11da177e4SLinus Torvalds /* 21da177e4SLinus Torvalds * INET An implementation of the TCP/IP protocol suite for the LINUX 31da177e4SLinus Torvalds * operating system. INET is implemented using the BSD Socket 41da177e4SLinus Torvalds * interface as the means of communication with the user level. 51da177e4SLinus Torvalds * 61da177e4SLinus Torvalds * IPv4 Forwarding Information Base: policy rules. 71da177e4SLinus Torvalds * 81da177e4SLinus Torvalds * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 9e1ef4bf2SThomas Graf * Thomas Graf <tgraf@suug.ch> 101da177e4SLinus Torvalds * 111da177e4SLinus Torvalds * This program is free software; you can redistribute it and/or 121da177e4SLinus Torvalds * modify it under the terms of the GNU General Public License 131da177e4SLinus Torvalds * as published by the Free Software Foundation; either version 141da177e4SLinus Torvalds * 2 of the License, or (at your option) any later version. 151da177e4SLinus Torvalds * 161da177e4SLinus Torvalds * Fixes: 171da177e4SLinus Torvalds * Rani Assaf : local_rule cannot be deleted 181da177e4SLinus Torvalds * Marc Boucher : routing by fwmark 191da177e4SLinus Torvalds */ 201da177e4SLinus Torvalds 211da177e4SLinus Torvalds #include <linux/types.h> 221da177e4SLinus Torvalds #include <linux/kernel.h> 231da177e4SLinus Torvalds #include <linux/netdevice.h> 241da177e4SLinus Torvalds #include <linux/netlink.h> 25e1ef4bf2SThomas Graf #include <linux/inetdevice.h> 261da177e4SLinus Torvalds #include <linux/init.h> 277b204afdSRobert Olsson #include <linux/list.h> 287b204afdSRobert Olsson #include <linux/rcupdate.h> 29bc3b2d7fSPaul Gortmaker #include <linux/export.h> 301da177e4SLinus Torvalds #include <net/ip.h> 311da177e4SLinus Torvalds #include <net/route.h> 321da177e4SLinus Torvalds #include <net/tcp.h> 331da177e4SLinus Torvalds #include <net/ip_fib.h> 34e1ef4bf2SThomas Graf #include <net/fib_rules.h> 351da177e4SLinus Torvalds 366a31d2a9SEric Dumazet struct fib4_rule { 37e1ef4bf2SThomas Graf struct fib_rule common; 38e1ef4bf2SThomas Graf u8 dst_len; 39e1ef4bf2SThomas Graf u8 src_len; 40e1ef4bf2SThomas Graf u8 tos; 4181f7bf6cSAl Viro __be32 src; 4281f7bf6cSAl Viro __be32 srcmask; 4381f7bf6cSAl Viro __be32 dst; 4481f7bf6cSAl Viro __be32 dstmask; 45c7066f70SPatrick McHardy #ifdef CONFIG_IP_ROUTE_CLASSID 46e1ef4bf2SThomas Graf u32 tclassid; 471da177e4SLinus Torvalds #endif 481da177e4SLinus Torvalds }; 491da177e4SLinus Torvalds 50f4530fa5SDavid S. Miller int __fib_lookup(struct net *net, struct flowi4 *flp, struct fib_result *res) 51e1ef4bf2SThomas Graf { 52e1ef4bf2SThomas Graf struct fib_lookup_arg arg = { 53e1ef4bf2SThomas Graf .result = res, 54ebc0ffaeSEric Dumazet .flags = FIB_LOOKUP_NOREF, 55e1ef4bf2SThomas Graf }; 56e1ef4bf2SThomas Graf int err; 57e1ef4bf2SThomas Graf 5822bd5b9bSDavid S. Miller err = fib_rules_lookup(net->ipv4.rules_ops, flowi4_to_flowi(flp), 0, &arg); 5985b91b03SDavid S. Miller #ifdef CONFIG_IP_ROUTE_CLASSID 6085b91b03SDavid S. Miller if (arg.rule) 6185b91b03SDavid S. Miller res->tclassid = ((struct fib4_rule *)arg.rule)->tclassid; 6285b91b03SDavid S. Miller else 6385b91b03SDavid S. Miller res->tclassid = 0; 6485b91b03SDavid S. Miller #endif 651da177e4SLinus Torvalds return err; 661da177e4SLinus Torvalds } 67f4530fa5SDavid S. Miller EXPORT_SYMBOL_GPL(__fib_lookup); 681da177e4SLinus Torvalds 698ce11e6aSAdrian Bunk static int fib4_rule_action(struct fib_rule *rule, struct flowi *flp, 708ce11e6aSAdrian Bunk int flags, struct fib_lookup_arg *arg) 71e1ef4bf2SThomas Graf { 72e1ef4bf2SThomas Graf int err = -EAGAIN; 73e1ef4bf2SThomas Graf struct fib_table *tbl; 74e1ef4bf2SThomas Graf 75e1ef4bf2SThomas Graf switch (rule->action) { 76e1ef4bf2SThomas Graf case FR_ACT_TO_TBL: 77e1ef4bf2SThomas Graf break; 78e1ef4bf2SThomas Graf 79e1ef4bf2SThomas Graf case FR_ACT_UNREACHABLE: 80e1ef4bf2SThomas Graf err = -ENETUNREACH; 81e1ef4bf2SThomas Graf goto errout; 82e1ef4bf2SThomas Graf 83e1ef4bf2SThomas Graf case FR_ACT_PROHIBIT: 84e1ef4bf2SThomas Graf err = -EACCES; 85e1ef4bf2SThomas Graf goto errout; 86e1ef4bf2SThomas Graf 87e1ef4bf2SThomas Graf case FR_ACT_BLACKHOLE: 88e1ef4bf2SThomas Graf default: 89e1ef4bf2SThomas Graf err = -EINVAL; 90e1ef4bf2SThomas Graf goto errout; 91e1ef4bf2SThomas Graf } 92e1ef4bf2SThomas Graf 936a31d2a9SEric Dumazet tbl = fib_get_table(rule->fr_net, rule->table); 946a31d2a9SEric Dumazet if (!tbl) 95e1ef4bf2SThomas Graf goto errout; 96e1ef4bf2SThomas Graf 9722bd5b9bSDavid S. Miller err = fib_table_lookup(tbl, &flp->u.ip4, (struct fib_result *) arg->result, arg->flags); 98e1ef4bf2SThomas Graf if (err > 0) 99e1ef4bf2SThomas Graf err = -EAGAIN; 100e1ef4bf2SThomas Graf errout: 101e1ef4bf2SThomas Graf return err; 102e1ef4bf2SThomas Graf } 103e1ef4bf2SThomas Graf 1047764a45aSStefan Tomanek static bool fib4_rule_suppress(struct fib_rule *rule, struct fib_lookup_arg *arg) 1057764a45aSStefan Tomanek { 1066ef94cfaSStefan Tomanek struct fib_result *result = (struct fib_result *) arg->result; 1076ef94cfaSStefan Tomanek struct net_device *dev = result->fi->fib_dev; 1086ef94cfaSStefan Tomanek 1097764a45aSStefan Tomanek /* do not accept result if the route does 1107764a45aSStefan Tomanek * not meet the required prefix length 1117764a45aSStefan Tomanek */ 112*73f5698eSStefan Tomanek if (result->prefixlen <= rule->suppress_prefixlen) 1136ef94cfaSStefan Tomanek goto suppress_route; 1146ef94cfaSStefan Tomanek 1156ef94cfaSStefan Tomanek /* do not accept result if the route uses a device 1166ef94cfaSStefan Tomanek * belonging to a forbidden interface group 1176ef94cfaSStefan Tomanek */ 1186ef94cfaSStefan Tomanek if (rule->suppress_ifgroup != -1 && dev && dev->group == rule->suppress_ifgroup) 1196ef94cfaSStefan Tomanek goto suppress_route; 1206ef94cfaSStefan Tomanek 1216ef94cfaSStefan Tomanek return false; 1226ef94cfaSStefan Tomanek 1236ef94cfaSStefan Tomanek suppress_route: 1247764a45aSStefan Tomanek if (!(arg->flags & FIB_LOOKUP_NOREF)) 1257764a45aSStefan Tomanek fib_info_put(result->fi); 1267764a45aSStefan Tomanek return true; 1277764a45aSStefan Tomanek } 128e1ef4bf2SThomas Graf 129e1ef4bf2SThomas Graf static int fib4_rule_match(struct fib_rule *rule, struct flowi *fl, int flags) 130e1ef4bf2SThomas Graf { 131e1ef4bf2SThomas Graf struct fib4_rule *r = (struct fib4_rule *) rule; 1329ade2286SDavid S. Miller struct flowi4 *fl4 = &fl->u.ip4; 1339ade2286SDavid S. Miller __be32 daddr = fl4->daddr; 1349ade2286SDavid S. Miller __be32 saddr = fl4->saddr; 135e1ef4bf2SThomas Graf 136e1ef4bf2SThomas Graf if (((saddr ^ r->src) & r->srcmask) || 137e1ef4bf2SThomas Graf ((daddr ^ r->dst) & r->dstmask)) 138e1ef4bf2SThomas Graf return 0; 139e1ef4bf2SThomas Graf 1409ade2286SDavid S. Miller if (r->tos && (r->tos != fl4->flowi4_tos)) 141e1ef4bf2SThomas Graf return 0; 142e1ef4bf2SThomas Graf 143e1ef4bf2SThomas Graf return 1; 144e1ef4bf2SThomas Graf } 1451da177e4SLinus Torvalds 1468ad4942cSDenis V. Lunev static struct fib_table *fib_empty_table(struct net *net) 1471da177e4SLinus Torvalds { 1482dfe55b4SPatrick McHardy u32 id; 1491da177e4SLinus Torvalds 1501da177e4SLinus Torvalds for (id = 1; id <= RT_TABLE_MAX; id++) 1518ad4942cSDenis V. Lunev if (fib_get_table(net, id) == NULL) 1528ad4942cSDenis V. Lunev return fib_new_table(net, id); 1531da177e4SLinus Torvalds return NULL; 1541da177e4SLinus Torvalds } 1551da177e4SLinus Torvalds 156ef7c79edSPatrick McHardy static const struct nla_policy fib4_rule_policy[FRA_MAX+1] = { 1571f6c9557SThomas Graf FRA_GENERIC_POLICY, 158e1ef4bf2SThomas Graf [FRA_FLOW] = { .type = NLA_U32 }, 1591da177e4SLinus Torvalds }; 1601da177e4SLinus Torvalds 161e1ef4bf2SThomas Graf static int fib4_rule_configure(struct fib_rule *rule, struct sk_buff *skb, 1628b3521eeSRami Rosen struct fib_rule_hdr *frh, 163e1ef4bf2SThomas Graf struct nlattr **tb) 1641da177e4SLinus Torvalds { 1653b1e0a65SYOSHIFUJI Hideaki struct net *net = sock_net(skb->sk); 166e1ef4bf2SThomas Graf int err = -EINVAL; 167e1ef4bf2SThomas Graf struct fib4_rule *rule4 = (struct fib4_rule *) rule; 1681da177e4SLinus Torvalds 169e1701c68SThomas Graf if (frh->tos & ~IPTOS_TOS_MASK) 170e1ef4bf2SThomas Graf goto errout; 171e1ef4bf2SThomas Graf 172e1ef4bf2SThomas Graf if (rule->table == RT_TABLE_UNSPEC) { 173e1ef4bf2SThomas Graf if (rule->action == FR_ACT_TO_TBL) { 174e1ef4bf2SThomas Graf struct fib_table *table; 175e1ef4bf2SThomas Graf 176e4e4971cSDenis V. Lunev table = fib_empty_table(net); 177e1ef4bf2SThomas Graf if (table == NULL) { 178e1ef4bf2SThomas Graf err = -ENOBUFS; 179e1ef4bf2SThomas Graf goto errout; 180e1ef4bf2SThomas Graf } 181e1ef4bf2SThomas Graf 182e1ef4bf2SThomas Graf rule->table = table->tb_id; 183e1ef4bf2SThomas Graf } 184e1ef4bf2SThomas Graf } 185e1ef4bf2SThomas Graf 186e1701c68SThomas Graf if (frh->src_len) 18745d60b9eSAl Viro rule4->src = nla_get_be32(tb[FRA_SRC]); 188e1ef4bf2SThomas Graf 189e1701c68SThomas Graf if (frh->dst_len) 19045d60b9eSAl Viro rule4->dst = nla_get_be32(tb[FRA_DST]); 191e1ef4bf2SThomas Graf 192c7066f70SPatrick McHardy #ifdef CONFIG_IP_ROUTE_CLASSID 1937a9bc9b8SDavid S. Miller if (tb[FRA_FLOW]) { 194e1ef4bf2SThomas Graf rule4->tclassid = nla_get_u32(tb[FRA_FLOW]); 1957a9bc9b8SDavid S. Miller if (rule4->tclassid) 196f4530fa5SDavid S. Miller net->ipv4.fib_num_tclassid_users++; 1977a9bc9b8SDavid S. Miller } 1981da177e4SLinus Torvalds #endif 1991da177e4SLinus Torvalds 200e1ef4bf2SThomas Graf rule4->src_len = frh->src_len; 201e1ef4bf2SThomas Graf rule4->srcmask = inet_make_mask(rule4->src_len); 202e1ef4bf2SThomas Graf rule4->dst_len = frh->dst_len; 203e1ef4bf2SThomas Graf rule4->dstmask = inet_make_mask(rule4->dst_len); 204e1ef4bf2SThomas Graf rule4->tos = frh->tos; 205e1ef4bf2SThomas Graf 206f4530fa5SDavid S. Miller net->ipv4.fib_has_custom_rules = true; 207e1ef4bf2SThomas Graf err = 0; 208e1ef4bf2SThomas Graf errout: 209e1ef4bf2SThomas Graf return err; 2101da177e4SLinus Torvalds } 2111da177e4SLinus Torvalds 2127a9bc9b8SDavid S. Miller static void fib4_rule_delete(struct fib_rule *rule) 2137a9bc9b8SDavid S. Miller { 214f4530fa5SDavid S. Miller struct net *net = rule->fr_net; 2157a9bc9b8SDavid S. Miller #ifdef CONFIG_IP_ROUTE_CLASSID 2167a9bc9b8SDavid S. Miller struct fib4_rule *rule4 = (struct fib4_rule *) rule; 2177a9bc9b8SDavid S. Miller 2187a9bc9b8SDavid S. Miller if (rule4->tclassid) 219f4530fa5SDavid S. Miller net->ipv4.fib_num_tclassid_users--; 2207a9bc9b8SDavid S. Miller #endif 221f4530fa5SDavid S. Miller net->ipv4.fib_has_custom_rules = true; 2227a9bc9b8SDavid S. Miller } 2237a9bc9b8SDavid S. Miller 224e1ef4bf2SThomas Graf static int fib4_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh, 225e1ef4bf2SThomas Graf struct nlattr **tb) 226a5cdc030SPatrick McHardy { 227e1ef4bf2SThomas Graf struct fib4_rule *rule4 = (struct fib4_rule *) rule; 228a5cdc030SPatrick McHardy 229e1ef4bf2SThomas Graf if (frh->src_len && (rule4->src_len != frh->src_len)) 230e1ef4bf2SThomas Graf return 0; 231e1ef4bf2SThomas Graf 232e1ef4bf2SThomas Graf if (frh->dst_len && (rule4->dst_len != frh->dst_len)) 233e1ef4bf2SThomas Graf return 0; 234e1ef4bf2SThomas Graf 235e1ef4bf2SThomas Graf if (frh->tos && (rule4->tos != frh->tos)) 236e1ef4bf2SThomas Graf return 0; 237e1ef4bf2SThomas Graf 238c7066f70SPatrick McHardy #ifdef CONFIG_IP_ROUTE_CLASSID 239e1ef4bf2SThomas Graf if (tb[FRA_FLOW] && (rule4->tclassid != nla_get_u32(tb[FRA_FLOW]))) 240e1ef4bf2SThomas Graf return 0; 241e1ef4bf2SThomas Graf #endif 242e1ef4bf2SThomas Graf 243e1701c68SThomas Graf if (frh->src_len && (rule4->src != nla_get_be32(tb[FRA_SRC]))) 244e1ef4bf2SThomas Graf return 0; 245e1ef4bf2SThomas Graf 246e1701c68SThomas Graf if (frh->dst_len && (rule4->dst != nla_get_be32(tb[FRA_DST]))) 247e1ef4bf2SThomas Graf return 0; 248e1ef4bf2SThomas Graf 249e1ef4bf2SThomas Graf return 1; 250a5cdc030SPatrick McHardy } 251a5cdc030SPatrick McHardy 252e1ef4bf2SThomas Graf static int fib4_rule_fill(struct fib_rule *rule, struct sk_buff *skb, 25304af8cf6SRami Rosen struct fib_rule_hdr *frh) 2541da177e4SLinus Torvalds { 255e1ef4bf2SThomas Graf struct fib4_rule *rule4 = (struct fib4_rule *) rule; 2561da177e4SLinus Torvalds 257e1ef4bf2SThomas Graf frh->dst_len = rule4->dst_len; 258e1ef4bf2SThomas Graf frh->src_len = rule4->src_len; 259e1ef4bf2SThomas Graf frh->tos = rule4->tos; 2601da177e4SLinus Torvalds 261f3756b79SDavid S. Miller if ((rule4->dst_len && 262f3756b79SDavid S. Miller nla_put_be32(skb, FRA_DST, rule4->dst)) || 263f3756b79SDavid S. Miller (rule4->src_len && 264f3756b79SDavid S. Miller nla_put_be32(skb, FRA_SRC, rule4->src))) 265f3756b79SDavid S. Miller goto nla_put_failure; 266c7066f70SPatrick McHardy #ifdef CONFIG_IP_ROUTE_CLASSID 267f3756b79SDavid S. Miller if (rule4->tclassid && 268f3756b79SDavid S. Miller nla_put_u32(skb, FRA_FLOW, rule4->tclassid)) 269f3756b79SDavid S. Miller goto nla_put_failure; 270e1ef4bf2SThomas Graf #endif 271e1ef4bf2SThomas Graf return 0; 272e1ef4bf2SThomas Graf 273e1ef4bf2SThomas Graf nla_put_failure: 274e1ef4bf2SThomas Graf return -ENOBUFS; 2751da177e4SLinus Torvalds } 2761da177e4SLinus Torvalds 277339bf98fSThomas Graf static size_t fib4_rule_nlmsg_payload(struct fib_rule *rule) 278339bf98fSThomas Graf { 279339bf98fSThomas Graf return nla_total_size(4) /* dst */ 280339bf98fSThomas Graf + nla_total_size(4) /* src */ 281339bf98fSThomas Graf + nla_total_size(4); /* flow */ 282339bf98fSThomas Graf } 283339bf98fSThomas Graf 284ae299fc0SDenis V. Lunev static void fib4_rule_flush_cache(struct fib_rules_ops *ops) 28573417f61SThomas Graf { 286bafa6d9dSNicolas Dichtel rt_cache_flush(ops->fro_net); 28773417f61SThomas Graf } 28873417f61SThomas Graf 28904a6f82cSAndi Kleen static const struct fib_rules_ops __net_initconst fib4_rules_ops_template = { 29025239ceeSPatrick McHardy .family = AF_INET, 291e1ef4bf2SThomas Graf .rule_size = sizeof(struct fib4_rule), 292e1701c68SThomas Graf .addr_size = sizeof(u32), 293e1ef4bf2SThomas Graf .action = fib4_rule_action, 2947764a45aSStefan Tomanek .suppress = fib4_rule_suppress, 295e1ef4bf2SThomas Graf .match = fib4_rule_match, 296e1ef4bf2SThomas Graf .configure = fib4_rule_configure, 2977a9bc9b8SDavid S. Miller .delete = fib4_rule_delete, 298e1ef4bf2SThomas Graf .compare = fib4_rule_compare, 299e1ef4bf2SThomas Graf .fill = fib4_rule_fill, 300d8a566beSPatrick McHardy .default_pref = fib_default_rule_pref, 301339bf98fSThomas Graf .nlmsg_payload = fib4_rule_nlmsg_payload, 30273417f61SThomas Graf .flush_cache = fib4_rule_flush_cache, 303e1ef4bf2SThomas Graf .nlgroup = RTNLGRP_IPV4_RULE, 304e1ef4bf2SThomas Graf .policy = fib4_rule_policy, 305e1ef4bf2SThomas Graf .owner = THIS_MODULE, 306e1ef4bf2SThomas Graf }; 307e1ef4bf2SThomas Graf 308e4e4971cSDenis V. Lunev static int fib_default_rules_init(struct fib_rules_ops *ops) 3092994c638SDenis V. Lunev { 3102994c638SDenis V. Lunev int err; 3112994c638SDenis V. Lunev 3125adef180SPatrick McHardy err = fib_default_rule_add(ops, 0, RT_TABLE_LOCAL, 0); 3132994c638SDenis V. Lunev if (err < 0) 3142994c638SDenis V. Lunev return err; 315e4e4971cSDenis V. Lunev err = fib_default_rule_add(ops, 0x7FFE, RT_TABLE_MAIN, 0); 3162994c638SDenis V. Lunev if (err < 0) 3172994c638SDenis V. Lunev return err; 318e4e4971cSDenis V. Lunev err = fib_default_rule_add(ops, 0x7FFF, RT_TABLE_DEFAULT, 0); 3192994c638SDenis V. Lunev if (err < 0) 3202994c638SDenis V. Lunev return err; 3212994c638SDenis V. Lunev return 0; 3222994c638SDenis V. Lunev } 3232994c638SDenis V. Lunev 3247b1a74fdSDenis V. Lunev int __net_init fib4_rules_init(struct net *net) 325e1ef4bf2SThomas Graf { 326dbb50165SDenis V. Lunev int err; 327e4e4971cSDenis V. Lunev struct fib_rules_ops *ops; 328dbb50165SDenis V. Lunev 329e9c5158aSEric W. Biederman ops = fib_rules_register(&fib4_rules_ops_template, net); 330e9c5158aSEric W. Biederman if (IS_ERR(ops)) 331e9c5158aSEric W. Biederman return PTR_ERR(ops); 332e4e4971cSDenis V. Lunev 333e4e4971cSDenis V. Lunev err = fib_default_rules_init(ops); 334dbb50165SDenis V. Lunev if (err < 0) 335dbb50165SDenis V. Lunev goto fail; 336e4e4971cSDenis V. Lunev net->ipv4.rules_ops = ops; 337f4530fa5SDavid S. Miller net->ipv4.fib_has_custom_rules = false; 338dbb50165SDenis V. Lunev return 0; 339dbb50165SDenis V. Lunev 340dbb50165SDenis V. Lunev fail: 341dbb50165SDenis V. Lunev /* also cleans all rules already added */ 3429e3a5487SDenis V. Lunev fib_rules_unregister(ops); 343dbb50165SDenis V. Lunev return err; 344e1ef4bf2SThomas Graf } 3457b1a74fdSDenis V. Lunev 3467b1a74fdSDenis V. Lunev void __net_exit fib4_rules_exit(struct net *net) 3477b1a74fdSDenis V. Lunev { 3489e3a5487SDenis V. Lunev fib_rules_unregister(net->ipv4.rules_ops); 3497b1a74fdSDenis V. Lunev } 350