diff options
Diffstat (limited to 'proto/ospf/topology.c')
-rw-r--r-- | proto/ospf/topology.c | 120 |
1 files changed, 67 insertions, 53 deletions
diff --git a/proto/ospf/topology.c b/proto/ospf/topology.c index 341eff87..ce77f57a 100644 --- a/proto/ospf/topology.c +++ b/proto/ospf/topology.c @@ -283,8 +283,8 @@ ospf_originate_lsa(struct ospf_proto *p, struct ospf_new_lsa *lsa) if (en->nf != lsa->nf) { - log(L_ERR "%s: LSA ID collision for %I/%d", - p->p.name, lsa->nf->fn.prefix, lsa->nf->fn.pxlen); + log(L_ERR "%s: LSA ID collision for %N", + p->p.name, lsa->nf->fn.addr); en = NULL; goto drop; @@ -514,14 +514,14 @@ ospf_update_lsadb(struct ospf_proto *p) } -static inline u32 -ort_to_lsaid(struct ospf_proto *p UNUSED4 UNUSED6, ort *nf) +static u32 +ort_to_lsaid(struct ospf_proto *p, ort *nf) { /* * In OSPFv2, We have to map IP prefixes to u32 in such manner that resulting * u32 interpreted as IP address is a member of given prefix. Therefore, /32 - * prefix have to be mapped on itself. All received prefixes have to be - * mapped on different u32s. + * prefix has to be mapped on itself. All received prefixes have to be mapped + * on different u32s. * * We have an assumption that if there is nontrivial (non-/32) network prefix, * then there is not /32 prefix for the first and the last IP address of the @@ -542,17 +542,21 @@ ort_to_lsaid(struct ospf_proto *p UNUSED4 UNUSED6, ort *nf) * network appeared, we choose a different way. * * In OSPFv3, it is simpler. There is not a requirement for membership of the - * result in the input network, so we just use a hash-based unique ID of a - * routing table entry for a route that originated given LSA. For ext-LSA, it - * is an imported route in the nest's routing table (p->table). For summary-LSA, - * it is a 'source' route in the protocol internal routing table (p->rtf). + * result in the input network, so we just allocate a unique ID from ID map + * and store it in nf->lsa_id for further reference. */ if (ospf_is_v3(p)) - return nf->fn.uid; + { + if (!nf->lsa_id) + nf->lsa_id = idm_alloc(&p->idm); - u32 id = ipa_to_u32(nf->fn.prefix); - int pxlen = nf->fn.pxlen; + return nf->lsa_id; + } + + net_addr_ip4 *net = (void *) nf->fn.addr; + u32 id = ip4_to_u32(net->prefix); + int pxlen = net->pxlen; if ((pxlen == 0) || (pxlen == 32)) return id; @@ -628,12 +632,12 @@ configured_stubnet(struct ospf_area *oa, struct ifa *a) { if (sn->summary) { - if (ipa_in_net(a->prefix, sn->px.addr, sn->px.len) && (a->pxlen >= sn->px.len)) + if (net_in_netX(&a->prefix, &sn->prefix)) return 1; } else { - if (ipa_equal(a->prefix, sn->px.addr) && (a->pxlen == sn->px.len)) + if (net_equal(&a->prefix, &sn->prefix)) return 1; } } @@ -781,7 +785,8 @@ prepare_rt2_lsa_body(struct ospf_proto *p, struct ospf_area *oa) (ifa->type == OSPF_IT_PTMP)) add_rt2_lsa_link(p, LSART_STUB, ipa_to_u32(ifa->addr->ip), 0xffffffff, 0); else - add_rt2_lsa_link(p, LSART_STUB, ipa_to_u32(ifa->addr->prefix), u32_mkmask(ifa->addr->pxlen), ifa->cost); + add_rt2_lsa_link(p, LSART_STUB, ip4_to_u32(net4_prefix(&ifa->addr->prefix)), + u32_mkmask(net4_pxlen(&ifa->addr->prefix)), ifa->cost); i++; ifa->rt_pos_end = i; @@ -790,7 +795,8 @@ prepare_rt2_lsa_body(struct ospf_proto *p, struct ospf_area *oa) struct ospf_stubnet_config *sn; WALK_LIST(sn, oa->ac->stubnet_list) if (!sn->hidden) - add_rt2_lsa_link(p, LSART_STUB, ipa_to_u32(sn->px.addr), u32_mkmask(sn->px.len), sn->cost), i++; + add_rt2_lsa_link(p, LSART_STUB, ip4_to_u32(net4_prefix(&sn->prefix)), + u32_mkmask(net4_pxlen(&sn->prefix)), sn->cost), i++; struct ospf_lsa_rt *rt = p->lsab; /* Store number of links in lower half of options */ @@ -907,7 +913,7 @@ prepare_net2_lsa_body(struct ospf_proto *p, struct ospf_iface *ifa) ASSERT(p->lsab_used == 0); net = lsab_alloc(p, sizeof(struct ospf_lsa_net) + 4 * nodes); - net->optx = u32_mkmask(ifa->addr->pxlen); + net->optx = u32_mkmask(ifa->addr->prefix.pxlen); net->routers[0] = p->router_id; WALK_LIST(n, ifa->neigh_list) @@ -999,9 +1005,10 @@ prepare_sum3_net_lsa_body(struct ospf_proto *p, ort *nf, u32 metric) { struct ospf_lsa_sum3_net *sum; - sum = lsab_allocz(p, sizeof(struct ospf_lsa_sum3_net) + IPV6_PREFIX_SPACE(nf->fn.pxlen)); + sum = lsab_allocz(p, sizeof(struct ospf_lsa_sum3_net) + + IPV6_PREFIX_SPACE(nf->fn.addr->pxlen)); sum->metric = metric; - put_ipv6_prefix(sum->prefix, nf->fn.prefix, nf->fn.pxlen, 0, 0); + ospf_put_ipv6_prefix(sum->prefix, nf->fn.addr, 0, 0); } static inline void @@ -1028,7 +1035,7 @@ ospf_originate_sum_net_lsa(struct ospf_proto *p, struct ospf_area *oa, ort *nf, }; if (ospf_is_v2(p)) - prepare_sum2_lsa_body(p, nf->fn.pxlen, metric); + prepare_sum2_lsa_body(p, nf->fn.addr->pxlen, metric); else prepare_sum3_net_lsa_body(p, nf, metric); @@ -1036,20 +1043,20 @@ ospf_originate_sum_net_lsa(struct ospf_proto *p, struct ospf_area *oa, ort *nf, } void -ospf_originate_sum_rt_lsa(struct ospf_proto *p, struct ospf_area *oa, ort *nf, int metric, u32 options) +ospf_originate_sum_rt_lsa(struct ospf_proto *p, struct ospf_area *oa, u32 drid, int metric, u32 options) { struct ospf_new_lsa lsa = { .type = LSA_T_SUM_RT, .mode = LSA_M_RTCALC, .dom = oa->areaid, - .id = ipa_to_rid(nf->fn.prefix), /* Router ID of ASBR, irrelevant for OSPFv3 */ + .id = drid, /* Router ID of ASBR, irrelevant for OSPFv3 */ .opts = oa->options }; if (ospf_is_v2(p)) prepare_sum2_lsa_body(p, 0, metric); else - prepare_sum3_rt_lsa_body(p, lsa.id, metric, options & LSA_OPTIONS_MASK); + prepare_sum3_rt_lsa_body(p, drid, metric, options & LSA_OPTIONS_MASK); ospf_originate_lsa(p, &lsa); } @@ -1082,7 +1089,7 @@ prepare_ext3_lsa_body(struct ospf_proto *p, ort *nf, { struct ospf_lsa_ext3 *ext; int bsize = sizeof(struct ospf_lsa_ext3) - + IPV6_PREFIX_SPACE(nf->fn.pxlen) + + IPV6_PREFIX_SPACE(nf->fn.addr->pxlen) + (ipa_nonzero(fwaddr) ? 16 : 0) + (tag ? 4 : 0); @@ -1090,7 +1097,7 @@ prepare_ext3_lsa_body(struct ospf_proto *p, ort *nf, ext->metric = metric & LSA_METRIC_MASK; u32 *buf = ext->rest; - buf = put_ipv6_prefix(buf, nf->fn.prefix, nf->fn.pxlen, pbit ? OPT_PX_P : 0, 0); + buf = ospf_put_ipv6_prefix(buf, nf->fn.addr, pbit ? OPT_PX_P : 0, 0); if (ebit) ext->metric |= LSA_EXT3_EBIT; @@ -1098,7 +1105,7 @@ prepare_ext3_lsa_body(struct ospf_proto *p, ort *nf, if (ipa_nonzero(fwaddr)) { ext->metric |= LSA_EXT3_FBIT; - buf = put_ipv6_addr(buf, fwaddr); + buf = ospf_put_ipv6_addr(buf, fwaddr); } if (tag) @@ -1140,7 +1147,7 @@ ospf_originate_ext_lsa(struct ospf_proto *p, struct ospf_area *oa, ort *nf, u8 m }; if (ospf_is_v2(p)) - prepare_ext2_lsa_body(p, nf->fn.pxlen, metric, ebit, fwaddr, tag); + prepare_ext2_lsa_body(p, nf->fn.addr->pxlen, metric, ebit, fwaddr, tag); else prepare_ext3_lsa_body(p, nf, metric, ebit, fwaddr, tag, oa && pbit); @@ -1177,7 +1184,7 @@ use_gw_for_fwaddr(struct ospf_proto *p, ip_addr gw, struct iface *iface) WALK_LIST(ifa, p->iface_list) if ((ifa->iface == iface) && - (!ospf_is_v2(p) || ipa_in_net(gw, ifa->addr->prefix, ifa->addr->pxlen))) + (!ospf_is_v2(p) || ipa_in_netX(gw, &ifa->addr->prefix))) return 1; return 0; @@ -1215,7 +1222,8 @@ find_surrogate_fwaddr(struct ospf_proto *p, struct ospf_area *oa) { WALK_LIST(a, ifa->iface->addrs) { - if ((a->flags & IA_SECONDARY) || + if ((a->prefix.type != NET_IP6) || + (a->flags & IA_SECONDARY) || (a->flags & IA_PEER) || (a->scope <= SCOPE_LINK)) continue; @@ -1234,7 +1242,7 @@ find_surrogate_fwaddr(struct ospf_proto *p, struct ospf_area *oa) } void -ospf_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old UNUSED, ea_list *ea) +ospf_rt_notify(struct proto *P, struct channel *ch UNUSED, net *n, rte *new, rte *old UNUSED, ea_list *ea) { struct ospf_proto *p = (struct ospf_proto *) P; struct ospf_area *oa = NULL; /* non-NULL for NSSA-LSA */ @@ -1253,7 +1261,7 @@ ospf_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old U if (!new) { - nf = (ort *) fib_find(&p->rtf, &n->n.prefix, n->n.pxlen); + nf = fib_find(&p->rtf, n->n.addr); if (!nf || !nf->external_rte) return; @@ -1280,8 +1288,8 @@ ospf_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old U ip_addr fwd = IPA_NONE; - if ((a->dest == RTD_ROUTER) && use_gw_for_fwaddr(p, a->gw, a->iface)) - fwd = a->gw; + if ((a->dest == RTD_UNICAST) && use_gw_for_fwaddr(p, a->nh.gw, a->nh.iface)) + fwd = a->nh.gw; /* NSSA-LSA with P-bit set must have non-zero forwarding address */ if (oa && ipa_zero(fwd)) @@ -1290,13 +1298,13 @@ ospf_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old U if (ipa_zero(fwd)) { - log(L_ERR "%s: Cannot find forwarding address for NSSA-LSA %I/%d", - p->p.name, n->n.prefix, n->n.pxlen); + log(L_ERR "%s: Cannot find forwarding address for NSSA-LSA %N", + p->p.name, n->n.addr); return; } } - nf = (ort *) fib_get(&p->rtf, &n->n.prefix, n->n.pxlen); + nf = fib_get(&p->rtf, n->n.addr); ospf_originate_ext_lsa(p, oa, nf, LSA_M_EXPORT, metric, ebit, fwd, tag, 1); nf->external_rte = 1; } @@ -1308,11 +1316,11 @@ ospf_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old U */ static inline void -lsab_put_prefix(struct ospf_proto *p, ip_addr prefix, u32 pxlen, u32 cost) +lsab_put_prefix(struct ospf_proto *p, net_addr *net, u32 cost) { - void *buf = lsab_alloc(p, IPV6_PREFIX_SPACE(pxlen)); - u8 flags = (pxlen < MAX_PREFIX_LENGTH) ? 0 : OPT_PX_LA; - put_ipv6_prefix(buf, prefix, pxlen, flags, cost); + void *buf = lsab_alloc(p, IPV6_PREFIX_SPACE(net6_pxlen(net))); + u8 flags = (net6_pxlen(net) < IP6_MAX_PREFIX_LENGTH) ? 0 : OPT_PX_LA; + ospf_put_ipv6_prefix(buf, net, flags, cost); } static void @@ -1330,11 +1338,12 @@ prepare_link_lsa_body(struct ospf_proto *p, struct ospf_iface *ifa) struct ifa *a; WALK_LIST(a, ifa->iface->addrs) { - if ((a->flags & IA_SECONDARY) || - (a->scope < SCOPE_SITE)) + if ((a->prefix.type != NET_IP6) || + (a->flags & IA_SECONDARY) || + (a->scope <= SCOPE_LINK)) continue; - lsab_put_prefix(p, a->prefix, a->pxlen, 0); + lsab_put_prefix(p, &a->prefix, 0); i++; } @@ -1401,12 +1410,13 @@ prepare_prefix_rt_lsa_body(struct ospf_proto *p, struct ospf_area *oa) struct ifa *a; WALK_LIST(a, ifa->iface->addrs) { - if ((a->flags & IA_SECONDARY) || + if ((a->prefix.type != NET_IP6) || + (a->flags & IA_SECONDARY) || (a->flags & IA_PEER) || (a->scope <= SCOPE_LINK)) continue; - if (((a->pxlen < MAX_PREFIX_LENGTH) && net_lsa) || + if (((a->prefix.pxlen < IP6_MAX_PREFIX_LENGTH) && net_lsa) || configured_stubnet(oa, a)) continue; @@ -1414,11 +1424,12 @@ prepare_prefix_rt_lsa_body(struct ospf_proto *p, struct ospf_area *oa) (ifa->state == OSPF_IS_LOOP) || (ifa->type == OSPF_IT_PTMP)) { - lsab_put_prefix(p, a->ip, MAX_PREFIX_LENGTH, 0); + net_addr_ip6 net = NET_ADDR_IP6(a->ip, IP6_MAX_PREFIX_LENGTH); + lsab_put_prefix(p, (net_addr *) &net, 0); host_addr = 1; } else - lsab_put_prefix(p, a->prefix, a->pxlen, ifa->cost); + lsab_put_prefix(p, &a->prefix, ifa->cost); i++; } @@ -1429,8 +1440,8 @@ prepare_prefix_rt_lsa_body(struct ospf_proto *p, struct ospf_area *oa) WALK_LIST(sn, oa->ac->stubnet_list) if (!sn->hidden) { - lsab_put_prefix(p, sn->px.addr, sn->px.len, sn->cost); - if (sn->px.len == MAX_PREFIX_LENGTH) + lsab_put_prefix(p, &sn->prefix, sn->cost); + if (sn->prefix.pxlen == IP6_MAX_PREFIX_LENGTH) host_addr = 1; i++; } @@ -1447,11 +1458,14 @@ prepare_prefix_rt_lsa_body(struct ospf_proto *p, struct ospf_area *oa) struct ifa *a; WALK_LIST(a, ifa->iface->addrs) { - if ((a->flags & IA_SECONDARY) || (a->scope <= SCOPE_LINK)) + if ((a->prefix.type != NET_IP6) || + (a->flags & IA_SECONDARY) || + (a->scope <= SCOPE_LINK)) continue; /* Found some IP */ - lsab_put_prefix(p, a->ip, MAX_PREFIX_LENGTH, 0); + net_addr_ip6 net = NET_ADDR_IP6(a->ip, IP6_MAX_PREFIX_LENGTH); + lsab_put_prefix(p, (net_addr *) &net, 0); i++; goto done; } @@ -1748,7 +1762,7 @@ ospf_top_hash(struct top_graph *f, u32 domain, u32 lsaid, u32 rtrid, u32 type) * and request lists of OSPF neighbors. */ struct top_graph * -ospf_top_new(struct ospf_proto *p UNUSED4 UNUSED6, pool *pool) +ospf_top_new(struct ospf_proto *p, pool *pool) { struct top_graph *f; |