From feda3ba2effd74936f7ba1b03da1e30a9dee3a16 Mon Sep 17 00:00:00 2001 From: Mike Evans Date: Thu, 4 Jun 2015 07:50:08 +0000 Subject: [PATCH 2/3] Add kernel tunnel dynamic attribute For IP-in-IP routes, use original next hop received in BGP UPDATE Instead of using a recursively resolved next hop, which will probably be to somewhere half way along the tunnel, from where packets won't be able to be routed any further. --- nest/route.h | 1 + proto/bgp/packets.c | 1 + sysdep/linux/krt-sys.h | 1 + sysdep/linux/netlink.Y | 4 +++- sysdep/linux/netlink.c | 23 +++++++++++++++++++++-- 5 files changed, 27 insertions(+), 3 deletions(-) diff --git a/nest/route.h b/nest/route.h index 2bfd066..3eb86c8 100644 --- a/nest/route.h +++ b/nest/route.h @@ -359,6 +359,7 @@ typedef struct rta { byte aflags; /* Attribute cache flags (RTAF_...) */ u16 hash_key; /* Hash over important fields */ u32 igp_metric; /* IGP metric to next hop (for iBGP routes) */ + ip_addr orig_gw; /* Original next hop from BGP UPDATE */ ip_addr gw; /* Next hop */ ip_addr from; /* Advertising router */ struct hostentry *hostentry; /* Hostentry for recursive next-hops */ diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c index 0cf38ed..fba9d60 100644 --- a/proto/bgp/packets.c +++ b/proto/bgp/packets.c @@ -1180,6 +1180,7 @@ bgp_set_next_hop(struct bgp_proto *p, rta *a) if (ipa_zero(*nexthop)) return 0; + a->orig_gw = *nexthop; rta_set_recursive_next_hop(p->p.table, a, p->igp_table, nexthop, nexthop + second); } diff --git a/sysdep/linux/krt-sys.h b/sysdep/linux/krt-sys.h index 6d6586d..cd4d2ec 100644 --- a/sysdep/linux/krt-sys.h +++ b/sysdep/linux/krt-sys.h @@ -37,6 +37,7 @@ static inline struct ifa * kif_get_primary_ip(struct iface *i) { return NULL; } #define EA_KRT_PREFSRC EA_CODE(EAP_KRT, 0x10) #define EA_KRT_REALM EA_CODE(EAP_KRT, 0x11) #define EA_KRT_SCOPE EA_CODE(EAP_KRT, 0x12) +#define EA_KRT_TUNNEL EA_CODE(EAP_KRT, 0x13) #define KRT_METRICS_MAX 0x10 /* RTAX_QUICKACK+1 */ diff --git a/sysdep/linux/netlink.Y b/sysdep/linux/netlink.Y index f577244..4266333 100644 --- a/sysdep/linux/netlink.Y +++ b/sysdep/linux/netlink.Y @@ -15,7 +15,8 @@ CF_KEYWORDS(KERNEL, TABLE, METRIC, KRT_PREFSRC, KRT_REALM, KRT_SCOPE, KRT_MTU, K KRT_HOPLIMIT, KRT_INITCWND, KRT_RTO_MIN, KRT_INITRWND, KRT_QUICKACK, KRT_LOCK_MTU, KRT_LOCK_WINDOW, KRT_LOCK_RTT, KRT_LOCK_RTTVAR, KRT_LOCK_SSTRESH, KRT_LOCK_CWND, KRT_LOCK_ADVMSS, KRT_LOCK_REORDERING, - KRT_LOCK_HOPLIMIT, KRT_LOCK_RTO_MIN, KRT_FEATURE_ECN, KRT_FEATURE_ALLFRAG) + KRT_LOCK_HOPLIMIT, KRT_LOCK_RTO_MIN, KRT_FEATURE_ECN, KRT_FEATURE_ALLFRAG, + KRT_TUNNEL) CF_GRAMMAR @@ -29,6 +30,7 @@ kern_sys_item: CF_ADDTO(dynamic_attr, KRT_PREFSRC { $$ = f_new_dynamic_attr(EAF_TYPE_IP_ADDRESS, T_IP, EA_KRT_PREFSRC); }) CF_ADDTO(dynamic_attr, KRT_REALM { $$ = f_new_dynamic_attr(EAF_TYPE_INT, T_INT, EA_KRT_REALM); }) CF_ADDTO(dynamic_attr, KRT_SCOPE { $$ = f_new_dynamic_attr(EAF_TYPE_INT, T_INT, EA_KRT_SCOPE); }) +CF_ADDTO(dynamic_attr, KRT_TUNNEL { $$ = f_new_dynamic_attr(EAF_TYPE_STRING, T_STRING, EA_KRT_TUNNEL); }) CF_ADDTO(dynamic_attr, KRT_MTU { $$ = f_new_dynamic_attr(EAF_TYPE_INT, T_INT, EA_KRT_MTU); }) CF_ADDTO(dynamic_attr, KRT_WINDOW { $$ = f_new_dynamic_attr(EAF_TYPE_INT, T_INT, EA_KRT_WINDOW); }) diff --git a/sysdep/linux/netlink.c b/sysdep/linux/netlink.c index 1490213..d754df7 100644 --- a/sysdep/linux/netlink.c +++ b/sysdep/linux/netlink.c @@ -887,6 +887,7 @@ nl_send_route(struct krt_proto *p, rte *e, struct ea_list *eattrs, int op, int d struct rtmsg r; char buf[128 + KRT_METRICS_MAX*8 + nh_bufsize(a->nexthops)]; } r; + struct iface* i; DBG("nl_send_route(%I/%d,op=%x)\n", net->n.prefix, net->n.pxlen, op); @@ -962,8 +963,22 @@ dest: { case RTD_ROUTER: r.r.rtm_type = RTN_UNICAST; - nl_add_attr_u32(&r.h, sizeof(r), RTA_OIF, iface->index); - nl_add_attr_ipa(&r.h, sizeof(r), RTA_GATEWAY, gw); + if ((ea = ea_find(eattrs, EA_KRT_TUNNEL)) && + (i = if_find_by_name(ea->u.ptr->data))) + { + /* + * Tunnel attribute is set, so set the route up using the specified tunnel device + * to the originator of the route. + */ + nl_add_attr_u32(&r.h, sizeof(r), RTA_OIF, i->index); + nl_add_attr_ipa(&r.h, sizeof(r), RTA_GATEWAY, a->orig_gw); + r.r.rtm_flags |= RTNH_F_ONLINK; + } + else + { + nl_add_attr_u32(&r.h, sizeof(r), RTA_OIF, iface->index); + nl_add_attr_ipa(&r.h, sizeof(r), RTA_GATEWAY, gw); + } break; case RTD_DEVICE: r.r.rtm_type = RTN_UNICAST; @@ -1652,6 +1667,10 @@ krt_sys_get_attr(eattr *a, byte *buf, int buflen UNUSED) bsprintf(buf, "scope"); return GA_NAME; + case EA_KRT_TUNNEL: + bsprintf(buf, "tunnel"); + return GA_NAME; + case EA_KRT_LOCK: buf += bsprintf(buf, "lock:"); ea_format_bitfield(a, buf, buflen, krt_metrics_names, 2, KRT_METRICS_MAX); -- 2.7.4