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
|
diff -urN linux-2.6.9.orig/include/linux/pkt_sched.h linux-2.6.9/include/linux/pkt_sched.h
--- linux-2.6.9.orig/include/linux/pkt_sched.h 2004-10-18 23:54:40.000000000 +0200
+++ linux-2.6.9/include/linux/pkt_sched.h 2004-12-20 13:04:48.000000000 +0100
@@ -431,4 +431,117 @@
#define NETEM_DIST_SCALE 8192
+
+/* WRR section */
+
+/* Other includes */
+#include <linux/if_ether.h>
+
+// A sub weight and of a class
+// All numbers are represented as parts of (2^64-1).
+struct tc_wrr_class_weight {
+ __u64 val; // Current value (0 is not valid)
+ __u64 decr; // Value pr bytes (2^64-1 is not valid)
+ __u64 incr; // Value pr seconds (2^64-1 is not valid)
+ __u64 min; // Minimal value (0 is not valid)
+ __u64 max; // Minimal value (0 is not valid)
+
+ // The time where the above information was correct:
+ time_t tim;
+};
+
+// Pakcet send when modifying a class:
+struct tc_wrr_class_modf {
+ // Not-valid values are ignored.
+ struct tc_wrr_class_weight weight1;
+ struct tc_wrr_class_weight weight2;
+};
+
+// Packet returned when quering a class:
+struct tc_wrr_class_stats {
+ char used; // If this is false the information below is invalid
+
+ struct tc_wrr_class_modf class_modf;
+
+ unsigned char addr[ETH_ALEN];
+ char usemac; // True if addr is a MAC address, else it is an IP address
+ // (this value is only for convience, it is always the same
+ // value as in the qdisc)
+ int heappos; // Current heap position or 0 if not in heap
+ __u64 penal_ls; // Penalty value in heap (ls)
+ __u64 penal_ms; // Penalty value in heap (ms)
+};
+
+// Qdisc-wide penalty information (boolean values - 2 not valid)
+struct tc_wrr_qdisc_weight {
+ char weight_mode; // 0=No automatic change to weight
+ // 1=Decrease normally
+ // 2=Also multiply with number of machines
+ // 3=Instead multiply with priority divided
+ // with priority of the other.
+ // -1=no change
+};
+
+// Packet send when modifing a qdisc:
+struct tc_wrr_qdisc_modf {
+ // Not-valid values are ignored:
+ struct tc_wrr_qdisc_weight weight1;
+ struct tc_wrr_qdisc_weight weight2;
+};
+
+// Packet send when creating a qdisc:
+struct tc_wrr_qdisc_crt {
+ struct tc_wrr_qdisc_modf qdisc_modf;
+
+ char srcaddr; // 1=lookup source, 0=lookup destination
+ char usemac; // 1=Classify on MAC addresses, 0=classify on IP
+ char usemasq; // 1=Classify based on masqgrading - only valid
+ // if usemac is zero
+ int bands_max; // Maximal number of bands (i.e.: classes)
+ int proxy_maxconn; // If differnt from 0 then we support proxy remapping
+ // of packets. And this is the number of maximal
+ // concurrent proxy connections.
+};
+
+// Packet returned when quering a qdisc:
+struct tc_wrr_qdisc_stats {
+ struct tc_wrr_qdisc_crt qdisc_crt;
+ int proxy_curconn;
+ int nodes_in_heap; // Current number of bands wanting to send something
+ int bands_cur; // Current number of bands used (i.e.: MAC/IP addresses seen)
+ int bands_reused; // Number of times this band has been reused.
+ int packets_requed; // Number of times packets have been requeued.
+ __u64 priosum; // Sum of priorities in heap where 1 is 2^32
+};
+
+struct tc_wrr_qdisc_modf_std {
+ // This indicates which of the tc_wrr_qdisc_modf structers this is:
+ char proxy; // 0=This struct
+
+ // Should we also change a class?
+ char change_class;
+
+ // Only valid if change_class is false
+ struct tc_wrr_qdisc_modf qdisc_modf;
+
+ // Only valid if change_class is true:
+ unsigned char addr[ETH_ALEN]; // Class to change (non-used bytes should be 0)
+ struct tc_wrr_class_modf class_modf; // The change
+};
+
+// Used for proxyrempping:
+struct tc_wrr_qdisc_modf_proxy {
+ // This indicates which of the tc_wrr_qdisc_modf structers this is:
+ char proxy; // 1=This struct
+
+ // This is 1 if the proxyremap information should be reset
+ char reset;
+
+ // changec is the number of elements in changes.
+ int changec;
+
+ // This is an array of type ProxyRemapBlock:
+ long changes[0];
+};
+
#endif
diff -urN linux-2.6.9.orig/net/sched/Kconfig linux-2.6.9/net/sched/Kconfig
--- linux-2.6.9.orig/net/sched/Kconfig 2004-10-18 23:55:06.000000000 +0200
+++ linux-2.6.9/net/sched/Kconfig 2004-12-20 12:32:26.000000000 +0100
@@ -129,6 +129,21 @@
To compile this code as a module, choose M here: the
module will be called sch_prio.
+config NET_SCH_WRR
+ tristate "WRR packet scheduler"
+ depends on NET_SCHED && IP_NF_CONNTRACK
+ help
+ The weighted robin-robin scheduling algorithm directs network
+ connections to different real servers based on server weights
+ in a round-robin manner. Servers with higher weights receive
+ new connections first than those with less weights, and servers
+ with higher weights get more connections than those with less
+ weights and servers with equal weights get equal connections.
+
+ If you want to compile it in kernel, say Y. If you want to compile
+ it as a module, say M here and read Documentation/modules.txt. If
+ unsure, say N.
+
config NET_SCH_RED
tristate "RED queue"
depends on NET_SCHED
diff -urN linux-2.6.9.orig/net/sched/Makefile linux-2.6.9/net/sched/Makefile
--- linux-2.6.9.orig/net/sched/Makefile 2004-10-18 23:54:08.000000000 +0200
+++ linux-2.6.9/net/sched/Makefile 2004-12-20 12:32:26.000000000 +0100
@@ -20,6 +20,7 @@
obj-$(CONFIG_NET_SCH_INGRESS) += sch_ingress.o
obj-$(CONFIG_NET_SCH_DSMARK) += sch_dsmark.o
obj-$(CONFIG_NET_SCH_SFQ) += sch_sfq.o
+obj-$(CONFIG_NET_SCH_WRR) += sch_wrr.o
obj-$(CONFIG_NET_SCH_TBF) += sch_tbf.o
obj-$(CONFIG_NET_SCH_TEQL) += sch_teql.o
obj-$(CONFIG_NET_SCH_PRIO) += sch_prio.o
diff -urN linux-2.6.9.orig/net/sched/proxydict.c linux-2.6.9/net/sched/proxydict.c
--- linux-2.6.9.orig/net/sched/proxydict.c 1970-01-01 01:00:00.000000000 +0100
+++ linux-2.6.9/net/sched/proxydict.c 2004-12-20 12:32:26.000000000 +0100
@@ -0,0 +1,153 @@
+#ifndef __KERNEL__
+#include <string.h>
+#include <netinet/in.h>
+#endif
+
+#include "proxyremap.h"
+#include "proxydict.h"
+
+
+/*--------------------------------------------------------------------------
+Implementation.
+*/
+
+// Hash function
+#define hash_fnc(m,server,port,proto) \
+ (((proto)*7+(server)*13+(port)*5)%m->hash_size)
+
+// Size of hash table given maximal number of connections:
+#define hash_size_max_con(max_con) (2*(max_con))
+
+// The memory area we maintain:
+typedef struct {
+ int hash_size;
+ int max_con;
+ int cur_con;
+
+ int free_first;
+
+ // Then we have:
+ // int hash_table[hash_size];
+ // int next[max_con];
+ // ProxyRemapBlock info[max_con];
+ //
+ // The idea is the following:
+ // Given a connection we map it by hash_fnc into hash_table. This gives an
+ // index in next which contains a -1 terminated linked list of connections
+ // mapping to that hash value.
+ //
+ // The entries in next not allocated is also in linked list where
+ // the first free index is free_first.
+} memory;
+
+#define Memory(m) ((memory*)m)
+#define Hash_table(m) ((int*)(((char*)m)+sizeof(memory)))
+#define Next(m) ((int*)(((char*)m)+sizeof(memory)+ \
+ sizeof(int)*((memory*)m)->hash_size))
+#define Info(m) ((ProxyRemapBlock*)(((char*)m)+ \
+ sizeof(memory)+ \
+ sizeof(int)*((memory*)m)->hash_size+\
+ sizeof(int)*((memory*)m)->max_con \
+ ))
+
+int proxyGetMemSize(int max_con) {
+ return sizeof(memory)+
+ sizeof(int)*hash_size_max_con(max_con)+
+ sizeof(int)*max_con+
+ sizeof(ProxyRemapBlock)*max_con;
+}
+
+void proxyInitMem(void* data, int max_con) {
+ // Init m:
+ memory* m=Memory(data);
+ m->max_con=max_con;
+ m->cur_con=0;
+ m->hash_size=hash_size_max_con(max_con);
+
+ {
+ // Get pointers:
+ int* hash_table=Hash_table(data);
+ int* next=Next(data);
+ int i;
+
+ // Init the hash table:
+ for(i=0; i<m->hash_size; i++) hash_table[i]=-1;
+
+ // Init the free-list
+ for(i=0; i<m->max_con; i++) next[i]=i+1;
+ m->free_first=0;
+ }
+}
+
+int proxyGetCurConn(void* data) {
+ return Memory(data)->cur_con;
+}
+
+int proxyGetMaxConn(void* data) {
+ return Memory(data)->max_con;
+}
+
+ProxyRemapBlock* proxyLookup(void* data, unsigned ipaddr, unsigned short port, char proto) {
+ memory* m=Memory(data);
+ int* hash_table=Hash_table(m);
+ int* next=Next(m);
+ ProxyRemapBlock* info=Info(m);
+ int i;
+
+ for(i=hash_table[hash_fnc(m,ipaddr,port,proto)]; i!=-1; i=next[i]) {
+ if(info[i].proto==proto &&
+ info[i].sport==port &&
+ info[i].saddr==ipaddr) return &info[i];
+ }
+
+ return 0;
+}
+
+int proxyConsumeBlock(void* data, ProxyRemapBlock* blk) {
+ memory* m=Memory(data);
+ int* hash_table=Hash_table(m);
+ int* next=Next(m);
+ ProxyRemapBlock* info=Info(m);
+ int hash=hash_fnc(m,blk->saddr,blk->sport,blk->proto);
+ int foo;
+
+ if(blk->open) {
+ if(m->cur_con == m->max_con) return -1;
+
+ // Insert the block at a free entry:
+ info[m->free_first]=*blk;
+ m->cur_con++;
+
+ foo=next[m->free_first];
+
+ // And insert it in the hash tabel:
+ next[m->free_first]=hash_table[hash];
+ hash_table[hash]=m->free_first;
+ m->free_first=foo;
+ } else {
+ int* toupdate;
+
+ // Find the block
+ for(toupdate=&hash_table[hash];
+ *toupdate!=-1;
+ toupdate=&next[*toupdate]) {
+ if(info[*toupdate].proto==blk->proto &&
+ info[*toupdate].sport==blk->sport &&
+ info[*toupdate].saddr==blk->saddr) break;
+ }
+ if(*toupdate==-1) return -1;
+
+ foo=*toupdate;
+
+ // Delete it from the hashing list:
+ *toupdate=next[*toupdate];
+
+ // And put it on the free list:
+ next[foo]=m->free_first;
+ m->free_first=foo;
+
+ m->cur_con--;
+ }
+
+ return 0;
+}
diff -urN linux-2.6.9.orig/net/sched/proxydict.h linux-2.6.9/net/sched/proxydict.h
--- linux-2.6.9.orig/net/sched/proxydict.h 1970-01-01 01:00:00.000000000 +0100
+++ linux-2.6.9/net/sched/proxydict.h 2004-12-20 12:32:26.000000000 +0100
@@ -0,0 +1,32 @@
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*--------------------------------------------------------------------------
+This is common code for for handling the tabels containing information about
+which proxyserver connections are associated with which machines..
+*/
+
+// Returns the number of bytes that should be available in the area
+// maintained by this module given the maximal number of concurrent
+// connections.
+int proxyGetMemSize(int max_connections);
+
+// Initializes a memory area to use. There must be as many bytes
+// available as returned by getMemSize.
+void proxyInitMem(void* data, int max_connections);
+
+// Queries:
+int proxyGetCurConn(void* data); // Returns current number of connections
+int proxyMaxCurConn(void* data); // Returns maximal number of connections
+
+// This is called to open and close conenctions. Returns -1 if
+// a protocol error occores (i.e.: If it is discovered)
+int proxyConsumeBlock(void* data, ProxyRemapBlock*);
+
+// Returns the RemapBlock associated with this connection or 0:
+ProxyRemapBlock* proxyLookup(void* data, unsigned ipaddr, unsigned short port, char proto);
+
+#ifdef __cplusplus
+}
+#endif
diff -urN linux-2.6.9.orig/net/sched/proxyremap.h linux-2.6.9/net/sched/proxyremap.h
--- linux-2.6.9.orig/net/sched/proxyremap.h 1970-01-01 01:00:00.000000000 +0100
+++ linux-2.6.9/net/sched/proxyremap.h 2004-12-20 12:32:26.000000000 +0100
@@ -0,0 +1,33 @@
+#ifndef PROXYREMAP_H
+#define PROXYREMAP_H
+
+// This describes the information that is written in proxyremap.log and which
+// are used in the communication between proxyremapserver and proxyremapclient.
+// Everything is in network order.
+
+// First this header is send:
+#define PROXY_WELCOME_LINE "ProxyRemap 1.02. This is a binary protocol.\r\n"
+
+// Then this block is send every time a connection is opened or closed.
+// Note how it is alligned to use small space usage - arrays of this
+// structure are saved in many places.
+typedef struct {
+ // Server endpoint of connection:
+ unsigned saddr;
+ unsigned short sport;
+
+ // IP protocol for this connection (typically udp or tcp):
+ unsigned char proto;
+
+ // Is the connection opened or closed?
+ unsigned char open;
+
+ // Client the packets should be accounted to:
+ unsigned caddr;
+ unsigned char macaddr[6]; // Might be 0.
+
+ // An informal two-charecter code from the proxyserver. Used for debugging.
+ char proxyinfo[2];
+} ProxyRemapBlock;
+
+#endif
diff -urN linux-2.6.9.orig/net/sched/sch_wrr.c linux-2.6.9/net/sched/sch_wrr.c
--- linux-2.6.9.orig/net/sched/sch_wrr.c 1970-01-01 01:00:00.000000000 +0100
+++ linux-2.6.9/net/sched/sch_wrr.c 2004-12-20 12:42:12.000000000 +0100
@@ -0,0 +1,1297 @@
+/*-----------------------------------------------------------------------------
+Weighted Round Robin scheduler.
+
+Written by Christian Worm Mortensen, cworm@it-c.dk.
+
+Introduction
+============
+This module implements a weighted round robin queue with build-in classifier.
+The classifier currently map each MAC or IP address (configurable either MAC
+or IP and either source or destination) to different classes. Each such class
+is called a band. Whan using MAC addresses only bridged packets can be
+classified other packets go to a default MAC address.
+
+Each band has a weight value, where 0<weight<=1. The bandwidth each band
+get is proportional to the weight as can be deduced from the next section.
+
+
+The queue
+=========
+Each band has a penalty value. Bands having something to sent are kept in
+a heap according to this value. The band with the lowest penalty value
+is in the root of the heap. The penalty value is a 128 bit number. Initially
+no bands are in the heap.
+
+Two global 64 bit values counter_low_penal and couter_high_penal are initialized
+to 0 and to 2^63 respectively.
+
+Enqueing:
+ The packet is inserted in the queue for the band it belongs to. If the band
+ is not in the heap it is inserted into it. In this case, the upper 64 bits
+ of its penalty value is set to the same as for the root-band of the heap.
+ If the heap is empty 0 is used. The lower 64 bit is set to couter_low_penal
+ and couter_low_penal is incremented by 1.
+
+Dequing:
+ If the heap is empty we have nothing to send.
+
+ If the root band has a non-empty queue a packet is dequeued from that.
+ The upper 64 bit of the penalty value of the band is incremented by the
+ packet size divided with the weight of the band. The lower 64 bit is set to
+ couter_high_penal and couter_high_penal is incremented by 1.
+
+ If the root element for some reason has an empty queue it is removed from
+ the heap and we try to dequeue again.
+
+The effect of the heap and the upper 64 bit of the penalty values is to
+implement a weighted round robin queue. The effect of counter_low_penal,
+counter_high_penal and the lower 64 bit of the penalty value is primarily to
+stabilize the queue and to give better quality of service to machines only
+sending a packet now and then. For example machines which have a single
+interactive connection such as telnet or simple text chatting.
+
+
+Setting weight
+==============
+The weight value can be changed dynamically by the queue itself. The weight
+value and how it is changed is described by the two members weight1 and
+weight2 which has type tc_wrr_class_weight and which are in each class. And
+by the two integer value members of the qdisc called penalfact1 and penalfact2.
+The structure is defined as:
+
+ struct tc_wrr_class_weight {
+ // All are represented as parts of (2^64-1).
+ __u64 val; // Current value (0 is not valid)
+ __u64 decr; // Value pr bytes (2^64-1 is not valid)
+ __u64 incr; // Value pr seconds (2^64-1 is not valid)
+ __u64 min; // Minimal value (0 is not valid)
+ __u64 max; // Minimal value (0 is not valid)
+
+ // The time where the above information was correct:
+ time_t tim;
+ };
+
+The weight value used by the dequeue operations is calculated as
+weight1.val*weight2.val. weight1 and weight2 and handled independently and in the
+same way as will be described now.
+
+Every second, the val parameter is incremented by incr.
+
+Every time a packet is transmitted the value is increment by decr times
+the packet size. Depending on the value of the weight_mode parameter it
+is also mulitplied with other numbers. This makes it possible to give
+penalty to machines transferring much data.
+
+-----------------------------------------------------------------------------*/
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <asm/uaccess.h>
+#include <asm/system.h>
+#include <asm/bitops.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/vmalloc.h>
+#include <linux/sched.h>
+#include <linux/string.h>
+#include <linux/mm.h>
+#include <linux/socket.h>
+#include <linux/sockios.h>
+#include <linux/in.h>
+#include <linux/errno.h>
+#include <linux/interrupt.h>
+#include <linux/if_ether.h>
+#include <linux/inet.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/notifier.h>
+#include <net/ip.h>
+#include <net/route.h>
+#include <linux/skbuff.h>
+#include <net/sock.h>
+#include <net/pkt_sched.h>
+
+#include <linux/if_arp.h>
+#include <linux/version.h>
+
+// There seems to be problems when calling functions from userspace when
+// using vmalloc and vfree.
+//#define my_malloc(size) vmalloc(size)
+//#define my_free(ptr) vfree(ptr)
+#define my_malloc(size) kmalloc(size,GFP_KERNEL)
+#define my_free(ptr) kfree(ptr)
+
+#define LOCK_START sch_tree_lock(sch);
+#define LOCK_END sch_tree_unlock(sch);
+#define ENQUEUE_SUCCESS 0
+#define ENQUEUE_FAIL NET_XMIT_DROP
+#ifdef CONFIG_IP_NF_CONNTRACK
+ #include <linux/netfilter_ipv4/ip_conntrack.h>
+ #define MASQ_SUPPORT
+#endif
+
+#include "proxydict.c"
+
+// The penalty (priority) type:
+typedef u64 penalty_base_t;
+#define penalty_base_t_max ((penalty_base_t)-1)
+typedef struct penalty_t {
+ penalty_base_t ms;
+ penalty_base_t ls;
+} penalty_t;
+#define penalty_leq(a,b) (a.ms<b.ms || (a.ms==b.ms && a.ls<=b.ls))
+#define penalty_le(a,b) (a.ms<b.ms || (a.ms==b.ms && a.ls<b.ls))
+static penalty_t penalty_max={penalty_base_t_max,penalty_base_t_max};
+
+//-----------------------------------------------------------------------------
+// A generel heap.
+
+struct heap;
+struct heap_element;
+
+// Initializes an empty heap:
+// he: A pointer to an unintialized heap structure identifying the heap
+// size: Maximal number of elements the heap can contain
+// poll: An array of size "size" used by the heap.
+static void heap_init(struct heap* he,int size, struct heap_element* poll);
+
+// Each element in the heap is identified by a user-assigned id which
+// should be a non negative integer less than the size argument
+// given to heap_init.
+static void heap_insert(struct heap*, int id, penalty_t);
+static void heap_remove(struct heap*, int id);
+static void heap_set_penalty(struct heap*, int id, penalty_t);
+
+// Retreviewing information:
+static char heap_empty(struct heap*); // Heap empty?
+static char heap_contains(struct heap*, int id); // Does heap contain
+ // the given id?
+static int heap_root(struct heap*); // Returns the id of the root
+static penalty_t heap_get_penalty(struct heap*, int id); // Returns penaly
+ // of root node
+
+//--------------------
+// Heap implementation
+
+struct heap_element {
+ penalty_t penalty;
+ int id; // The user-assigned id of this element
+ int id2idx; // Maps from user-assigned ids to indices in root_1
+};
+
+struct heap {
+ struct heap_element* root_1;
+ int elements;
+};
+
+// Heap implementation:
+static void heap_init(struct heap* h, int size, struct heap_element* poll) {
+ int i;
+
+ h->elements=0;
+ h->root_1=poll-1;
+
+ for(i=0; i<size; i++) poll[i].id2idx=0;
+};
+
+static char heap_empty(struct heap* h) {
+ return h->elements==0;
+}
+
+static char heap_contains(struct heap* h, int id) {
+ return h->root_1[id+1].id2idx!=0;
+}
+
+static int heap_root(struct heap* h) {
+ return h->root_1[1].id;
+}
+
+static penalty_t heap_get_penalty(struct heap* h, int id) {
+ return h->root_1[ h->root_1[id+1].id2idx ].penalty;
+}
+
+static void heap_penalty_changed_internal(struct heap* h,int idx);
+
+static void heap_set_penalty(struct heap* h, int id, penalty_t p) {
+ int idx=h->root_1[id+1].id2idx;
+ h->root_1[idx].penalty=p;
+ heap_penalty_changed_internal(h,idx);
+}
+
+static void heap_insert(struct heap* h, int id, penalty_t p) {
+ // Insert at the end of the heap:
+ h->elements++;
+ h->root_1[h->elements].id=id;
+ h->root_1[h->elements].penalty=p;
+ h->root_1[id+1].id2idx=h->elements;
+
+ // And put it in the right position:
+ heap_penalty_changed_internal(h,h->elements);
+}
+
+static void heap_remove(struct heap* h, int id) {
+ int idx=h->root_1[id+1].id2idx;
+ int mvid;
+ h->root_1[id+1].id2idx=0;
+
+ if(h->elements==idx) { h->elements--; return; }
+
+ mvid=h->root_1[h->elements].id;
+ h->root_1[idx].id=mvid;
+ h->root_1[idx].penalty=h->root_1[h->elements].penalty;
+ h->root_1[mvid+1].id2idx=idx;
+
+ h->elements--;
+ heap_penalty_changed_internal(h,idx);
+}
+
+static void heap_swap(struct heap* h, int idx0, int idx1) {
+ penalty_t tmp_p;
+ int tmp_id;
+ int id0,id1;
+
+ // Simple content:
+ tmp_p=h->root_1[idx0].penalty;
+ tmp_id=h->root_1[idx0].id;
+ h->root_1[idx0].penalty=h->root_1[idx1].penalty;
+ h->root_1[idx0].id=h->root_1[idx1].id;
+ h->root_1[idx1].penalty=tmp_p;
+ h->root_1[idx1].id=tmp_id;
+
+ // Update reverse pointers:
+ id0=h->root_1[idx0].id;
+ id1=h->root_1[idx1].id;
+ h->root_1[id0+1].id2idx=idx0;
+ h->root_1[id1+1].id2idx=idx1;
+}
+
+static void heap_penalty_changed_internal(struct heap* h,int cur) {
+ if(cur==1 || penalty_leq(h->root_1[cur>>1].penalty,h->root_1[cur].penalty)) {
+ // We are in heap order upwards - so we should move the element down
+ for(;;) {
+ int nxt0=cur<<1;
+ int nxt1=nxt0+1;
+ penalty_t pen_c=h->root_1[cur].penalty;
+ penalty_t pen_0=nxt0<=h->elements ? h->root_1[nxt0].penalty : penalty_max;
+ penalty_t pen_1=nxt1<=h->elements ? h->root_1[nxt1].penalty : penalty_max;
+
+ if(penalty_le(pen_0,pen_c) && penalty_leq(pen_0,pen_1)) {
+ // Swap with child 0:
+ heap_swap(h,cur,nxt0);
+ cur=nxt0;
+ } else if(penalty_le(pen_1,pen_c)) {
+ // Swap with child 1:
+ heap_swap(h,cur,nxt1);
+ cur=nxt1;
+ } else {
+ // Heap in heap order:
+ return;
+ }
+ }
+ } else {
+ // We are not in heap order upwards (and thus we must be it downwards).
+ // We move up:
+ while(cur!=1) { // While not root
+ int nxt=cur>>1;
+ if(penalty_leq(h->root_1[nxt].penalty,h->root_1[cur].penalty)) return;
+ heap_swap(h,cur,nxt);
+ cur=nxt;
+ }
+ }
+};
+
+//-----------------------------------------------------------------------------
+// Classification based on MAC or IP adresses. Note that of historical reason
+// these are prefixed with mac_ since originally only MAC bases classification
+// was supported.
+//
+// This code should be in a separate filter module - but it isn't.
+
+// Interface:
+
+struct mac_head;
+
+// Initialices/destroys the structure we maintain.
+// Returns -1 on error
+static int mac_init(struct mac_head*, int max_macs, char srcaddr,
+ char usemac, char usemasq, void* proxyremap);
+static void mac_done(struct mac_head*);
+static void mac_reset(struct mac_head*);
+
+// Classify a packet. Returns a number n where 0<=n<max_macs. Or -1 if
+// the packet should be dropped.
+static int mac_classify(struct mac_head*, struct sk_buff *skb);
+
+//-------------
+// Implementation:
+
+struct mac_addr {
+ unsigned char addr[ETH_ALEN]; // Address of this band (last two are 0 on IP)
+ unsigned long lastused; // Last time a packet was encountered
+ int class; // Classid of this band (0<=classid<max_macs)
+};
+
+static int mac_compare(const void* a, const void* b) {
+ return memcmp(a,b,ETH_ALEN);
+}
+
+struct mac_head {
+ int mac_max; // Maximal number of MAC addresses/classes allowed
+ int mac_cur; // Current number of MAC addresses/classes
+ int mac_reused; // Number of times we have reused a class with a new
+ // address.
+ u64 incr_time;
+ char srcaddr; // True if we classify on the source address of packets,
+ // else we use destination address.
+ char usemac; // If true we use mac, else we use IP
+ char usemasq; // If true we try to demasqgrade
+ struct mac_addr* macs; // Allocated mac_max elements, used max_cur
+ char* cls2mac; // Mapping from classnumbers to addresses -
+ // there is 6 bytes in each entry
+
+ void* proxyremap; // Information on proxy remapping of data or 0
+};
+
+// This is as the standard C library function with the same name:
+static const void* bsearch(const void* key, const void* base, int nmemb,
+ size_t size,
+ int (*compare)(const void*, const void*)) {
+ int m_idx;
+ const void* m_ptr;
+ int i;
+
+ if(nmemb<=0) return 0;
+
+ m_idx=nmemb>>1;
+ m_ptr=((const char*)base)+m_idx*size;
+
+ i=compare(key,m_ptr);
+ if(i<0) // key is less
+ return bsearch(key,base,m_idx,size,compare);
+ else if(i>0)
+ return bsearch(key,((const char*)m_ptr)+size,nmemb-m_idx-1,size,compare);
+
+ return m_ptr;
+}
+
+static int mac_init(struct mac_head* h, int max_macs, char srcaddr,
+ char usemac, char usemasq,void* proxyremap) {
+ h->mac_cur=0;
+ h->mac_reused=0;
+ h->incr_time=0;
+ h->srcaddr=srcaddr;
+ h->usemac=usemac;
+ h->usemasq=usemasq;
+ h->mac_max=max_macs;
+ h->proxyremap=proxyremap;
+
+ h->macs=(struct mac_addr*)
+ my_malloc( sizeof(struct mac_addr)*max_macs);
+ h->cls2mac=(char*)my_malloc( 6*max_macs);
+ if(!h->macs || !h->cls2mac) {
+ if(h->macs) my_free(h->macs);
+ if(h->cls2mac) my_free(h->cls2mac);
+ return -1;
+ }
+ return 0;
+}
+
+static void mac_done(struct mac_head* h) {
+ my_free(h->macs);
+ my_free(h->cls2mac);
+}
+
+static void mac_reset(struct mac_head* h) {
+ h->mac_cur=0;
+ h->mac_reused=0;
+ h->incr_time=0;
+}
+
+static int lookup_mac(struct mac_head* h, unsigned char* addr) {
+ int i;
+ int class;
+
+ // First try to find the address in the table:
+ struct mac_addr* m=(struct mac_addr*)
+ bsearch(addr,h->macs,h->mac_cur,sizeof(struct mac_addr),mac_compare);
+ if(m) {
+ // Found:
+ m->lastused=h->incr_time++;
+ return m->class;
+ }
+
+ // Okay - the MAC adress was not in table
+ if(h->mac_cur==h->mac_max) {
+ // And the table is full - delete the oldest entry:
+
+ // Find the oldest entry:
+ int lowidx=0;
+ int i;
+ for(i=1; i<h->mac_cur; i++)
+ if(h->macs[i].lastused < h->macs[lowidx].lastused) lowidx=i;
+
+ class=h->macs[lowidx].class;
+
+ // And delete it:
+ memmove(&h->macs[lowidx],&h->macs[lowidx+1],
+ (h->mac_cur-lowidx-1)*sizeof(struct mac_addr));
+ h->mac_reused++;
+ h->mac_cur--;
+ } else {
+ class=h->mac_cur;
+ }
+
+ // The table is now not full - find the position we should put the address in:
+ for(i=0; i<h->mac_cur; i++) if(mac_compare(addr,&h->macs[i])<0) break;
+
+ // We should insert at position i:
+ memmove(&h->macs[i+1],&h->macs[i],(h->mac_cur-i)*sizeof(struct mac_addr));
+ m=&h->macs[i];
+ memcpy(m->addr,addr,ETH_ALEN);
+ m->lastused=h->incr_time++;
+ m->class=class;
+ h->mac_cur++;
+
+ // Finally update the cls2mac variabel:
+ memcpy(h->cls2mac+ETH_ALEN*class,addr,ETH_ALEN);
+
+ return m->class;
+}
+
+int valid_ip_checksum(struct iphdr* ip, int size) {
+ __u16 header_len=ip->ihl<<2;
+ __u16 c=0;
+ __u16* ipu=(u16*)ip;
+ int a;
+
+ // We require 4 bytes in the packet since we access the port numbers:
+ if((size<header_len) || size<sizeof(struct iphdr)+4) return 0;
+
+ for(a=0; a<(header_len>>1); a++, ipu++) {
+ if(a!=5) { // If not the checksum field
+ __u16 oldc=c;
+ c+=(*ipu);
+ if(c<oldc) c++;
+ }
+ }
+
+ return ip->check==(__u16)~c;
+}
+
+static int mac_classify(struct mac_head* head, struct sk_buff *skb)
+{
+ // We set this to the address we map to. In case we map to an IP
+ // address the last two entries are set to 0.
+ unsigned char addr[ETH_ALEN];
+
+
+ // This is the size of the network part of the packet, I think:
+ int size=((char*)skb->data+skb->len)-((char*)skb->nh.iph);
+
+ // Set a default value for the address:
+ memset(addr,0,ETH_ALEN);
+
+ // Accept IP-ARP traffic with big-enough packets:
+ if(ntohs(skb->protocol)==ETH_P_ARP &&
+ ntohs(skb->nh.arph->ar_pro)==ETH_P_IP) {
+ // Map all ARP trafic to a default adress to make sure
+ // it goes through
+ } else if ((ntohs(skb->protocol)==ETH_P_IP) &&
+ valid_ip_checksum(skb->nh.iph,size)) {
+ // Accept IP packets which have correct checksum.
+
+ // This is the IP header:
+ struct iphdr* iph=skb->nh.iph;
+
+ // And this is the port numbers:
+ const __u16 *portp = (__u16 *)&(((char *)iph)[iph->ihl*4]);
+ __u16 sport=portp[0];
+ __u16 dport=portp[1];
+
+ // We will set this to the IP address of the packet that should be
+ // accounted to:
+ unsigned ipaddr;
+
+ // Used below:
+ ProxyRemapBlock* prm;
+
+ // Set ipaddr:
+ if(head->srcaddr)
+ ipaddr=iph->saddr;
+ else
+ ipaddr=iph->daddr;
+
+#ifdef MASQ_SUPPORT
+ // Update ipaddr if packet is masqgraded:
+ if(head->usemasq) {
+ // Thanks to Rusty Russell for help with the following code:
+ enum ip_conntrack_info ctinfo;
+ struct ip_conntrack *ct;
+ ct = ip_conntrack_get(skb, &ctinfo);
+ if (ct) {
+ if(head->srcaddr)
+ ipaddr=ct->tuplehash[CTINFO2DIR(ctinfo)].tuple.src.ip;
+ else
+ ipaddr=ct->tuplehash[CTINFO2DIR(ctinfo)].tuple.dst.ip;
+ }
+ }
+#endif
+
+ // Set prm based on ipaddr:
+ prm=0;
+ if(head->proxyremap) {
+ if(head->srcaddr) {
+ prm=proxyLookup(head->proxyremap,ipaddr,sport,skb->nh.iph->protocol);
+ } else {
+ prm=proxyLookup(head->proxyremap,ipaddr,dport,skb->nh.iph->protocol);
+ }
+ }
+
+ // And finally set addr to the address:
+ memset(addr,0,ETH_ALEN);
+ if(prm) {
+ // This package should be remapped:
+ if(head->usemac)
+ memcpy(addr,prm->macaddr,ETH_ALEN);
+ else {
+ memcpy(addr,&prm->caddr,sizeof(unsigned));
+ }
+ } else {
+ // This packet should not be remapped:
+ if(head->usemac) {
+ // We should find MAC address of packet.
+ // Unfortunatly, this is not always available.
+ // On bridged packets it always is, however..
+ } else {
+ memcpy(addr,&ipaddr,4);
+ }
+ }
+ } else {
+ // All other traffic is dropped - this ensures that packets
+ // we consider probably have valid addresses so we don't
+ // get to many strange addresses into our table. And that we
+ // don't use bandwidth on strange packets..
+ return -1;
+ }
+
+ return lookup_mac(head,addr);
+}
+
+//-----------------------------------------------------------------------------
+// The qdisc itself
+
+// Pr-class information.
+struct wrrc_sched_data {
+ struct Qdisc* que; // The queue for this class
+ struct tc_wrr_class_modf class_modf; // Information about the class.
+
+ // For classes in the heap this is the priority value priosum
+ // was updated with for this class:
+ u64 priosum_val;
+};
+
+// Pr-qdisc information:
+struct wrr_sched_data
+{
+ // A heap containing all the bands that will send something
+ struct heap h;
+ struct heap_element* poll; // bandc elements
+
+ // The sum of the prioities of the elements in the heap where
+ // a priority of 1 is saved as 2^32
+ u64 priosum;
+
+ // A class for each band
+ struct wrrc_sched_data* bands; // bandc elements
+
+ // Information maintained by the proxydict module of 0 if we
+ // have no proxy remapping
+ void* proxydict;
+
+ // Always incrementning counters, we always have that any value of
+ // counter_low_penal < any value of counter_high_penal.
+ penalty_base_t counter_low_penal;
+ penalty_base_t counter_high_penal;
+
+ // Penalty updating:
+ struct tc_wrr_qdisc_modf qdisc_modf;
+
+ // Statistics:
+ int packets_requed;
+
+ // The filter:
+ struct mac_head filter;
+ int bandc; // Number of bands
+};
+
+// Priority handling.
+// weight is in interval [0..2^32]
+// priosum has whole numbers in the upper and fragments in the lower 32 bits.
+static void weight_transmit(struct tc_wrr_class_weight* p,
+ struct tc_wrr_qdisc_weight q,
+ unsigned heapsize,
+ u64 priosum, u64 weight,
+ unsigned size) {
+
+ unsigned long now=jiffies/HZ;
+
+ // Penalty for transmitting:
+ u64 change,old;
+ u32 divisor;
+
+ change=0;
+ switch(q.weight_mode) {
+ case 1: change=p->decr*size; break;
+ case 2: change=p->decr*size*heapsize; break;
+ case 3: // Note: 64 bit division is not always available..
+ divisor=(u32)(weight>>16);
+ if(divisor<=0) divisor=1;
+ change=p->decr*size*(((u32)(priosum>>16))/divisor); break;
+ }
+ old=p->val;
+ p->val-=change;
+ if(p->val>old || p->val<p->min) p->val=p->min;
+
+ // Credit for time went:
+ change=(now-p->tim)*p->incr;
+ p->tim=now;
+ old=p->val;
+ p->val+=change;
+ if(p->val<old || p->val>p->max) p->val=p->max;
+}
+
+static void weight_setdefault(struct tc_wrr_class_weight* p) {
+ p->val=(u64)-1;
+ p->decr=0;
+ p->incr=0;
+ p->min=(u64)-1;
+ p->max=(u64)-1;
+ p->tim=jiffies/HZ;
+}
+
+static void weight_setvalue(struct tc_wrr_class_weight* dst,
+ struct tc_wrr_class_weight* src) {
+ if(src->val!=0) {
+ dst->val=src->val;
+ dst->tim=jiffies/HZ;
+ }
+ if(src->min!=0) dst->min=src->min;
+ if(src->max!=0) dst->max=src->max;
+ if(src->decr!=((u64)-1)) dst->decr=src->decr;
+ if(src->incr!=((u64)-1)) dst->incr=src->incr;
+ if(dst->val<dst->min) dst->val=dst->min;
+ if(dst->val>dst->max) dst->val=dst->max;
+}
+
+static void wrr_destroy(struct Qdisc *sch)
+{
+ struct wrr_sched_data *q=qdisc_priv(sch);
+ int i;
+
+ // Destroy our filter:
+ mac_done(&q->filter);
+
+ // Destroy all our childre ques:
+ for(i=0; i<q->bandc; i++)
+ qdisc_destroy(q->bands[i].que);
+
+ // And free memory:
+ my_free(q->bands);
+ my_free(q->poll);
+ if(q->proxydict) my_free(q->proxydict);
+}
+
+static int wrr_init(struct Qdisc *sch, struct rtattr *opt)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ int i,maciniterr;
+ char crterr;
+ struct tc_wrr_qdisc_crt *qopt;
+
+ // Parse options:
+ if (!opt) return -EINVAL; // Options must be specified
+ if (opt->rta_len < RTA_LENGTH(sizeof(*qopt))) return -EINVAL;
+ qopt = RTA_DATA(opt);
+
+ if(qopt->bands_max>8192 || qopt->bands_max<2) {
+ // More than 8192 queues or less than 2? That cannot be true - it must be
+ // an error...
+ return -EINVAL;
+ }
+
+ if(qopt->proxy_maxconn<0 || qopt->proxy_maxconn>20000) {
+ // More than this number of maximal concurrent connections is unrealistic
+ return -EINVAL;
+ }
+
+#ifndef MASQ_SUPPORT
+ if(qopt->usemasq) {
+ return -ENOSYS;
+ }
+#endif
+
+ if(qopt->usemac) { // Not supported - please fix this!
+ return -ENOSYS;
+ }
+
+ q->bandc=qopt->bands_max;
+ q->qdisc_modf=qopt->qdisc_modf;
+
+ // Create structures:
+ q->poll=(struct heap_element*)
+ my_malloc( sizeof(struct heap_element)*q->bandc);
+ q->bands=(struct wrrc_sched_data*)
+ my_malloc( sizeof(struct wrrc_sched_data)*q->bandc);
+
+ if(qopt->proxy_maxconn>0) {
+ q->proxydict=my_malloc(proxyGetMemSize(qopt->proxy_maxconn));
+ } else {
+ q->proxydict=0;
+ }
+
+ // Init mac module:
+ maciniterr=mac_init(&q->filter,qopt->bands_max,qopt->srcaddr,
+ qopt->usemac,qopt->usemasq,q->proxydict);
+
+ // See if we got the memory we wanted:
+ if(!q->poll || !q->bands ||
+ (qopt->proxy_maxconn>0 && !q->proxydict) || maciniterr<0) {
+ if(q->poll) my_free(q->poll);
+ if(q->bands) my_free(q->bands);
+ if(q->proxydict) my_free(q->proxydict);
+ if(maciniterr>=0) mac_done(&q->filter);
+ return -ENOMEM;
+ }
+
+ // Initialize proxy:
+ if(q->proxydict) {
+ proxyInitMem(q->proxydict,qopt->proxy_maxconn);
+ }
+
+ // Initialize values:
+ q->counter_low_penal=0;
+ q->counter_high_penal=penalty_base_t_max>>1;
+ q->packets_requed=0;
+
+ // Initialize empty heap:
+ heap_init(&q->h,q->bandc,q->poll);
+ q->priosum=0;
+
+ // Initialize each band:
+ crterr=0;
+ for (i=0; i<q->bandc; i++) {
+ weight_setdefault(&q->bands[i].class_modf.weight1);
+ weight_setdefault(&q->bands[i].class_modf.weight2);
+ if(!crterr) {
+ struct Qdisc *child=qdisc_create_dflt(sch->dev, &pfifo_qdisc_ops);
+ if(child)
+ q->bands[i].que = child;
+ else {
+ // Queue couldn't be created :-(
+ crterr=1;
+ }
+ }
+ if(crterr) q->bands[i].que = &noop_qdisc;
+ }
+
+ if(crterr) {
+ // Destroy again:
+ wrr_destroy(sch);
+ return -ENOMEM;
+ }
+
+ return 0;
+}
+
+static void wrr_reset(struct Qdisc* sch)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ int i;
+
+ // Reset own values:
+ q->counter_low_penal=0;
+ q->counter_high_penal=penalty_base_t_max>>1;
+ q->packets_requed=0;
+
+ // Reset filter:
+ mac_reset(&q->filter);
+
+ // Reinitialize heap:
+ heap_init(&q->h,q->bandc,q->poll);
+ q->priosum=0;
+
+ // Reset all bands:
+ for (i=0; i<q->bandc; i++) {
+ weight_setdefault(&q->bands[i].class_modf.weight1);
+ weight_setdefault(&q->bands[i].class_modf.weight2);
+ qdisc_reset(q->bands[i].que);
+ }
+
+ // Reset proxy remapping information:
+ if(q->proxydict)
+ proxyInitMem(q->proxydict,proxyGetMaxConn(q->proxydict));
+}
+
+static int wrr_enqueue(struct sk_buff *skb, struct Qdisc* sch)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ int retvalue=ENQUEUE_FAIL;
+
+ // The packet is in skb.
+ int band=mac_classify(&q->filter,skb);
+
+ if(band>=0) {
+ // Enque packet for this band:
+ struct Qdisc* qdisc = q->bands[band].que;
+
+ if ((retvalue=qdisc->enqueue(skb, qdisc)) == ENQUEUE_SUCCESS) {
+ // Successfull
+ sch->stats.bytes += skb->len;
+ sch->stats.packets++;
+ sch->q.qlen++;
+
+ // Insert band into heap if not already there:
+ if(!heap_contains(&q->h,band)) {
+ penalty_t p;
+ if(!heap_empty(&q->h))
+ p.ms=heap_get_penalty(&q->h,heap_root(&q->h)).ms;
+ else
+ p.ms=0;
+ p.ls=q->counter_low_penal++;
+ heap_insert(&q->h,band,p);
+ q->bands[band].priosum_val=
+ ((q->bands[band].class_modf.weight1.val>>48)+1)*
+ ((q->bands[band].class_modf.weight2.val>>48)+1);
+ q->priosum+=q->bands[band].priosum_val;
+ }
+ }
+ } else {
+ // If we decide not to enque it seems like we also need to free the packet:
+ kfree_skb(skb);
+ }
+
+ if(retvalue!=ENQUEUE_SUCCESS) {
+ // Packet not enqued:
+ sch->stats.drops++;
+ }
+
+ return retvalue;
+}
+
+static struct sk_buff *wrr_dequeue(struct Qdisc* sch)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ struct sk_buff* skb;
+ int band;
+ u64 weight,priosum;
+ struct wrrc_sched_data* b;
+
+ // Return if heap is empty:
+ if(heap_empty(&q->h)) return 0;
+
+ // Find root element:
+ band=heap_root(&q->h);
+
+ // Find priority of this element in interval [1;2^32]
+ b=&q->bands[band];
+ weight=((b->class_modf.weight1.val>>48)+1)*
+ ((b->class_modf.weight2.val>>48)+1); //weight is in interval [1;2^32]
+ priosum=q->priosum;
+ q->priosum-=q->bands[band].priosum_val;
+
+ // Deque the packet from the root:
+ skb=q->bands[band].que->dequeue(q->bands[band].que);
+
+ if(skb) {
+ // There was a packet in this que.
+ unsigned adjlen;
+ penalty_t p;
+
+ // Find length of packet adjusted with priority:
+ adjlen=(u32)(weight>>(32-16));
+ if(adjlen==0) adjlen=1;
+ adjlen=(skb->len<<16)/adjlen;
+
+ // Update penalty information for this class:
+ weight_transmit(&b->class_modf.weight1,q->qdisc_modf.weight1,q->h.elements,priosum,weight,skb->len);
+ weight_transmit(&b->class_modf.weight2,q->qdisc_modf.weight2,q->h.elements,priosum,weight,skb->len);
+ q->bands[band].priosum_val=((b->class_modf.weight1.val>>48)+1)*
+ ((b->class_modf.weight2.val>>48)+1);
+ q->priosum+=q->bands[band].priosum_val;
+
+ // And update the class in the heap
+ p=heap_get_penalty(&q->h,band);
+ p.ms+=adjlen;
+ p.ls=q->counter_high_penal++;
+ heap_set_penalty(&q->h,band,p);
+
+ // Return packet:
+ sch->q.qlen--;
+ return skb;
+ }
+
+ // No packet - so machine should be removed from heap:
+ heap_remove(&q->h,band);
+
+ // And try again:
+ return wrr_dequeue(sch);
+}
+
+static int wrr_requeue(struct sk_buff *skb, struct Qdisc* sch)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ struct Qdisc* qdisc;
+ int ret;
+
+ // Find band we took it from:
+ int band=mac_classify(&q->filter,skb);
+ if(band<0) {
+ // Who should now free the pakcet?
+ printk(KERN_DEBUG "sch_wrr: Oops - packet requed could never have been queued.\n");
+ sch->stats.drops++;
+ return ENQUEUE_FAIL;
+ }
+
+ q->packets_requed++;
+
+ // Try to requeue it on that machine:
+ qdisc=q->bands[band].que;
+
+ if((ret=qdisc->ops->requeue(skb,qdisc))==ENQUEUE_SUCCESS) {
+ // On success:
+ sch->q.qlen++;
+
+ // We should restore priority information - but we don't
+ //
+ // p=heap_get_penalty(&q->h,band);
+ // ...
+ // heap_set_penalty(&q->h,band,p);
+
+ return ENQUEUE_SUCCESS;
+ } else {
+ sch->stats.drops++;
+ return ret;
+ }
+}
+
+static unsigned int wrr_drop(struct Qdisc* sch)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+
+ // Ugly... Drop button up in heap:
+ int i;
+
+ for(i=q->h.elements; i>=1; i--) {
+ int band=q->h.root_1[i].id;
+ if(q->bands[band].que->ops->drop(q->bands[band].que)) {
+ // On success
+ sch->q.qlen--;
+ return 1;
+ }
+ }
+
+ return 0;
+}
+
+static int wrr_dump(struct Qdisc *sch, struct sk_buff *skb)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ unsigned char *b = skb->tail;
+ struct tc_wrr_qdisc_stats opt;
+
+ opt.qdisc_crt.qdisc_modf=q->qdisc_modf;
+ opt.qdisc_crt.srcaddr=q->filter.srcaddr;
+ opt.qdisc_crt.usemac=q->filter.usemac;
+ opt.qdisc_crt.usemasq=q->filter.usemasq;
+ opt.qdisc_crt.bands_max=q->filter.mac_max;
+ opt.nodes_in_heap=q->h.elements;
+ opt.bands_cur=q->filter.mac_cur;
+ opt.bands_reused=q->filter.mac_reused;
+ opt.packets_requed=q->packets_requed;
+ opt.priosum=q->priosum;
+
+ if(q->proxydict) {
+ opt.qdisc_crt.proxy_maxconn=proxyGetMaxConn(q->proxydict);
+ opt.proxy_curconn=proxyGetCurConn(q->proxydict);
+ } else {
+ opt.qdisc_crt.proxy_maxconn=0;
+ opt.proxy_curconn=0;
+ }
+
+ RTA_PUT(skb, TCA_OPTIONS, sizeof(opt), &opt);
+ return skb->len;
+
+rtattr_failure: // seems like RTA_PUT jump to this label..
+ skb_trim(skb, b - skb->data);
+ return -1;
+}
+
+static int wrr_tune_std(struct Qdisc *sch, struct rtattr *opt)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ struct tc_wrr_qdisc_modf_std *qopt = RTA_DATA(opt);
+
+ if(opt->rta_len < RTA_LENGTH(sizeof(*qopt))) return -EINVAL;
+
+ LOCK_START
+
+ if(qopt->change_class) {
+ int idx=lookup_mac(&q->filter,qopt->addr);
+ weight_setvalue
+ (&q->bands[idx].class_modf.weight1,&qopt->class_modf.weight1);
+ weight_setvalue
+ (&q->bands[idx].class_modf.weight2,&qopt->class_modf.weight2);
+ } else {
+ if(qopt->qdisc_modf.weight1.weight_mode!=-1)
+ q->qdisc_modf.weight1.weight_mode=qopt->qdisc_modf.weight1.weight_mode;
+ if(qopt->qdisc_modf.weight2.weight_mode!=-1)
+ q->qdisc_modf.weight2.weight_mode=qopt->qdisc_modf.weight2.weight_mode;
+ }
+
+ LOCK_END
+
+ return 0;
+}
+
+static int wrr_tune_proxy(struct Qdisc *sch, struct rtattr *opt)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ struct tc_wrr_qdisc_modf_proxy *qopt = RTA_DATA(opt);
+ int i;
+
+ // Return if we are not configured with proxy support:
+ if(!q->proxydict) return -ENOSYS;
+
+ // Return if not enough data given:
+ if(opt->rta_len<RTA_LENGTH(sizeof(*qopt)) ||
+ opt->rta_len<
+ RTA_LENGTH(sizeof(*qopt)+sizeof(ProxyRemapBlock)*qopt->changec))
+ return -EINVAL;
+
+ LOCK_START;
+
+ if(qopt->reset) {
+ proxyInitMem(q->proxydict,proxyGetMaxConn(q->proxydict));
+ }
+
+ // Do all the changes:
+ for(i=0; i<qopt->changec; i++) {
+ proxyConsumeBlock(q->proxydict,&((ProxyRemapBlock*)&qopt->changes)[i]);
+ }
+
+ LOCK_END;
+
+ return 0;
+}
+
+static int wrr_tune(struct Qdisc *sch, struct rtattr *opt) {
+ if(((struct tc_wrr_qdisc_modf_std*)RTA_DATA(opt))->proxy) {
+ return wrr_tune_proxy(sch,opt);
+ } else {
+ return wrr_tune_std(sch,opt);
+ }
+}
+
+//-----------------------------------------------------------------------------
+// Classes.
+// External and internal IDs are equal. They are the band number plus 1.
+
+// Replace a class with another:
+static int wrr_graft(struct Qdisc *sch, unsigned long arg, struct Qdisc *new,
+ struct Qdisc **old)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ if(arg>q->bandc || arg==0) return -EINVAL;
+ arg--;
+
+ if (new == NULL)
+ new = &noop_qdisc;
+
+ LOCK_START
+ *old = q->bands[arg].que;
+ q->bands[arg].que = new;
+ qdisc_reset(*old);
+ LOCK_END
+
+ return 0;
+}
+
+// Returns the qdisc for a class:
+static struct Qdisc * wrr_leaf(struct Qdisc *sch, unsigned long arg)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ if(arg>q->bandc || arg==0) return NULL;
+ arg--;
+ return q->bands[arg].que;
+}
+
+static unsigned long wrr_get(struct Qdisc *sch, u32 classid)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ unsigned long band = TC_H_MIN(classid);
+ if(band>q->bandc || band==0) return 0;
+ return band;
+}
+
+static void wrr_put(struct Qdisc *q, unsigned long cl)
+{
+ return;
+}
+
+static int wrr_delete(struct Qdisc *sch, unsigned long cl)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ if(cl==0 || cl>q->bandc) return -ENOENT;
+ cl--;
+ return 0;
+}
+
+static int wrr_dump_class(struct Qdisc *sch, unsigned long cl,
+ struct sk_buff *skb, struct tcmsg *tcm)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ unsigned char *b = skb->tail;
+ struct tc_wrr_class_stats opt;
+
+ // Handle of this class:
+ tcm->tcm_handle = sch->handle|cl;
+
+ if(cl==0 || cl>q->bandc)
+ goto rtattr_failure;
+ cl--;
+
+ if(cl>=q->filter.mac_cur) {
+ // Band is unused:
+ memset(&opt,0,sizeof(opt));
+ opt.used=0;
+ } else {
+ opt.used=1;
+ opt.class_modf.weight1=q->bands[cl].class_modf.weight1;
+ opt.class_modf.weight2=q->bands[cl].class_modf.weight2;
+ weight_transmit(&opt.class_modf.weight1,q->qdisc_modf.weight1,0,0,0,0);
+ weight_transmit(&opt.class_modf.weight2,q->qdisc_modf.weight2,0,0,0,0);
+ memcpy(opt.addr,q->filter.cls2mac+cl*ETH_ALEN,ETH_ALEN);
+ opt.usemac=q->filter.usemac;
+ opt.heappos=q->h.root_1[cl+1].id2idx;
+ if(opt.heappos!=0) { // Is in heap
+ opt.penal_ls=heap_get_penalty(&q->h,cl).ls;
+ opt.penal_ms=heap_get_penalty(&q->h,cl).ms;
+ } else {
+ opt.penal_ls=0;
+ opt.penal_ms=0;
+ }
+ }
+
+ // Put quing information:
+ RTA_PUT(skb, TCA_OPTIONS, sizeof(opt), &opt);
+ return skb->len;
+
+rtattr_failure:
+ skb_trim(skb, b - skb->data);
+ return -1;
+}
+
+static int wrr_change(struct Qdisc *sch, u32 handle, u32 parent,
+ struct rtattr **tca, unsigned long *arg)
+{
+ unsigned long cl = *arg;
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ struct rtattr *opt = tca[TCA_OPTIONS-1];
+ struct tc_wrr_class_modf *copt = RTA_DATA(opt);
+
+ if(cl==0 || cl>q->bandc) return -EINVAL;
+ cl--;
+
+ if (opt->rta_len < RTA_LENGTH(sizeof(*copt))) return -EINVAL;
+
+ LOCK_START;
+
+ weight_setvalue(&q->bands[cl].class_modf.weight1,&copt->weight1);
+ weight_setvalue(&q->bands[cl].class_modf.weight2,&copt->weight2);
+
+ LOCK_END;
+
+ return 0;
+}
+
+static void wrr_walk(struct Qdisc *sch, struct qdisc_walker *arg)
+{
+ struct wrr_sched_data *q = qdisc_priv(sch);
+ int prio;
+
+ if (arg->stop) return;
+
+ for (prio = 1; prio <= q->bandc; prio++) {
+ if (arg->count < arg->skip) {
+ arg->count++;
+ continue;
+ }
+ if (arg->fn(sch, prio, arg) < 0) {
+ arg->stop = 1;
+ break;
+ }
+ arg->count++;
+ }
+}
+
+static struct tcf_proto ** wrr_find_tcf(struct Qdisc *sch, unsigned long cl)
+{
+ return NULL;
+}
+
+static unsigned long wrr_bind(struct Qdisc *sch,
+ unsigned long parent, u32 classid)
+{
+ return wrr_get(sch, classid);
+}
+
+//-----------------------------------------------------------------------------
+// General
+
+static struct Qdisc_class_ops wrr_class_ops =
+{
+ .graft = wrr_graft,
+ .leaf = wrr_leaf,
+ .get = wrr_get,
+ .put = wrr_put,
+ .change = wrr_change,
+ .delete = wrr_delete,
+ .walk = wrr_walk,
+ .tcf_chain = wrr_find_tcf,
+ .bind_tcf = wrr_bind,
+ .unbind_tcf = wrr_put,
+ .dump = wrr_dump_class,
+};
+
+static struct Qdisc_ops wrr_qdisc_ops =
+{
+ .next = NULL,
+ .cl_ops = &wrr_class_ops,
+ .id = "wrr",
+ .priv_size = sizeof(struct wrr_sched_data),
+ .enqueue = wrr_enqueue,
+ .dequeue = wrr_dequeue,
+ .requeue = wrr_requeue,
+ .drop = wrr_drop,
+ .init = wrr_init,
+ .reset = wrr_reset,
+ .destroy = wrr_destroy,
+ .change = wrr_tune,
+ .dump = wrr_dump,
+ .owner = THIS_MODULE,
+};
+
+static int __init wrr_module_init(void)
+{
+ return register_qdisc(&wrr_qdisc_ops);
+}
+
+static void __exit wrr_module_exit(void)
+{
+ unregister_qdisc(&wrr_qdisc_ops);
+}
+module_init(wrr_module_init)
+module_exit(wrr_module_exit)
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Christian Worm Mortensen");
|