding vtable function.) */
static void s_s3_client_schedule_process_work_synced(struct aws_s3_client *client);
/* Default implementation for scheduling processing of work. */
static void s_s3_client_schedule_process_work_synced_default(struct aws_s3_client *client);
/* Actual task function that processes work. */
static void s_s3_client_process_work_task(struct aws_task *task, void *arg, enum aws_task_status task_status);
static void s_s3_client_process_work_default(struct aws_s3_client *client);
static void s_s3_client_endpoint_shutdown_callback(struct aws_s3_client *client);
/* Default factory function for creating a meta request. */
static struct aws_s3_meta_request *s_s3_client_meta_request_factory_default(
struct aws_s3_client *client,
const struct aws_s3_meta_request_options *options);
static struct aws_s3_client_vtable s_s3_client_default_vtable = {
.meta_request_factory = s_s3_client_meta_request_factory_default,
.acquire_http_connection = aws_http_connection_manager_acquire_connection,
.get_host_address_count = aws_host_resolver_get_host_address_count,
.schedule_process_work_synced = s_s3_client_schedule_process_work_synced_default,
.process_work = s_s3_client_process_work_default,
.endpoint_shutdown_callback = s_s3_client_endpoint_shutdown_callback,
.finish_destroy = s_s3_client_finish_destroy_default,
.parallel_input_stream_new_from_file = aws_parallel_input_stream_new_from_file,
};
void aws_s3_set_dns_ttl(size_t ttl) {
s_dns_host_address_ttl_seconds = ttl;
}
/* Returns the max number of connections allowed.
*
* When meta request is NULL, this will return the overall allowed number of connections.
*
* If meta_request is not NULL, this will give the max number of connections allowed for that meta request type on
* that endpoint.
*/
uint32_t aws_s3_client_get_max_active_connections(
struct aws_s3_client *client,
struct aws_s3_meta_request *meta_request) {
AWS_PRECONDITION(client);
uint32_t num_connections_per_vip = g_max_num_connections_per_vip;
uint32_t num_vips = client->ideal_vip_count;
if (meta_request != NULL) {
num_connections_per_vip = g_num_conns_per_vip_meta_request_look_up[meta_request->type];
struct aws_s3_endpoint *endpoint = meta_request->endpoint;
AWS_ASSERT(endpoint != NULL);
AWS_ASSERT(client->vtable->get_host_address_count);
size_t num_known_vips = client->vtable->get_host_address_count(
client->client_bootstrap->host_resolver, endpoint->host_name, AWS_GET_HOST_ADDRESS_COUNT_RECORD_TYPE_A);
/* If the number of known vips is less than our ideal VIP count, clamp it. */
if (num_known_vips < (size_t)num_vips) {
num_vips = (uint32_t)num_known_vips;
}
}
/* We always want to allow for at least one VIP worth of connections. */
if (num_vips == 0) {
num_vips = 1;
}
uint32_t max_active_connections = num_vips * num_connections_per_vip;
if (client->max_active_connections_override > 0 &&
client->max_active_connections_override < max_active_connections) {
max_active_connections = client->max_active_connections_override;
}
return max_active_connections;
}
/* Returns the max number of requests allowed to be in memory */
uint32_t aws_s3_client_get_max_requests_in_flight(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
return aws_s3_client_get_max_active_connections(client, NULL) * s_max_requests_multiplier;
}
/* Returns the max number of requests that should be in preparation stage (ie: reading from a stream, being signed,
* etc.) */
uint32_t aws_s3_client_get_max_requests_prepare(struct aws_s3_client *client) {
return aws_s3_client_get_max_active_connections(client, NULL);
}
static uint32_t s_s3_client_get_num_requests_network_io(
struct aws_s3_client *client,
enum aws_s3_meta_request_type meta_request_type) {
AWS_PRECONDITION(client);
uint32_t num_requests_network_io = 0;
if (meta_request_type == AWS_S3_META_REQUEST_TYPE_MAX) {
for (uint32_t i = 0; i < AWS_S3_META_REQUEST_TYPE_MAX; ++i) {
num_requests_network_io += (uint32_t)aws_atomic_load_int(&client->stats.num_requests_network_io[i]);
}
} else {
num_requests_network_io =
(uint32_t)aws_atomic_load_int(&client->stats.num_requests_network_io[meta_request_type]);
}
return num_requests_network_io;
}
void aws_s3_client_lock_synced_data(struct aws_s3_client *client) {
aws_mutex_lock(&client->synced_data.lock);
}
void aws_s3_client_unlock_synced_data(struct aws_s3_client *client) {
aws_mutex_unlock(&client->synced_data.lock);
}
static void s_s3express_provider_finish_destroy(void *user_data) {
struct aws_s3_client *client = user_data;
AWS_PRECONDITION(client);
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
client->synced_data.s3express_provider_active = false;
/* Schedule the work task to call s_s3_client_finish_destroy function if
* everything cleaning up asynchronously has finished. */
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
}
struct aws_s3express_credentials_provider *s_s3express_provider_default_factory(
struct aws_allocator *allocator,
struct aws_s3_client *client,
aws_simple_completion_callback on_provider_shutdown_callback,
void *shutdown_user_data,
void *factory_user_data) {
(void)factory_user_data;
struct aws_s3express_credentials_provider_default_options options = {
.client = client,
.shutdown_complete_callback = on_provider_shutdown_callback,
.shutdown_user_data = shutdown_user_data,
};
struct aws_s3express_credentials_provider *s3express_provider =
aws_s3express_credentials_provider_new_default(allocator, &options);
return s3express_provider;
}
struct aws_s3_client *aws_s3_client_new(
struct aws_allocator *allocator,
const struct aws_s3_client_config *client_config) {
AWS_PRECONDITION(allocator);
AWS_PRECONDITION(client_config);
if (client_config->client_bootstrap == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"Cannot create client from client_config; client_bootstrap provided in options is invalid.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
/* Cannot be less than zero. If zero, use default. */
if (client_config->throughput_target_gbps < 0.0) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"Cannot create client from client_config; throughput_target_gbps cannot less than or equal to 0.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (client_config->signing_config == NULL) {
AWS_LOGF_ERROR(AWS_LS_S3_CLIENT, "Cannot create client from client_config; signing_config is required.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (client_config->signing_config->credentials == NULL &&
client_config->signing_config->credentials_provider == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"Cannot create client from client_config; Invalid signing_config provided, either credentials or "
"credentials provider has to be set.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (!client_config->enable_s3express &&
client_config->signing_config->algorithm == AWS_SIGNING_ALGORITHM_V4_S3EXPRESS) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"Cannot create client from client_config; Client config is set use S3 Express signing, but S3 Express "
"support is "
"not configured.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
#ifdef BYO_CRYPTO
if (client_config->tls_mode == AWS_MR_TLS_ENABLED && client_config->tls_connection_options == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"Cannot create client from client_config; when using BYO_CRYPTO, tls_connection_options can not be "
"NULL when TLS is enabled.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
#endif
struct aws_s3_client *client = aws_mem_calloc(allocator, 1, sizeof(struct aws_s3_client));
client->allocator = allocator;
size_t mem_limit = 0;
if (client_config->memory_limit_in_bytes == 0) {
#if SIZE_BITS == 32
if (client_config->throughput_target_gbps > 25.0) {
mem_limit = GB_TO_BYTES(2);
} else {
mem_limit = GB_TO_BYTES(1);
}
#else
if (client_config->throughput_target_gbps > 75.0) {
mem_limit = GB_TO_BYTES(8);
} else if (client_config->throughput_target_gbps > 25.0) {
mem_limit = GB_TO_BYTES(4);
} else {
mem_limit = GB_TO_BYTES(2);
}
#endif
} else {
// cap memory limit to SIZE_MAX
if (client_config->memory_limit_in_bytes > SIZE_MAX) {
mem_limit = SIZE_MAX;
} else {
mem_limit = (size_t)client_config->memory_limit_in_bytes;
}
}
size_t part_size = s_default_part_size;
if (client_config->part_size != 0) {
if (client_config->part_size > SIZE_MAX) {
part_size = SIZE_MAX;
} else {
part_size = (size_t)client_config->part_size;
}
}
client->buffer_pool = aws_s3_buffer_pool_new(allocator, part_size, mem_limit);
if (client->buffer_pool == NULL) {
goto on_early_fail;
}
struct aws_s3_buffer_pool_usage_stats pool_usage = aws_s3_buffer_pool_get_usage(client->buffer_pool);
if (client_config->max_part_size > pool_usage.mem_limit) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"Cannot create client from client_config; configured max part size should not exceed memory limit."
"size.");
aws_raise_error(AWS_ERROR_S3_INVALID_MEMORY_LIMIT_CONFIG);
goto on_early_fail;
}
client->vtable = &s_s3_client_default_vtable;
aws_ref_count_init(&client->ref_count, client, (aws_simple_completion_callback *)s_s3_client_start_destroy);
if (aws_mutex_init(&client->synced_data.lock) != AWS_OP_SUCCESS) {
goto on_early_fail;
}
aws_linked_list_init(&client->synced_data.pending_meta_request_work);
aws_linked_list_init(&client->synced_data.prepared_requests);
aws_linked_list_init(&client->threaded_data.meta_requests);
aws_linked_list_init(&client->threaded_data.request_queue);
aws_atomic_init_int(&client->stats.num_requests_in_flight, 0);
for (uint32_t i = 0; i < (uint32_t)AWS_S3_META_REQUEST_TYPE_MAX; ++i) {
aws_atomic_init_int(&client->stats.num_requests_network_io[i], 0);
}
aws_atomic_init_int(&client->stats.num_requests_stream_queued_waiting, 0);
aws_atomic_init_int(&client->stats.num_requests_streaming_response, 0);
*((uint32_t *)&client->max_active_connections_override) = client_config->max_active_connections_override;
/* Store our client bootstrap. */
client->client_bootstrap = aws_client_bootstrap_acquire(client_config->client_bootstrap);
struct aws_event_loop_group *event_loop_group = client_config->client_bootstrap->event_loop_group;
aws_event_loop_group_acquire(event_loop_group);
client->process_work_event_loop = aws_event_loop_group_get_next_loop(event_loop_group);
/* Make a copy of the region string. */
client->region = aws_string_new_from_array(allocator, client_config->region.ptr, client_config->region.len);
*((size_t *)&client->part_size) = part_size;
if (client_config->max_part_size != 0) {
*((uint64_t *)&client->max_part_size) = client_config->max_part_size;
} else {
*((uint64_t *)&client->max_part_size) = s_default_max_part_size;
}
if (client_config->max_part_size > pool_usage.mem_limit) {
*((uint64_t *)&client->max_part_size) = pool_usage.mem_limit;
}
if (client->max_part_size > SIZE_MAX) {
/* For the 32bit max part size to be SIZE_MAX */
*((uint64_t *)&client->max_part_size) = SIZE_MAX;
}
if (client_config->multipart_upload_threshold != 0) {
*((uint64_t *)&client->multipart_upload_threshold) = client_config->multipart_upload_threshold;
} else {
*((uint64_t *)&client->multipart_upload_threshold) =
part_size > g_s3_min_upload_part_size ? part_size : g_s3_min_upload_part_size;
}
if (client_config->max_part_size < client_config->part_size) {
*((uint64_t *)&client_config->max_part_size) = client_config->part_size;
}
client->connect_timeout_ms = client_config->connect_timeout_ms;
if (client_config->proxy_ev_settings) {
client->proxy_ev_settings = aws_mem_calloc(allocator, 1, sizeof(struct proxy_env_var_settings));
*client->proxy_ev_settings = *client_config->proxy_ev_settings;
if (client_config->proxy_ev_settings->tls_options) {
client->proxy_ev_tls_options = aws_mem_calloc(allocator, 1, sizeof(struct aws_tls_connection_options));
if (aws_tls_connection_options_copy(client->proxy_ev_tls_options, client->proxy_ev_settings->tls_options)) {
goto on_error;
}
client->proxy_ev_settings->tls_options = client->proxy_ev_tls_options;
}
}
if (client_config->tcp_keep_alive_options) {
client->tcp_keep_alive_options = aws_mem_calloc(allocator, 1, sizeof(struct aws_s3_tcp_keep_alive_options));
*client->tcp_keep_alive_options = *client_config->tcp_keep_alive_options;
}
if (client_config->monitoring_options) {
client->monitoring_options = *client_config->monitoring_options;
} else {
client->monitoring_options.minimum_throughput_bytes_per_second = 1;
client->monitoring_options.allowable_throughput_failure_interval_seconds =
s_default_throughput_failure_interval_seconds;
}
if (client_config->tls_mode == AWS_MR_TLS_ENABLED) {
client->tls_connection_options =
aws_mem_calloc(client->allocator, 1, sizeof(struct aws_tls_connection_options));
if (client_config->tls_connection_options != NULL) {
aws_tls_connection_options_copy(client->tls_connection_options, client_config->tls_connection_options);
} else {
#ifdef BYO_CRYPTO
AWS_FATAL_ASSERT(false);
goto on_error;
#else
struct aws_tls_ctx_options default_tls_ctx_options;
AWS_ZERO_STRUCT(default_tls_ctx_options);
aws_tls_ctx_options_init_default_client(&default_tls_ctx_options, allocator);
struct aws_tls_ctx *default_tls_ctx = aws_tls_client_ctx_new(allocator, &default_tls_ctx_options);
if (default_tls_ctx == NULL) {
goto on_error;
}
aws_tls_connection_options_init_from_ctx(client->tls_connection_options, default_tls_ctx);
aws_tls_ctx_release(default_tls_ctx);
aws_tls_ctx_options_clean_up(&default_tls_ctx_options);
#endif
}
}
if (client_config->proxy_options) {
client->proxy_config = aws_http_proxy_config_new_from_proxy_options_with_tls_info(
allocator, client_config->proxy_options, client_config->tls_mode == AWS_MR_TLS_ENABLED);
if (client->proxy_config == NULL) {
goto on_error;
}
}
/* Set up body streaming ELG */
{
uint16_t num_event_loops =
(uint16_t)aws_array_list_length(&client->client_bootstrap->event_loop_group->event_loops);
uint16_t num_streaming_threads = num_event_loops;
if (num_streaming_threads < 1) {
num_streaming_threads = 1;
}
struct aws_shutdown_callback_options body_streaming_elg_shutdown_options = {
.shutdown_callback_fn = s_s3_client_body_streaming_elg_shutdown,
.shutdown_callback_user_data = client,
};
client->body_streaming_elg = aws_event_loop_group_new_default(
client->allocator, num_streaming_threads, &body_streaming_elg_shutdown_options);
if (!client->body_streaming_elg) {
/* Fail to create elg, we should fail the call */
goto on_error;
}
client->synced_data.body_streaming_elg_allocated = true;
}
/* Setup cannot fail after this point. */
if (client_config->throughput_target_gbps != 0.0) {
*((double *)&client->throughput_target_gbps) = client_config->throughput_target_gbps;
} else {
*((double *)&client->throughput_target_gbps) = s_default_throughput_target_gbps;
}
*((enum aws_s3_meta_request_compute_content_md5 *)&client->compute_content_md5) =
client_config->compute_content_md5;
/* Determine how many vips are ideal by dividing target-throughput by throughput-per-vip. */
{
double ideal_vip_count_double = client->throughput_target_gbps / s_throughput_per_vip_gbps;
*((uint32_t *)&client->ideal_vip_count) = (uint32_t)ceil(ideal_vip_count_double);
}
client->cached_signing_config = aws_cached_signing_config_new(client, client_config->signing_config);
if (client_config->enable_s3express) {
if (client_config->s3express_provider_override_factory) {
client->s3express_provider_factory = client_config->s3express_provider_override_factory;
client->factory_user_data = client_config->factory_user_data;
} else {
client->s3express_provider_factory = s_s3express_provider_default_factory;
}
}
client->synced_data.active = true;
if (client_config->retry_strategy != NULL) {
aws_retry_strategy_acquire(client_config->retry_strategy);
client->retry_strategy = client_config->retry_strategy;
} else {
struct aws_exponential_backoff_retry_options backoff_retry_options = {
.el_group = client_config->client_bootstrap->event_loop_group,
.max_retries = s_default_max_retries,
};
struct aws_standard_retry_options retry_options = {
.backoff_retry_options = backoff_retry_options,
};
client->retry_strategy = aws_retry_strategy_new_standard(allocator, &retry_options);
}
aws_hash_table_init(
&client->synced_data.endpoints,
client->allocator,
10,
aws_hash_string,
aws_hash_callback_string_eq,
aws_hash_callback_string_destroy,
NULL);
/* Initialize shutdown options and tracking. */
client->shutdown_callback = client_config->shutdown_callback;
client->shutdown_callback_user_data = client_config->shutdown_callback_user_data;
*((bool *)&client->enable_read_backpressure) = client_config->enable_read_backpressure;
*((size_t *)&client->initial_read_window) = client_config->initial_read_window;
return client;
on_error:
aws_string_destroy(client->region);
if (client->tls_connection_options) {
aws_tls_connection_options_clean_up(client->tls_connection_options);
aws_mem_release(client->allocator, client->tls_connection_options);
client->tls_connection_options = NULL;
}
if (client->proxy_config) {
aws_http_proxy_config_destroy(client->proxy_config);
}
if (client->proxy_ev_tls_options) {
aws_tls_connection_options_clean_up(client->proxy_ev_tls_options);
aws_mem_release(client->allocator, client->proxy_ev_tls_options);
client->proxy_ev_settings->tls_options = NULL;
}
aws_mem_release(client->allocator, client->proxy_ev_settings);
aws_mem_release(client->allocator, client->tcp_keep_alive_options);
aws_event_loop_group_release(client->client_bootstrap->event_loop_group);
aws_client_bootstrap_release(client->client_bootstrap);
aws_mutex_clean_up(&client->synced_data.lock);
on_early_fail:
aws_mem_release(client->allocator, client);
return NULL;
}
struct aws_s3_client *aws_s3_client_acquire(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
aws_ref_count_acquire(&client->ref_count);
return client;
}
struct aws_s3_client *aws_s3_client_release(struct aws_s3_client *client) {
if (client != NULL) {
aws_ref_count_release(&client->ref_count);
}
return NULL;
}
static void s_s3_client_start_destroy(void *user_data) {
struct aws_s3_client *client = user_data;
AWS_PRECONDITION(client);
AWS_LOGF_DEBUG(AWS_LS_S3_CLIENT, "id=%p Client starting destruction.", (void *)client);
struct aws_linked_list local_vip_list;
aws_linked_list_init(&local_vip_list);
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
client->synced_data.active = false;
/* Prevent the client from cleaning up in between the mutex unlock/re-lock below.*/
client->synced_data.start_destroy_executing = true;
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
aws_event_loop_group_release(client->body_streaming_elg);
client->body_streaming_elg = NULL;
aws_s3express_credentials_provider_release(client->s3express_provider);
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
client->synced_data.start_destroy_executing = false;
/* Schedule the work task to clean up outstanding connections and to call s_s3_client_finish_destroy function if
* everything cleaning up asynchronously has finished. */
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
}
static void s_s3_client_finish_destroy_default(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
AWS_LOGF_DEBUG(AWS_LS_S3_CLIENT, "id=%p Client finishing destruction.", (void *)client);
if (client->threaded_data.trim_buffer_pool_task_scheduled) {
aws_event_loop_cancel_task(client->process_work_event_loop, &client->synced_data.trim_buffer_pool_task);
}
aws_string_destroy(client->region);
client->region = NULL;
if (client->tls_connection_options) {
aws_tls_connection_options_clean_up(client->tls_connection_options);
aws_mem_release(client->allocator, client->tls_connection_options);
client->tls_connection_options = NULL;
}
if (client->proxy_config) {
aws_http_proxy_config_destroy(client->proxy_config);
}
if (client->proxy_ev_tls_options) {
aws_tls_connection_options_clean_up(client->proxy_ev_tls_options);
aws_mem_release(client->allocator, client->proxy_ev_tls_options);
client->proxy_ev_settings->tls_options = NULL;
}
aws_mem_release(client->allocator, client->proxy_ev_settings);
aws_mem_release(client->allocator, client->tcp_keep_alive_options);
aws_mutex_clean_up(&client->synced_data.lock);
AWS_ASSERT(aws_linked_list_empty(&client->synced_data.pending_meta_request_work));
AWS_ASSERT(aws_linked_list_empty(&client->threaded_data.meta_requests));
aws_hash_table_clean_up(&client->synced_data.endpoints);
aws_retry_strategy_release(client->retry_strategy);
aws_event_loop_group_release(client->client_bootstrap->event_loop_group);
aws_client_bootstrap_release(client->client_bootstrap);
aws_cached_signing_config_destroy(client->cached_signing_config);
aws_s3_client_shutdown_complete_callback_fn *shutdown_callback = client->shutdown_callback;
void *shutdown_user_data = client->shutdown_callback_user_data;
aws_s3_buffer_pool_destroy(client->buffer_pool);
aws_mem_release(client->allocator, client);
client = NULL;
if (shutdown_callback != NULL) {
shutdown_callback(shutdown_user_data);
}
}
static void s_s3_client_body_streaming_elg_shutdown(void *user_data) {
struct aws_s3_client *client = user_data;
AWS_PRECONDITION(client);
AWS_LOGF_DEBUG(AWS_LS_S3_CLIENT, "id=%p Client body streaming ELG shutdown.", (void *)client);
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
client->synced_data.body_streaming_elg_allocated = false;
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
}
uint32_t aws_s3_client_queue_requests_threaded(
struct aws_s3_client *client,
struct aws_linked_list *request_list,
bool queue_front) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(request_list);
if (aws_linked_list_empty(request_list)) {
return 0;
}
uint32_t request_list_size = 0;
for (struct aws_linked_list_node *node = aws_linked_list_begin(request_list);
node != aws_linked_list_end(request_list);
node = aws_linked_list_next(node)) {
++request_list_size;
}
if (queue_front) {
aws_linked_list_move_all_front(&client->threaded_data.request_queue, request_list);
} else {
aws_linked_list_move_all_back(&client->threaded_data.request_queue, request_list);
}
client->threaded_data.request_queue_size += request_list_size;
return request_list_size;
}
struct aws_s3_request *aws_s3_client_dequeue_request_threaded(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
if (aws_linked_list_empty(&client->threaded_data.request_queue)) {
return NULL;
}
struct aws_linked_list_node *request_node = aws_linked_list_pop_front(&client->threaded_data.request_queue);
struct aws_s3_request *request = AWS_CONTAINER_OF(request_node, struct aws_s3_request, node);
--client->threaded_data.request_queue_size;
return request;
}
/*
* There is currently some overlap between user provided Host header and endpoint
* override. This function handles the corner cases for when either or both are provided.
*/
int s_apply_endpoint_override(
const struct aws_s3_client *client,
struct aws_http_headers *message_headers,
const struct aws_uri *endpoint) {
AWS_PRECONDITION(message_headers);
const struct aws_byte_cursor *endpoint_authority = endpoint == NULL ? NULL : aws_uri_authority(endpoint);
if (!aws_http_headers_has(message_headers, g_host_header_name)) {
if (endpoint_authority == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; message provided in options does not have either 'Host' header "
"set or endpoint override.",
(void *)client);
return aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
}
if (aws_http_headers_set(message_headers, g_host_header_name, *endpoint_authority)) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; failed to set 'Host' header based on endpoint override.",
(void *)client);
return aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
}
}
struct aws_byte_cursor host_value;
AWS_FATAL_ASSERT(aws_http_headers_get(message_headers, g_host_header_name, &host_value) == AWS_OP_SUCCESS);
if (endpoint_authority != NULL && !aws_byte_cursor_eq(&host_value, endpoint_authority)) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; host header value " PRInSTR
" does not match endpoint override " PRInSTR,
(void *)client,
AWS_BYTE_CURSOR_PRI(host_value),
AWS_BYTE_CURSOR_PRI(*endpoint_authority));
return aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
}
return AWS_OP_SUCCESS;
}
/* Public facing make-meta-request function. */
struct aws_s3_meta_request *aws_s3_client_make_meta_request(
struct aws_s3_client *client,
const struct aws_s3_meta_request_options *options) {
AWS_LOGF_INFO(AWS_LS_S3_CLIENT, "id=%p Initiating making of meta request", (void *)client);
AWS_PRECONDITION(client);
AWS_PRECONDITION(client->vtable);
AWS_PRECONDITION(client->vtable->meta_request_factory);
AWS_PRECONDITION(options);
bool use_s3express_signing = false;
if (options->signing_config != NULL) {
use_s3express_signing = options->signing_config->algorithm == AWS_SIGNING_ALGORITHM_V4_S3EXPRESS;
} else if (client->cached_signing_config) {
use_s3express_signing = client->cached_signing_config->config.algorithm == AWS_SIGNING_ALGORITHM_V4_S3EXPRESS;
}
if (options->type >= AWS_S3_META_REQUEST_TYPE_MAX) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; invalid meta request type specified.",
(void *)client);
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (options->message == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; message provided in options is invalid.",
(void *)client);
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (use_s3express_signing && client->s3express_provider_factory == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; client doesn't support S3 Express signing.",
(void *)client);
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
struct aws_http_headers *message_headers = aws_http_message_get_headers(options->message);
if (message_headers == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; message provided in options does not contain headers.",
(void *)client);
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (options->checksum_config) {
if (options->checksum_config->location == AWS_SCL_TRAILER) {
struct aws_http_headers *headers = aws_http_message_get_headers(options->message);
struct aws_byte_cursor existing_encoding;
AWS_ZERO_STRUCT(existing_encoding);
if (aws_http_headers_get(headers, g_content_encoding_header_name, &existing_encoding) == AWS_OP_SUCCESS) {
if (aws_byte_cursor_find_exact(&existing_encoding, &g_content_encoding_header_aws_chunked, NULL) ==
AWS_OP_SUCCESS) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; for trailer checksum, the original request cannot be "
"aws-chunked encoding. The client will encode the request instead.",
(void *)client);
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
}
}
if (options->checksum_config->location == AWS_SCL_HEADER) {
/* TODO: support calculate checksum to add to header */
aws_raise_error(AWS_ERROR_UNSUPPORTED_OPERATION);
return NULL;
}
if (options->checksum_config->location != AWS_SCL_NONE &&
options->checksum_config->checksum_algorithm == AWS_SCA_NONE) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; checksum location is set, but no checksum algorithm selected.",
(void *)client);
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (options->checksum_config->checksum_algorithm != AWS_SCA_NONE &&
options->checksum_config->location == AWS_SCL_NONE) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; checksum algorithm is set, but no checksum location selected.",
(void *)client);
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
}
if (s_apply_endpoint_override(client, message_headers, options->endpoint)) {
return NULL;
}
struct aws_byte_cursor host_header_value;
/* The Host header must be set from s_apply_endpoint_override, if not errored out */
AWS_FATAL_ASSERT(aws_http_headers_get(message_headers, g_host_header_name, &host_header_value) == AWS_OP_SUCCESS);
bool is_https = true;
uint32_t port = 0;
if (options->endpoint != NULL) {
struct aws_byte_cursor https_scheme = aws_byte_cursor_from_c_str("https");
struct aws_byte_cursor http_scheme = aws_byte_cursor_from_c_str("http");
const struct aws_byte_cursor *scheme = aws_uri_scheme(options->endpoint);
is_https = aws_byte_cursor_eq_ignore_case(scheme, &https_scheme);
if (!is_https && !aws_byte_cursor_eq_ignore_case(scheme, &http_scheme)) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Cannot create meta s3 request; unexpected scheme '" PRInSTR "' in endpoint override.",
(void *)client,
AWS_BYTE_CURSOR_PRI(*scheme));
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
port = aws_uri_port(options->endpoint);
}
struct aws_s3_meta_request *meta_request = client->vtable->meta_request_factory(client, options);
if (meta_request == NULL) {
AWS_LOGF_ERROR(AWS_LS_S3_CLIENT, "id=%p: Could not create new meta request.", (void *)client);
return NULL;
}
bool error_occurred = false;
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
if (use_s3express_signing && !client->synced_data.s3express_provider_active) {
AWS_LOGF_TRACE(AWS_LS_S3_CLIENT, "id=%p Create S3 Express provider for the client.", (void *)client);
/**
* Invoke the factory within the lock. We WARNED people uses their own factory to not use ANY client related
* api during the factory.
*
* We cannot just release the lock and invoke the factory, because it can lead to the other request assume
* the provider is active, and not waiting for the provider to be created. And lead to unexpected behavior.
*/
client->s3express_provider = client->s3express_provider_factory(
client->allocator, client, s_s3express_provider_finish_destroy, client, client->factory_user_data);
/* Provider is related to client, we don't need to clean it up if meta request failed. But, if provider
* failed to be created, let's bail out earlier. */
if (!client->s3express_provider) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Failed to create S3 Express provider for client due to error %d (%s)",
(void *)client,
aws_last_error_or_unknown(),
aws_error_str(aws_last_error_or_unknown()));
error_occurred = true;
goto unlock;
}
client->synced_data.s3express_provider_active = true;
}
struct aws_string *endpoint_host_name = NULL;
if (options->endpoint != NULL) {
endpoint_host_name = aws_string_new_from_cursor(client->allocator, aws_uri_host_name(options->endpoint));
} else {
struct aws_uri host_uri;
if (aws_uri_init_parse(&host_uri, client->allocator, &host_header_value)) {
error_occurred = true;
goto unlock;
}
endpoint_host_name = aws_string_new_from_cursor(client->allocator, aws_uri_host_name(&host_uri));
aws_uri_clean_up(&host_uri);
}
struct aws_s3_endpoint *endpoint = NULL;
struct aws_hash_element *endpoint_hash_element = NULL;
if (use_s3express_signing) {
meta_request->s3express_session_host = aws_string_new_from_string(client->allocator, endpoint_host_name);
}
int was_created = 0;
if (aws_hash_table_create(
&client->synced_data.endpoints, endpoint_host_name, &endpoint_hash_element, &was_created)) {
aws_string_destroy(endpoint_host_name);
error_occurred = true;
goto unlock;
}
if (was_created) {
struct aws_s3_endpoint_options endpoint_options = {
.host_name = endpoint_host_name,
.client_bootstrap = client->client_bootstrap,
.tls_connection_options = is_https ? client->tls_connection_options : NULL,
.dns_host_address_ttl_seconds = s_dns_host_address_ttl_seconds,
.client = client,
.max_connections = aws_s3_client_get_max_active_connections(client, NULL),
.port = port,
.proxy_config = client->proxy_config,
.proxy_ev_settings = client->proxy_ev_settings,
.connect_timeout_ms = client->connect_timeout_ms,
.tcp_keep_alive_options = client->tcp_keep_alive_options,
.monitoring_options = &client->monitoring_options,
};
endpoint = aws_s3_endpoint_new(client->allocator, &endpoint_options);
if (endpoint == NULL) {
aws_hash_table_remove(&client->synced_data.endpoints, endpoint_host_name, NULL, NULL);
aws_string_destroy(endpoint_host_name);
error_occurred = true;
goto unlock;
}
endpoint_hash_element->value = endpoint;
++client->synced_data.num_endpoints_allocated;
} else {
endpoint = endpoint_hash_element->value;
aws_s3_endpoint_acquire(endpoint, true /*already_holding_lock*/);
aws_string_destroy(endpoint_host_name);
endpoint_host_name = NULL;
}
meta_request->endpoint = endpoint;
s_s3_client_push_meta_request_synced(client, meta_request);
s_s3_client_schedule_process_work_synced(client);
unlock:
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
if (error_occurred) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Could not create meta request due to error %d (%s)",
(void *)client,
aws_last_error(),
aws_error_str(aws_last_error()));
meta_request = aws_s3_meta_request_release(meta_request);
} else {
AWS_LOGF_INFO(AWS_LS_S3_CLIENT, "id=%p: Created meta request %p", (void *)client, (void *)meta_request);
}
return meta_request;
}
static void s_s3_client_endpoint_shutdown_callback(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
--client->synced_data.num_endpoints_allocated;
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
}
static struct aws_s3_meta_request *s_s3_client_meta_request_factory_default(
struct aws_s3_client *client,
const struct aws_s3_meta_request_options *options) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(options);
const struct aws_http_headers *initial_message_headers = aws_http_message_get_headers(options->message);
AWS_ASSERT(initial_message_headers);
uint64_t content_length = 0;
struct aws_byte_cursor content_length_cursor;
bool content_length_found = false;
if (!aws_http_headers_get(initial_message_headers, g_content_length_header_name, &content_length_cursor)) {
if (aws_byte_cursor_utf8_parse_u64(content_length_cursor, &content_length)) {
AWS_LOGF_ERROR(
AWS_LS_S3_META_REQUEST,
"Could not parse Content-Length header. header value is:" PRInSTR "",
AWS_BYTE_CURSOR_PRI(content_length_cursor));
aws_raise_error(AWS_ERROR_S3_INVALID_CONTENT_LENGTH_HEADER);
return NULL;
}
content_length_found = true;
}
/* There are multiple ways to pass the body in, ensure only 1 was used */
int body_source_count = 0;
if (aws_http_message_get_body_stream(options->message) != NULL) {
++body_source_count;
}
if (options->send_filepath.len > 0) {
++body_source_count;
}
if (options->send_async_stream != NULL) {
++body_source_count;
}
if (body_source_count > 1) {
AWS_LOGF_ERROR(
AWS_LS_S3_META_REQUEST,
"Could not create auto-ranged-put meta request."
" More than one data source is set (filepath, async stream, body stream).");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
size_t part_size = client->part_size;
if (options->part_size != 0) {
if (options->part_size > SIZE_MAX) {
part_size = SIZE_MAX;
} else {
part_size = (size_t)options->part_size;
}
}
/* Call the appropriate meta-request new function. */
switch (options->type) {
case AWS_S3_META_REQUEST_TYPE_GET_OBJECT: {
struct aws_byte_cursor path_and_query;
if (aws_http_message_get_request_path(options->message, &path_and_query) == AWS_OP_SUCCESS) {
/* If the initial request already has partNumber, the request is not
* splittable(?). Treat it as a Default request.
* TODO: Still need tests to verify that the request of a part is
* splittable or not */
struct aws_byte_cursor sub_string;
AWS_ZERO_STRUCT(sub_string);
/* The first split on '?' for path and query is path, the second is query */
if (aws_byte_cursor_next_split(&path_and_query, '?', &sub_string) == true) {
aws_byte_cursor_next_split(&path_and_query, '?', &sub_string);
struct aws_uri_param param;
AWS_ZERO_STRUCT(param);
struct aws_byte_cursor part_number_query_str = aws_byte_cursor_from_c_str("partNumber");
while (aws_query_string_next_param(sub_string, ¶m)) {
if (aws_byte_cursor_eq(¶m.key, &part_number_query_str)) {
return aws_s3_meta_request_default_new(
client->allocator,
client,
AWS_S3_REQUEST_TYPE_GET_OBJECT,
content_length,
false /*should_compute_content_md5*/,
options);
}
}
}
}
return aws_s3_meta_request_auto_ranged_get_new(client->allocator, client, part_size, options);
}
case AWS_S3_META_REQUEST_TYPE_PUT_OBJECT: {
if (body_source_count == 0) {
AWS_LOGF_ERROR(
AWS_LS_S3_META_REQUEST,
"Could not create auto-ranged-put meta request."
" Body must be set via filepath, async stream, or body stream.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
if (options->resume_token == NULL) {
uint64_t client_max_part_size = client->max_part_size;
if (part_size < g_s3_min_upload_part_size) {
AWS_LOGF_WARN(
AWS_LS_S3_META_REQUEST,
"Config part size of %" PRIu64 " is less than the minimum upload part size of %" PRIu64
". Using to the minimum part-size for upload.",
(uint64_t)part_size,
(uint64_t)g_s3_min_upload_part_size);
part_size = g_s3_min_upload_part_size;
}
if (client_max_part_size < (uint64_t)g_s3_min_upload_part_size) {
AWS_LOGF_WARN(
AWS_LS_S3_META_REQUEST,
"Client config max part size of %" PRIu64
" is less than the minimum upload part size of %" PRIu64
". Clamping to the minimum part-size for upload.",
(uint64_t)client_max_part_size,
(uint64_t)g_s3_min_upload_part_size);
client_max_part_size = (uint64_t)g_s3_min_upload_part_size;
}
uint32_t num_parts = 0;
if (content_length_found) {
size_t out_part_size = 0;
if (aws_s3_calculate_optimal_mpu_part_size_and_num_parts(
content_length, part_size, client_max_part_size, &out_part_size, &num_parts)) {
return NULL;
}
part_size = out_part_size;
}
if (part_size != options->part_size && part_size != client->part_size) {
AWS_LOGF_DEBUG(
AWS_LS_S3_META_REQUEST,
"The multipart upload part size has been adjusted to %" PRIu64 "",
(uint64_t)part_size);
}
/* Default to client level setting */
uint64_t multipart_upload_threshold = client->multipart_upload_threshold;
if (options->multipart_upload_threshold != 0) {
/* If the threshold is set for the meta request, use it */
multipart_upload_threshold = options->multipart_upload_threshold;
} else if (options->part_size != 0) {
/* If the threshold is not set, but the part size is set for the meta request, use it */
multipart_upload_threshold = part_size;
}
if (content_length_found && content_length <= multipart_upload_threshold) {
return aws_s3_meta_request_default_new(
client->allocator,
client,
AWS_S3_REQUEST_TYPE_PUT_OBJECT,
content_length,
client->compute_content_md5 == AWS_MR_CONTENT_MD5_ENABLED &&
!aws_http_headers_has(initial_message_headers, g_content_md5_header_name),
options);
} else {
if (aws_s3_message_util_check_checksum_header(options->message)) {
/* The checksum header has been set and the request will be split. We fail the request */
AWS_LOGF_ERROR(
AWS_LS_S3_META_REQUEST,
"Could not create auto-ranged-put meta request; checksum headers has been set for "
"auto-ranged-put that will be split. Pre-calculated checksums are only supported for "
"single part upload.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
}
return aws_s3_meta_request_auto_ranged_put_new(
client->allocator, client, part_size, content_length_found, content_length, num_parts, options);
} else { /* else using resume token */
if (!content_length_found) {
AWS_LOGF_ERROR(
AWS_LS_S3_META_REQUEST,
"Could not create auto-ranged-put resume meta request; content_length must be specified.");
aws_raise_error(AWS_ERROR_INVALID_ARGUMENT);
return NULL;
}
/* don't pass part size and total num parts. constructor will pick it up from token */
return aws_s3_meta_request_auto_ranged_put_new(
client->allocator, client, 0, true, content_length, 0, options);
}
}
case AWS_S3_META_REQUEST_TYPE_COPY_OBJECT: {
return aws_s3_meta_request_copy_object_new(client->allocator, client, options);
}
case AWS_S3_META_REQUEST_TYPE_DEFAULT:
return aws_s3_meta_request_default_new(
client->allocator,
client,
AWS_S3_REQUEST_TYPE_UNKNOWN,
content_length,
false /*should_compute_content_md5*/,
options);
default:
AWS_FATAL_ASSERT(false);
}
return NULL;
}
static void s_s3_client_push_meta_request_synced(
struct aws_s3_client *client,
struct aws_s3_meta_request *meta_request) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(meta_request);
ASSERT_SYNCED_DATA_LOCK_HELD(client);
struct aws_s3_meta_request_work *meta_request_work =
aws_mem_calloc(client->allocator, 1, sizeof(struct aws_s3_meta_request_work));
meta_request_work->meta_request = aws_s3_meta_request_acquire(meta_request);
aws_linked_list_push_back(&client->synced_data.pending_meta_request_work, &meta_request_work->node);
}
static void s_s3_client_schedule_process_work_synced(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(client->vtable);
AWS_PRECONDITION(client->vtable->schedule_process_work_synced);
ASSERT_SYNCED_DATA_LOCK_HELD(client);
client->vtable->schedule_process_work_synced(client);
}
static void s_s3_client_schedule_process_work_synced_default(struct aws_s3_client *client) {
ASSERT_SYNCED_DATA_LOCK_HELD(client);
if (client->synced_data.process_work_task_scheduled) {
return;
}
aws_task_init(
&client->synced_data.process_work_task, s_s3_client_process_work_task, client, "s3_client_process_work_task");
aws_event_loop_schedule_task_now(client->process_work_event_loop, &client->synced_data.process_work_task);
client->synced_data.process_work_task_scheduled = true;
}
/* Task function for trying to find a request that can be processed. */
static void s_s3_client_trim_buffer_pool_task(struct aws_task *task, void *arg, enum aws_task_status task_status) {
AWS_PRECONDITION(task);
(void)task;
(void)task_status;
if (task_status != AWS_TASK_STATUS_RUN_READY) {
return;
}
struct aws_s3_client *client = arg;
AWS_PRECONDITION(client);
client->threaded_data.trim_buffer_pool_task_scheduled = false;
uint32_t num_reqs_in_flight = (uint32_t)aws_atomic_load_int(&client->stats.num_requests_in_flight);
if (num_reqs_in_flight == 0) {
aws_s3_buffer_pool_trim(client->buffer_pool);
}
}
static void s_s3_client_schedule_buffer_pool_trim_synced(struct aws_s3_client *client) {
ASSERT_SYNCED_DATA_LOCK_HELD(client);
if (client->threaded_data.trim_buffer_pool_task_scheduled) {
return;
}
uint32_t num_reqs_in_flight = (uint32_t)aws_atomic_load_int(&client->stats.num_requests_in_flight);
if (num_reqs_in_flight > 0) {
return;
}
aws_task_init(
&client->synced_data.trim_buffer_pool_task,
s_s3_client_trim_buffer_pool_task,
client,
"s3_client_buffer_pool_trim_task");
uint64_t trim_time = 0;
aws_event_loop_current_clock_time(client->process_work_event_loop, &trim_time);
trim_time +=
aws_timestamp_convert(s_buffer_pool_trim_time_offset_in_s, AWS_TIMESTAMP_SECS, AWS_TIMESTAMP_NANOS, NULL);
aws_event_loop_schedule_task_future(
client->process_work_event_loop, &client->synced_data.trim_buffer_pool_task, trim_time);
client->threaded_data.trim_buffer_pool_task_scheduled = true;
}
void aws_s3_client_schedule_process_work(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
}
static void s_s3_client_remove_meta_request_threaded(
struct aws_s3_client *client,
struct aws_s3_meta_request *meta_request) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(meta_request);
(void)client;
aws_linked_list_remove(&meta_request->client_process_work_threaded_data.node);
meta_request->client_process_work_threaded_data.scheduled = false;
aws_s3_meta_request_release(meta_request);
}
/* Task function for trying to find a request that can be processed. */
static void s_s3_client_process_work_task(struct aws_task *task, void *arg, enum aws_task_status task_status) {
AWS_PRECONDITION(task);
(void)task;
(void)task_status;
/* Client keeps a reference to the event loop group; a 'canceled' status should not happen.*/
AWS_ASSERT(task_status == AWS_TASK_STATUS_RUN_READY);
struct aws_s3_client *client = arg;
AWS_PRECONDITION(client);
AWS_PRECONDITION(client->vtable);
AWS_PRECONDITION(client->vtable->process_work);
client->vtable->process_work(client);
}
static void s_s3_client_process_work_default(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(client->vtable);
AWS_PRECONDITION(client->vtable->finish_destroy);
struct aws_linked_list meta_request_work_list;
aws_linked_list_init(&meta_request_work_list);
/*******************/
/* Step 1: Move relevant data into thread local memory. */
/*******************/
AWS_LOGF_DEBUG(
AWS_LS_S3_CLIENT,
"id=%p s_s3_client_process_work_default - Moving relevant synced_data into threaded_data.",
(void *)client);
/* BEGIN CRITICAL SECTION */
aws_s3_client_lock_synced_data(client);
/* Once we exit this mutex, someone can reschedule this task. */
client->synced_data.process_work_task_scheduled = false;
client->synced_data.process_work_task_in_progress = true;
if (client->synced_data.active) {
s_s3_client_schedule_buffer_pool_trim_synced(client);
}
aws_linked_list_swap_contents(&meta_request_work_list, &client->synced_data.pending_meta_request_work);
uint32_t num_requests_queued =
aws_s3_client_queue_requests_threaded(client, &client->synced_data.prepared_requests, false);
{
int sub_result = aws_sub_u32_checked(
client->threaded_data.num_requests_being_prepared,
num_requests_queued,
&client->threaded_data.num_requests_being_prepared);
AWS_ASSERT(sub_result == AWS_OP_SUCCESS);
(void)sub_result;
}
{
int sub_result = aws_sub_u32_checked(
client->threaded_data.num_requests_being_prepared,
client->synced_data.num_failed_prepare_requests,
&client->threaded_data.num_requests_being_prepared);
client->synced_data.num_failed_prepare_requests = 0;
AWS_ASSERT(sub_result == AWS_OP_SUCCESS);
(void)sub_result;
}
uint32_t num_endpoints_in_table = (uint32_t)aws_hash_table_get_entry_count(&client->synced_data.endpoints);
uint32_t num_endpoints_allocated = client->synced_data.num_endpoints_allocated;
aws_s3_client_unlock_synced_data(client);
/* END CRITICAL SECTION */
/*******************/
/* Step 2: Push meta requests into the thread local list if they haven't already been scheduled. */
/*******************/
AWS_LOGF_DEBUG(
AWS_LS_S3_CLIENT, "id=%p s_s3_client_process_work_default - Processing any new meta requests.", (void *)client);
while (!aws_linked_list_empty(&meta_request_work_list)) {
struct aws_linked_list_node *node = aws_linked_list_pop_back(&meta_request_work_list);
struct aws_s3_meta_request_work *meta_request_work =
AWS_CONTAINER_OF(node, struct aws_s3_meta_request_work, node);
AWS_FATAL_ASSERT(meta_request_work != NULL);
AWS_FATAL_ASSERT(meta_request_work->meta_request != NULL);
struct aws_s3_meta_request *meta_request = meta_request_work->meta_request;
if (!meta_request->client_process_work_threaded_data.scheduled) {
aws_linked_list_push_back(
&client->threaded_data.meta_requests, &meta_request->client_process_work_threaded_data.node);
meta_request->client_process_work_threaded_data.scheduled = true;
} else {
meta_request = aws_s3_meta_request_release(meta_request);
}
aws_mem_release(client->allocator, meta_request_work);
}
/*******************/
/* Step 3: Update relevant meta requests and connections. */
/*******************/
{
AWS_LOGF_DEBUG(AWS_LS_S3_CLIENT, "id=%p Updating meta requests.", (void *)client);
aws_s3_client_update_meta_requests_threaded(client);
AWS_LOGF_DEBUG(
AWS_LS_S3_CLIENT, "id=%p Updating connections, assigning requests where possible.", (void *)client);
aws_s3_client_update_connections_threaded(client);
}
/*******************/
/* Step 4: Log client stats. */
/*******************/
{
uint32_t num_requests_tracked_requests = (uint32_t)aws_atomic_load_int(&client->stats.num_requests_in_flight);
uint32_t num_auto_ranged_get_network_io =
s_s3_client_get_num_requests_network_io(client, AWS_S3_META_REQUEST_TYPE_GET_OBJECT);
uint32_t num_auto_ranged_put_network_io =
s_s3_client_get_num_requests_network_io(client, AWS_S3_META_REQUEST_TYPE_PUT_OBJECT);
uint32_t num_auto_default_network_io =
s_s3_client_get_num_requests_network_io(client, AWS_S3_META_REQUEST_TYPE_DEFAULT);
uint32_t num_requests_network_io =
s_s3_client_get_num_requests_network_io(client, AWS_S3_META_REQUEST_TYPE_MAX);
uint32_t num_requests_stream_queued_waiting =
(uint32_t)aws_atomic_load_int(&client->stats.num_requests_stream_queued_waiting);
uint32_t num_requests_being_prepared = client->threaded_data.num_requests_being_prepared;
uint32_t num_requests_streaming_response =
(uint32_t)aws_atomic_load_int(&client->stats.num_requests_streaming_response);
uint32_t total_approx_requests = num_requests_network_io + num_requests_stream_queued_waiting +
num_requests_streaming_response + num_requests_being_prepared +
client->threaded_data.request_queue_size;
AWS_LOGF(
s_log_level_client_stats,
AWS_LS_S3_CLIENT_STATS,
"id=%p Requests-in-flight(approx/exact):%d/%d Requests-preparing:%d Requests-queued:%d "
"Requests-network(get/put/default/total):%d/%d/%d/%d Requests-streaming-waiting:%d "
"Requests-streaming-response:%d "
" Endpoints(in-table/allocated):%d/%d",
(void *)client,
total_approx_requests,
num_requests_tracked_requests,
num_requests_being_prepared,
client->threaded_data.request_queue_size,
num_auto_ranged_get_network_io,
num_auto_ranged_put_network_io,
num_auto_default_network_io,
num_requests_network_io,
num_requests_stream_queued_waiting,
num_requests_streaming_response,
num_endpoints_in_table,
num_endpoints_allocated);
}
/*******************/
/* Step 5: Check for client shutdown. */
/*******************/
{
/* BEGIN CRITICAL SECTION */
aws_s3_client_lock_synced_data(client);
client->synced_data.process_work_task_in_progress = false;
/* This flag should never be set twice. If it was, that means a double-free could occur.*/
AWS_ASSERT(!client->synced_data.finish_destroy);
bool finish_destroy =
client->synced_data.active == false && client->synced_data.start_destroy_executing == false &&
client->synced_data.body_streaming_elg_allocated == false &&
client->synced_data.process_work_task_scheduled == false &&
client->synced_data.process_work_task_in_progress == false &&
client->synced_data.s3express_provider_active == false && client->synced_data.num_endpoints_allocated == 0;
client->synced_data.finish_destroy = finish_destroy;
if (!client->synced_data.active) {
AWS_LOGF_DEBUG(
AWS_LS_S3_CLIENT,
"id=%p Client shutdown progress: starting_destroy_executing=%d body_streaming_elg_allocated=%d "
"process_work_task_scheduled=%d process_work_task_in_progress=%d num_endpoints_allocated=%d "
"s3express_provider_active=%d finish_destroy=%d",
(void *)client,
(int)client->synced_data.start_destroy_executing,
(int)client->synced_data.body_streaming_elg_allocated,
(int)client->synced_data.process_work_task_scheduled,
(int)client->synced_data.process_work_task_in_progress,
(int)client->synced_data.num_endpoints_allocated,
(int)client->synced_data.s3express_provider_active,
(int)client->synced_data.finish_destroy);
}
aws_s3_client_unlock_synced_data(client);
/* END CRITICAL SECTION */
if (finish_destroy) {
client->vtable->finish_destroy(client);
}
}
}
static void s_s3_client_prepare_callback_queue_request(
struct aws_s3_meta_request *meta_request,
struct aws_s3_request *request,
int error_code,
void *user_data);
static bool s_s3_client_should_update_meta_request(
struct aws_s3_client *client,
struct aws_s3_meta_request *meta_request,
uint32_t num_requests_in_flight,
const uint32_t max_requests_in_flight,
const uint32_t max_requests_prepare) {
/* CreateSession has high priority to bypass the checks. */
if (meta_request->type == AWS_S3_META_REQUEST_TYPE_DEFAULT) {
struct aws_s3_meta_request_default *meta_request_default = meta_request->impl;
if (aws_string_eq_c_str(meta_request_default->operation_name, "CreateSession")) {
return true;
}
}
/**
* If number of being-prepared + already-prepared-and-queued requests is more than the max that can
* be in the preparation stage.
* Or total number of requests tracked by the client is more than the max tracked ("in flight")
* requests.
*
* We cannot create more requests for this meta request.
*/
if ((client->threaded_data.num_requests_being_prepared + client->threaded_data.request_queue_size) >=
max_requests_prepare) {
return false;
}
if (num_requests_in_flight >= max_requests_in_flight) {
return false;
}
/* If this particular endpoint doesn't have any known addresses yet, then we don't want to go full speed in
* ramping up requests just yet. If there is already enough in the queue for one address (even if those
* aren't for this particular endpoint) we skip over this meta request for now. */
struct aws_s3_endpoint *endpoint = meta_request->endpoint;
AWS_ASSERT(endpoint != NULL);
AWS_ASSERT(client->vtable->get_host_address_count);
size_t num_known_vips = client->vtable->get_host_address_count(
client->client_bootstrap->host_resolver, endpoint->host_name, AWS_GET_HOST_ADDRESS_COUNT_RECORD_TYPE_A);
if (num_known_vips == 0 && (client->threaded_data.num_requests_being_prepared +
client->threaded_data.request_queue_size) >= g_max_num_connections_per_vip) {
return false;
}
/* Nothing blocks the meta request to create more requests */
return true;
}
void aws_s3_client_update_meta_requests_threaded(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
const uint32_t max_requests_in_flight = aws_s3_client_get_max_requests_in_flight(client);
const uint32_t max_requests_prepare = aws_s3_client_get_max_requests_prepare(client);
struct aws_linked_list meta_requests_work_remaining;
aws_linked_list_init(&meta_requests_work_remaining);
uint32_t num_requests_in_flight = (uint32_t)aws_atomic_load_int(&client->stats.num_requests_in_flight);
const uint32_t pass_flags[] = {
AWS_S3_META_REQUEST_UPDATE_FLAG_CONSERVATIVE,
0,
};
const uint32_t num_passes = AWS_ARRAY_SIZE(pass_flags);
aws_s3_buffer_pool_remove_reservation_hold(client->buffer_pool);
for (uint32_t pass_index = 0; pass_index < num_passes; ++pass_index) {
/**
* Iterate through the meta requests to update meta requests and get new requests that can then be prepared
+ * (reading from any streams, signing, etc.) for sending.
*/
while (!aws_linked_list_empty(&client->threaded_data.meta_requests)) {
struct aws_linked_list_node *meta_request_node =
aws_linked_list_begin(&client->threaded_data.meta_requests);
struct aws_s3_meta_request *meta_request =
AWS_CONTAINER_OF(meta_request_node, struct aws_s3_meta_request, client_process_work_threaded_data);
if (!s_s3_client_should_update_meta_request(
client, meta_request, num_requests_in_flight, max_requests_in_flight, max_requests_prepare)) {
/* Move the meta request to be processed from next loop. */
aws_linked_list_remove(&meta_request->client_process_work_threaded_data.node);
aws_linked_list_push_back(
&meta_requests_work_remaining, &meta_request->client_process_work_threaded_data.node);
continue;
}
struct aws_s3_request *request = NULL;
/* Try to grab the next request from the meta request. */
/* TODO: should we bail out if request fails to update due to mem or
* continue going and hopping that following reqs can fit into mem?
* check if avail space is at least part size? */
bool work_remaining = aws_s3_meta_request_update(meta_request, pass_flags[pass_index], &request);
if (work_remaining) {
/* If there is work remaining, but we didn't get a request back, take the meta request out of the
* list so that we don't use it again during this function, with the intention of putting it back in
* the list before this function ends. */
if (request == NULL) {
aws_linked_list_remove(&meta_request->client_process_work_threaded_data.node);
aws_linked_list_push_back(
&meta_requests_work_remaining, &meta_request->client_process_work_threaded_data.node);
} else {
request->tracked_by_client = true;
++client->threaded_data.num_requests_being_prepared;
num_requests_in_flight =
(uint32_t)aws_atomic_fetch_add(&client->stats.num_requests_in_flight, 1) + 1;
aws_s3_meta_request_prepare_request(
meta_request, request, s_s3_client_prepare_callback_queue_request, client);
}
} else {
s_s3_client_remove_meta_request_threaded(client, meta_request);
}
}
aws_linked_list_move_all_front(&client->threaded_data.meta_requests, &meta_requests_work_remaining);
}
}
static void s_s3_client_meta_request_finished_request(
struct aws_s3_client *client,
struct aws_s3_meta_request *meta_request,
struct aws_s3_request *request,
int error_code) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(request);
if (request->tracked_by_client) {
/* BEGIN CRITICAL SECTION */
aws_s3_client_lock_synced_data(client);
aws_atomic_fetch_sub(&client->stats.num_requests_in_flight, 1);
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
/* END CRITICAL SECTION */
}
aws_s3_meta_request_finished_request(meta_request, request, error_code);
}
static void s_s3_client_prepare_callback_queue_request(
struct aws_s3_meta_request *meta_request,
struct aws_s3_request *request,
int error_code,
void *user_data) {
AWS_PRECONDITION(meta_request);
AWS_PRECONDITION(request);
struct aws_s3_client *client = user_data;
AWS_PRECONDITION(client);
if (error_code != AWS_ERROR_SUCCESS) {
s_s3_client_meta_request_finished_request(client, meta_request, request, error_code);
request = aws_s3_request_release(request);
}
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
if (error_code == AWS_ERROR_SUCCESS) {
aws_linked_list_push_back(&client->synced_data.prepared_requests, &request->node);
} else {
++client->synced_data.num_failed_prepare_requests;
}
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
}
void aws_s3_client_update_connections_threaded(struct aws_s3_client *client) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(client->vtable);
struct aws_linked_list left_over_requests;
aws_linked_list_init(&left_over_requests);
while (s_s3_client_get_num_requests_network_io(client, AWS_S3_META_REQUEST_TYPE_MAX) <
aws_s3_client_get_max_active_connections(client, NULL) &&
!aws_linked_list_empty(&client->threaded_data.request_queue)) {
struct aws_s3_request *request = aws_s3_client_dequeue_request_threaded(client);
const uint32_t max_active_connections = aws_s3_client_get_max_active_connections(client, request->meta_request);
if (request->is_noop) {
/* If request is no-op, finishes and cleans up the request */
s_s3_client_meta_request_finished_request(client, request->meta_request, request, AWS_ERROR_SUCCESS);
request = aws_s3_request_release(request);
} else if (!request->always_send && aws_s3_meta_request_has_finish_result(request->meta_request)) {
/* Unless the request is marked "always send", if this meta request has a finish result, then finish the
* request now and release it. */
s_s3_client_meta_request_finished_request(client, request->meta_request, request, AWS_ERROR_S3_CANCELED);
request = aws_s3_request_release(request);
} else if (
s_s3_client_get_num_requests_network_io(client, request->meta_request->type) < max_active_connections) {
s_s3_client_create_connection_for_request(client, request);
} else {
/* Push the request into the left-over list to be used in a future call of this function. */
aws_linked_list_push_back(&left_over_requests, &request->node);
}
}
aws_s3_client_queue_requests_threaded(client, &left_over_requests, true);
}
static void s_s3_client_acquired_retry_token(
struct aws_retry_strategy *retry_strategy,
int error_code,
struct aws_retry_token *token,
void *user_data);
static void s_s3_client_retry_ready(struct aws_retry_token *token, int error_code, void *user_data);
static void s_s3_client_create_connection_for_request_default(
struct aws_s3_client *client,
struct aws_s3_request *request);
static void s_s3_client_create_connection_for_request(struct aws_s3_client *client, struct aws_s3_request *request) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(client->vtable);
if (client->vtable->create_connection_for_request) {
client->vtable->create_connection_for_request(client, request);
return;
}
s_s3_client_create_connection_for_request_default(client, request);
}
static void s_s3_client_create_connection_for_request_default(
struct aws_s3_client *client,
struct aws_s3_request *request) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(request);
struct aws_s3_meta_request *meta_request = request->meta_request;
AWS_PRECONDITION(meta_request);
aws_atomic_fetch_add(&client->stats.num_requests_network_io[meta_request->type], 1);
struct aws_s3_connection *connection = aws_mem_calloc(client->allocator, 1, sizeof(struct aws_s3_connection));
connection->endpoint = aws_s3_endpoint_acquire(meta_request->endpoint, false /*already_holding_lock*/);
connection->request = request;
struct aws_byte_cursor host_header_value;
AWS_ZERO_STRUCT(host_header_value);
struct aws_http_headers *message_headers = aws_http_message_get_headers(meta_request->initial_request_message);
AWS_ASSERT(message_headers);
int result = aws_http_headers_get(message_headers, g_host_header_name, &host_header_value);
AWS_ASSERT(result == AWS_OP_SUCCESS);
(void)result;
if (aws_retry_strategy_acquire_retry_token(
client->retry_strategy, &host_header_value, s_s3_client_acquired_retry_token, connection, 0)) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Client could not acquire retry token for request %p due to error %d (%s)",
(void *)client,
(void *)request,
aws_last_error_or_unknown(),
aws_error_str(aws_last_error_or_unknown()));
goto reset_connection;
}
return;
reset_connection:
aws_s3_client_notify_connection_finished(
client, connection, aws_last_error_or_unknown(), AWS_S3_CONNECTION_FINISH_CODE_FAILED);
}
static void s_s3_client_acquired_retry_token(
struct aws_retry_strategy *retry_strategy,
int error_code,
struct aws_retry_token *token,
void *user_data) {
AWS_PRECONDITION(retry_strategy);
(void)retry_strategy;
struct aws_s3_connection *connection = user_data;
AWS_PRECONDITION(connection);
struct aws_s3_request *request = connection->request;
AWS_PRECONDITION(request);
struct aws_s3_meta_request *meta_request = request->meta_request;
AWS_PRECONDITION(meta_request);
struct aws_s3_endpoint *endpoint = meta_request->endpoint;
AWS_ASSERT(endpoint != NULL);
struct aws_s3_client *client = endpoint->client;
AWS_ASSERT(client != NULL);
if (error_code != AWS_ERROR_SUCCESS) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Client could not get retry token for connection %p processing request %p due to error %d (%s)",
(void *)client,
(void *)connection,
(void *)request,
error_code,
aws_error_str(error_code));
goto error_clean_up;
}
AWS_ASSERT(token);
connection->retry_token = token;
AWS_ASSERT(client->vtable->acquire_http_connection);
/* client needs to be kept alive until s_s3_client_on_acquire_http_connection completes */
/* TODO: not a blocker, consider managing the life time of aws_s3_client from aws_s3_endpoint to simplify usage */
aws_s3_client_acquire(client);
client->vtable->acquire_http_connection(
endpoint->http_connection_manager, s_s3_client_on_acquire_http_connection, connection);
return;
error_clean_up:
aws_s3_client_notify_connection_finished(client, connection, error_code, AWS_S3_CONNECTION_FINISH_CODE_FAILED);
}
static void s_s3_client_on_acquire_http_connection(
struct aws_http_connection *incoming_http_connection,
int error_code,
void *user_data) {
struct aws_s3_connection *connection = user_data;
AWS_PRECONDITION(connection);
struct aws_s3_request *request = connection->request;
AWS_PRECONDITION(request);
struct aws_s3_meta_request *meta_request = request->meta_request;
AWS_PRECONDITION(meta_request);
struct aws_s3_endpoint *endpoint = meta_request->endpoint;
AWS_ASSERT(endpoint != NULL);
struct aws_s3_client *client = endpoint->client;
AWS_ASSERT(client != NULL);
if (error_code != AWS_ERROR_SUCCESS) {
AWS_LOGF_ERROR(
AWS_LS_S3_ENDPOINT,
"id=%p: Could not acquire connection due to error code %d (%s)",
(void *)endpoint,
error_code,
aws_error_str(error_code));
if (error_code == AWS_IO_DNS_INVALID_NAME || error_code == AWS_IO_TLS_ERROR_NEGOTIATION_FAILURE) {
/**
* Fall fast without retry
* - Invalid DNS name will not change after retry.
* - TLS negotiation is expensive and retry will not help in most case.
*/
goto error_fail;
}
goto error_retry;
}
connection->http_connection = incoming_http_connection;
aws_s3_meta_request_send_request(meta_request, connection);
aws_s3_client_release(client); /* kept since this callback was registered */
return;
error_retry:
aws_s3_client_notify_connection_finished(client, connection, error_code, AWS_S3_CONNECTION_FINISH_CODE_RETRY);
aws_s3_client_release(client); /* kept since this callback was registered */
return;
error_fail:
aws_s3_client_notify_connection_finished(client, connection, error_code, AWS_S3_CONNECTION_FINISH_CODE_FAILED);
aws_s3_client_release(client); /* kept since this callback was registered */
}
/* Called by aws_s3_meta_request when it has finished using this connection for a single request. */
void aws_s3_client_notify_connection_finished(
struct aws_s3_client *client,
struct aws_s3_connection *connection,
int error_code,
enum aws_s3_connection_finish_code finish_code) {
AWS_PRECONDITION(client);
AWS_PRECONDITION(connection);
struct aws_s3_request *request = connection->request;
AWS_PRECONDITION(request);
struct aws_s3_meta_request *meta_request = request->meta_request;
AWS_PRECONDITION(meta_request);
AWS_PRECONDITION(meta_request->initial_request_message);
struct aws_s3_endpoint *endpoint = meta_request->endpoint;
AWS_PRECONDITION(endpoint);
if (request->send_data.metrics) {
request->send_data.metrics->crt_info_metrics.error_code = error_code;
}
/* If we're trying to set up a retry... */
if (finish_code == AWS_S3_CONNECTION_FINISH_CODE_RETRY) {
if (connection->retry_token == NULL) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Client could not schedule retry of request %p for meta request %p, as retry token is NULL.",
(void *)client,
(void *)request,
(void *)meta_request);
goto reset_connection;
}
if (aws_s3_meta_request_is_finished(meta_request)) {
AWS_LOGF_DEBUG(
AWS_LS_S3_CLIENT,
"id=%p Client not scheduling retry of request %p for meta request %p with token %p because meta "
"request has been flagged as finished.",
(void *)client,
(void *)request,
(void *)meta_request,
(void *)connection->retry_token);
goto reset_connection;
}
AWS_LOGF_DEBUG(
AWS_LS_S3_CLIENT,
"id=%p Client scheduling retry of request %p for meta request %p with token %p with error code %d (%s).",
(void *)client,
(void *)request,
(void *)meta_request,
(void *)connection->retry_token,
error_code,
aws_error_str(error_code));
enum aws_retry_error_type error_type = AWS_RETRY_ERROR_TYPE_TRANSIENT;
switch (error_code) {
case AWS_ERROR_S3_INTERNAL_ERROR:
error_type = AWS_RETRY_ERROR_TYPE_SERVER_ERROR;
break;
case AWS_ERROR_S3_SLOW_DOWN:
error_type = AWS_RETRY_ERROR_TYPE_THROTTLING;
break;
}
if (connection->http_connection != NULL) {
AWS_ASSERT(endpoint->http_connection_manager);
aws_http_connection_manager_release_connection(
endpoint->http_connection_manager, connection->http_connection);
connection->http_connection = NULL;
}
/* Ask the retry strategy to schedule a retry of the request. */
if (aws_retry_strategy_schedule_retry(
connection->retry_token, error_type, s_s3_client_retry_ready, connection)) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Client could not retry request %p for meta request %p with token %p due to error %d (%s)",
(void *)client,
(void *)request,
(void *)meta_request,
(void *)connection->retry_token,
aws_last_error_or_unknown(),
aws_error_str(aws_last_error_or_unknown()));
goto reset_connection;
}
return;
}
reset_connection:
if (connection->retry_token != NULL) {
/* If we have a retry token and successfully finished, record that success. */
if (finish_code == AWS_S3_CONNECTION_FINISH_CODE_SUCCESS) {
aws_retry_token_record_success(connection->retry_token);
}
aws_retry_token_release(connection->retry_token);
connection->retry_token = NULL;
}
/* If we weren't successful, and we're here, that means this failure is not eligible for a retry. So finish the
* request, and close our HTTP connection. */
if (finish_code != AWS_S3_CONNECTION_FINISH_CODE_SUCCESS) {
if (connection->http_connection != NULL) {
aws_http_connection_close(connection->http_connection);
}
}
aws_atomic_fetch_sub(&client->stats.num_requests_network_io[meta_request->type], 1);
s_s3_client_meta_request_finished_request(client, meta_request, request, error_code);
if (connection->http_connection != NULL) {
AWS_ASSERT(endpoint->http_connection_manager);
aws_http_connection_manager_release_connection(endpoint->http_connection_manager, connection->http_connection);
connection->http_connection = NULL;
}
if (connection->request != NULL) {
connection->request = aws_s3_request_release(connection->request);
}
aws_retry_token_release(connection->retry_token);
connection->retry_token = NULL;
aws_s3_endpoint_release(connection->endpoint);
connection->endpoint = NULL;
aws_mem_release(client->allocator, connection);
connection = NULL;
/* BEGIN CRITICAL SECTION */
{
aws_s3_client_lock_synced_data(client);
s_s3_client_schedule_process_work_synced(client);
aws_s3_client_unlock_synced_data(client);
}
/* END CRITICAL SECTION */
}
static void s_s3_client_prepare_request_callback_retry_request(
struct aws_s3_meta_request *meta_request,
struct aws_s3_request *request,
int error_code,
void *user_data);
static void s_s3_client_retry_ready(struct aws_retry_token *token, int error_code, void *user_data) {
AWS_PRECONDITION(token);
(void)token;
struct aws_s3_connection *connection = user_data;
AWS_PRECONDITION(connection);
struct aws_s3_request *request = connection->request;
AWS_PRECONDITION(request);
struct aws_s3_meta_request *meta_request = request->meta_request;
AWS_PRECONDITION(meta_request);
struct aws_s3_endpoint *endpoint = meta_request->endpoint;
AWS_PRECONDITION(endpoint);
struct aws_s3_client *client = endpoint->client;
AWS_PRECONDITION(client);
/* If we couldn't retry this request, then bail on the entire meta request. */
if (error_code != AWS_ERROR_SUCCESS) {
AWS_LOGF_ERROR(
AWS_LS_S3_CLIENT,
"id=%p Client could not retry request %p for meta request %p due to error %d (%s)",
(void *)client,
(void *)meta_request,
(void *)request,
error_code,
aws_error_str(error_code));
goto error_clean_up;
}
AWS_LOGF_DEBUG(
AWS_LS_S3_META_REQUEST,
"id=%p Client retrying request %p for meta request %p on connection %p with retry token %p",
(void *)client,
(void *)request,
(void *)meta_request,
(void *)connection,
(void *)connection->retry_token);
aws_s3_meta_request_prepare_request(
meta_request, request, s_s3_client_prepare_request_callback_retry_request, connection);
return;
error_clean_up:
aws_s3_client_notify_connection_finished(client, connection, error_code, AWS_S3_CONNECTION_FINISH_CODE_FAILED);
}
static void s_s3_client_prepare_request_callback_retry_request(
struct aws_s3_meta_request *meta_request,
struct aws_s3_request *request,
int error_code,
void *user_data) {
AWS_PRECONDITION(meta_request);
(void)meta_request;
AWS_PRECONDITION(request);
(void)request;
struct aws_s3_connection *connection = user_data;
AWS_PRECONDITION(connection);
struct aws_s3_endpoint *endpoint = meta_request->endpoint;
AWS_ASSERT(endpoint != NULL);
struct aws_s3_client *client = endpoint->client;
AWS_ASSERT(client != NULL);
if (error_code == AWS_ERROR_SUCCESS) {
AWS_ASSERT(connection->retry_token);
s_s3_client_acquired_retry_token(
client->retry_strategy, AWS_ERROR_SUCCESS, connection->retry_token, connection);
} else {
aws_s3_client_notify_connection_finished(client, connection, error_code, AWS_S3_CONNECTION_FINISH_CODE_FAILED);
}
}
static void s_resume_token_ref_count_zero_callback(void *arg) {
struct aws_s3_meta_request_resume_token *token = arg;
aws_string_destroy(token->multipart_upload_id);
aws_mem_release(token->allocator, token);
}
struct aws_s3_meta_request_resume_token *aws_s3_meta_request_resume_token_new(struct aws_allocator *allocator) {
struct aws_s3_meta_request_resume_token *token =
aws_mem_calloc(allocator, 1, sizeof(struct aws_s3_meta_request_resume_token));
token->allocator = allocator;
aws_ref_count_init(&token->ref_count, token, s_resume_token_ref_count_zero_callback);
return token;
}
struct aws_s3_meta_request_resume_token *aws_s3_meta_request_resume_token_new_upload(
struct aws_allocator *allocator,
const struct aws_s3_upload_resume_token_options *options) {
AWS_PRECONDITION(allocator);
AWS_PRECONDITION(options);
if (options->part_size > SIZE_MAX) {
aws_raise_error(AWS_ERROR_OVERFLOW_DETECTED);
return NULL;
}
struct aws_s3_meta_request_resume_token *token = aws_s3_meta_request_resume_token_new(allocator);
token->multipart_upload_id = aws_string_new_from_cursor(allocator, &options->upload_id);
token->part_size = (size_t)options->part_size;
token->total_num_parts = options->total_num_parts;
token->num_parts_completed = options->num_parts_completed;
token->type = AWS_S3_META_REQUEST_TYPE_PUT_OBJECT;
return token;
}
struct aws_s3_meta_request_resume_token *aws_s3_meta_request_resume_token_acquire(
struct aws_s3_meta_request_resume_token *resume_token) {
if (resume_token) {
aws_ref_count_acquire(&resume_token->ref_count);
}
return resume_token;
}
struct aws_s3_meta_request_resume_token *aws_s3_meta_request_resume_token_release(
struct aws_s3_meta_request_resume_token *resume_token) {
if (resume_token) {
aws_ref_count_release(&resume_token->ref_count);
}
return NULL;
}
enum aws_s3_meta_request_type aws_s3_meta_request_resume_token_type(
struct aws_s3_meta_request_resume_token *resume_token) {
AWS_FATAL_PRECONDITION(resume_token);
return resume_token->type;
}
uint64_t aws_s3_meta_request_resume_token_part_size(struct aws_s3_meta_request_resume_token *resume_token) {
AWS_FATAL_PRECONDITION(resume_token);
return (uint64_t)resume_token->part_size;
}
size_t aws_s3_meta_request_resume_token_total_num_parts(struct aws_s3_meta_request_resume_token *resume_token) {
AWS_FATAL_PRECONDITION(resume_token);
return resume_token->total_num_parts;
}
size_t aws_s3_meta_request_resume_token_num_parts_completed(struct aws_s3_meta_request_resume_token *resume_token) {
AWS_FATAL_PRECONDITION(resume_token);
return resume_token->num_parts_completed;
}
struct aws_byte_cursor aws_s3_meta_request_resume_token_upload_id(
struct aws_s3_meta_request_resume_token *resume_token) {
AWS_FATAL_PRECONDITION(resume_token);
if (resume_token->type == AWS_S3_META_REQUEST_TYPE_PUT_OBJECT && resume_token->multipart_upload_id != NULL) {
return aws_byte_cursor_from_string(resume_token->multipart_upload_id);
}
return aws_byte_cursor_from_c_str("");
}
static uint64_t s_upload_timeout_threshold_ns = 5000000000; /* 5 Secs */
const size_t g_expect_timeout_offset_ms =
700; /* 0.7 Secs. From experiments on c5n.18xlarge machine for 30 GiB upload, it gave us best performance. */
/**
* The upload timeout optimization: explained.
*
* Sometimes, S3 is extremely slow responding to an upload.
* In these cases, it's much faster to cancel and resend the upload,
* vs waiting 5sec for the slow response.
*
* Typically, S3 responds to an upload in 0.2sec after the request is fully received.
* But occasionally (about 0.1%) it takes 5sec to respond.
* In a large 30GiB file upload, you can expect about 4 parts to suffer from
* a slow response. If one of these parts is near the end of the file,
* then we end up sitting around doing nothing for up to 5sec, waiting
* for this final slow upload to complete.
*
* We use the response_first_byte_timeout HTTP option to cancel uploads
* suffering from a slow response. But how should we set it? A fast 100Gbps
* machine definitely wants it! But a slow computer does not. A slow computer
* would be better off waiting 5sec for the response, vs re-uploading the whole request.
*
* The current algorithm:
* 1. Start without a timeout value. After 10 requests completed, we know the average of how long the
* request takes. We decide if it's worth to set a timeout value or not. (If the average of request takes more than
* 5 secs or not) TODO: if the client have different part size, this doesn't make sense
* 2. If it is worth to retry, start with a default timeout value, 1 sec.
* 3. If a request finishes successfully, use the average response_to_first_byte_time + g_expect_timeout_offset_ms as
* our expected timeout value. (TODO: The real expected timeout value should be a P99 of all the requests.)
* 3.1 Adjust the current timeout value against the expected timeout value, via 0.99 * <current timeout> + 0.01 *
* <expected timeout> to get closer to the expected timeout value.
* 4. If request had timed out. We check the timeout rate.
* 4.1 If timeout rate is larger than 0.1%, we increase the timeout value by 100ms (Check the timeout value when the
* request was made, if the updated timeout value is larger than the expected, skip update).
* 4.2 If timeout rate is larger than 1%, we increase the timeout value by 1 secs (If needed). And clear the rate
* to get the exact rate with new timeout value.
* 4.3 Once the timeout value is larger than 5 secs, we stop the process.
*
* Invoked from `s_s3_auto_ranged_put_send_request_finish`.
*/
void aws_s3_client_update_upload_part_timeout(
struct aws_s3_client *client,
struct aws_s3_request *finished_upload_part_request,
int finished_error_code) {
aws_s3_client_lock_synced_data(client);
struct aws_s3_upload_part_timeout_stats *stats = &client->synced_data.upload_part_stats;
if (stats->stop_timeout) {
/* Timeout was disabled */
goto unlock;
}
struct aws_s3_request_metrics *metrics = finished_upload_part_request->send_data.metrics;
size_t current_timeout_ms = aws_atomic_load_int(&client->upload_timeout_ms);
uint64_t current_timeout_ns =
aws_timestamp_convert(current_timeout_ms, AWS_TIMESTAMP_MILLIS, AWS_TIMESTAMP_NANOS, NULL);
uint64_t updated_timeout_ns = 0;
uint64_t expect_timeout_offset_ns =
aws_timestamp_convert(g_expect_timeout_offset_ms, AWS_TIMESTAMP_MILLIS, AWS_TIMESTAMP_NANOS, NULL);
switch (finished_error_code) {
case AWS_ERROR_SUCCESS:
/* We only interested in request succeed */
stats->num_successful_upload_requests = aws_add_u64_saturating(stats->num_successful_upload_requests, 1);
if (stats->num_successful_upload_requests <= 10) {
/* Gether the data */
uint64_t request_time_ns =
metrics->time_metrics.receive_end_timestamp_ns - metrics->time_metrics.send_start_timestamp_ns;
stats->initial_request_time.sum_ns =
aws_add_u64_saturating(stats->initial_request_time.sum_ns, request_time_ns);
++stats->initial_request_time.num_samples;
if (stats->num_successful_upload_requests == 10) {
/* Decide we need a timeout or not */
uint64_t average_request_time_ns =
stats->initial_request_time.sum_ns / stats->initial_request_time.num_samples;
if (average_request_time_ns >= s_upload_timeout_threshold_ns) {
/* We don't need a timeout, as retry will be slower than just wait for the server to response */
stats->stop_timeout = true;
} else {
/* Start the timeout at 1 secs */
aws_atomic_store_int(&client->upload_timeout_ms, 1000);
}
}
goto unlock;
}
/* Starts to update timeout on case of succeed */
stats->timeout_rate_tracking.num_completed =
aws_add_u64_saturating(stats->timeout_rate_tracking.num_completed, 1);
/* Response to first byte is time taken for the first byte data received from the request finished
* sending */
uint64_t response_to_first_byte_time_ns =
metrics->time_metrics.receive_start_timestamp_ns - metrics->time_metrics.send_end_timestamp_ns;
stats->response_to_first_byte_time.sum_ns =
aws_add_u64_saturating(stats->response_to_first_byte_time.sum_ns, response_to_first_byte_time_ns);
stats->response_to_first_byte_time.num_samples =
aws_add_u64_saturating(stats->response_to_first_byte_time.num_samples, 1);
uint64_t average_response_to_first_byte_time_ns =
stats->response_to_first_byte_time.sum_ns / stats->response_to_first_byte_time.num_samples;
uint64_t expected_timeout_ns = average_response_to_first_byte_time_ns + expect_timeout_offset_ns;
double timeout_ns_double = (double)current_timeout_ns * 0.99 + (double)expected_timeout_ns * 0.01;
updated_timeout_ns = (uint64_t)timeout_ns_double;
break;
case AWS_ERROR_HTTP_RESPONSE_FIRST_BYTE_TIMEOUT:
if (stats->num_successful_upload_requests < 10) {
goto unlock;
}
/* Starts to update timeout on case of timed out */
stats->timeout_rate_tracking.num_completed =
aws_add_u64_saturating(stats->timeout_rate_tracking.num_completed, 1);
stats->timeout_rate_tracking.num_failed =
aws_add_u64_saturating(stats->timeout_rate_tracking.num_failed, 1);
uint64_t timeout_threshold = (uint64_t)ceil((double)stats->timeout_rate_tracking.num_completed / 100);
uint64_t warning_threshold = (uint64_t)ceil((double)stats->timeout_rate_tracking.num_completed / 1000);
if (stats->timeout_rate_tracking.num_failed > timeout_threshold) {
/**
* Restore the rate track, as we are larger than 1%, it goes off the record.
*/
AWS_LOGF_WARN(
AWS_LS_S3_CLIENT,
"id=%p Client upload part timeout rate is larger than expected, current timeout is %zu, bump it "
"up. Request original timeout is: %zu",
(void *)client,
current_timeout_ms,
finished_upload_part_request->upload_timeout_ms);
stats->timeout_rate_tracking.num_completed = 0;
stats->timeout_rate_tracking.num_failed = 0;
if (finished_upload_part_request->upload_timeout_ms + 1000 > current_timeout_ms) {
/* Update the timeout by adding 1 secs only when it's worth to do so */
updated_timeout_ns = aws_add_u64_saturating(
current_timeout_ns, aws_timestamp_convert(1, AWS_TIMESTAMP_SECS, AWS_TIMESTAMP_NANOS, NULL));
}
} else if (stats->timeout_rate_tracking.num_failed > warning_threshold) {
if (finished_upload_part_request->upload_timeout_ms + 100 > current_timeout_ms) {
/* Only update the timeout by adding 100 ms if the request was made with a longer time out. */
updated_timeout_ns = aws_add_u64_saturating(
current_timeout_ns,
aws_timestamp_convert(100, AWS_TIMESTAMP_MILLIS, AWS_TIMESTAMP_NANOS, NULL));
}
}
break;
default:
break;
}
if (updated_timeout_ns != 0) {
if (updated_timeout_ns > s_upload_timeout_threshold_ns) {
/* Stops timeout, as wait for server to response will be faster to set our own timeout */
stats->stop_timeout = true;
/* Unset the upload_timeout */
updated_timeout_ns = 0;
}
/* Apply the updated timeout */
aws_atomic_store_int(
&client->upload_timeout_ms,
(size_t)aws_timestamp_convert(updated_timeout_ns, AWS_TIMESTAMP_NANOS, AWS_TIMESTAMP_MILLIS, NULL));
}
unlock:
aws_s3_client_unlock_synced_data(client);
}
|