From 11cb620266035ffbe17b21c4a174380cb8b6a521 Mon Sep 17 00:00:00 2001 From: Ondrej Zajicek Date: Sun, 26 Oct 2008 22:36:08 +0100 Subject: Implementation of 4B ASN support for BGP --- proto/bgp/packets.c | 151 ++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 116 insertions(+), 35 deletions(-) (limited to 'proto/bgp/packets.c') diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c index 2e6f0b60..0dd920e4 100644 --- a/proto/bgp/packets.c +++ b/proto/bgp/packets.c @@ -12,6 +12,7 @@ #include "nest/iface.h" #include "nest/protocol.h" #include "nest/route.h" +#include "nest/attrs.h" #include "conf/conf.h" #include "lib/unaligned.h" #include "lib/socket.h" @@ -30,33 +31,64 @@ bgp_create_notification(struct bgp_conn *conn, byte *buf) return buf + 2 + conn->notify_size; } +#ifdef IPV6 +static byte * +bgp_put_cap_ipv6(struct bgp_conn *conn UNUSED, byte *buf) +{ + *buf++ = 1; /* Capability 1: Multiprotocol extensions */ + *buf++ = 4; /* Capability data length */ + *buf++ = 0; /* We support AF IPv6 */ + *buf++ = BGP_AF_IPV6; + *buf++ = 0; /* RFU */ + *buf++ = 1; /* and SAFI 1 */ + return buf; +} +#endif + +static byte * +bgp_put_cap_as4(struct bgp_conn *conn, byte *buf) +{ + *buf++ = 65; /* Capability 65: Support for 4-octet AS number */ + *buf++ = 4; /* Capability data length */ + put_u32(buf, conn->bgp->local_as); + return buf + 4; +} + static byte * bgp_create_open(struct bgp_conn *conn, byte *buf) { struct bgp_proto *p = conn->bgp; + byte *cap; + int cap_len; BGP_TRACE(D_PACKETS, "Sending OPEN(ver=%d,as=%d,hold=%d,id=%08x)", BGP_VERSION, p->local_as, p->cf->hold_time, p->local_id); buf[0] = BGP_VERSION; - put_u16(buf+1, p->local_as); + put_u16(buf+1, (p->local_as < 0xFFFF) ? p->local_as : AS_TRANS); put_u16(buf+3, p->cf->hold_time); put_u32(buf+5, p->local_id); -#ifndef IPV6 - buf[9] = 0; /* No optional parameters */ - return buf+10; -#else - buf += 9; - *buf++ = 8; /* Optional params len */ - *buf++ = 2; /* Option: Capability list */ - *buf++ = 6; /* Option length */ - *buf++ = 1; /* Capability 1: Multiprotocol extensions */ - *buf++ = 4; /* Capability data length */ - *buf++ = 0; /* We support AF IPv6 */ - *buf++ = BGP_AF_IPV6; - *buf++ = 0; /* RFU */ - *buf++ = 1; /* and SAFI 1 */ - return buf; + /* Skipped 3 B for length field and Capabilities parameter header */ + cap = buf + 12; + +#ifdef IPV6 + cap = bgp_put_cap_ipv6(conn, cap); #endif + if (bgp_as4_support) + cap = bgp_put_cap_as4(conn, cap); + + cap_len = cap - buf - 12; + if (cap_len > 0) + { + buf[9] = cap_len + 2; /* Optional params len */ + buf[10] = 2; /* Option: Capability list */ + buf[11] = cap_len; /* Option length */ + return cap; + } + else + { + buf[9] = 0; /* No optional parameters */ + return buf + 10; + } } static unsigned int @@ -118,7 +150,7 @@ bgp_create_update(struct bgp_conn *conn, byte *buf) continue; } DBG("Processing bucket %p\n", buck); - a_size = bgp_encode_attrs(w+2, buck->eattrs, 1024); + a_size = bgp_encode_attrs(p, w+2, buck->eattrs, 1024); put_u16(w, a_size); w += a_size + 2; r_size = bgp_encode_prefixes(p, w, buck, remains - a_size); @@ -166,7 +198,7 @@ bgp_create_update(struct bgp_conn *conn, byte *buf) *tmp++ = BGP_AF_IPV6; *tmp++ = 1; ea->attrs[0].u.ptr->length = bgp_encode_prefixes(p, tmp, buck, remains-11); - size = bgp_encode_attrs(w, ea, remains); + size = bgp_encode_attrs(p, w, ea, remains); w += size; remains -= size; } @@ -183,7 +215,7 @@ bgp_create_update(struct bgp_conn *conn, byte *buf) continue; } DBG("Processing bucket %p\n", buck); - size = bgp_encode_attrs(w, buck->eattrs, 1024); + size = bgp_encode_attrs(p, w, buck->eattrs, 1024); w += size; remains -= size; tstart = tmp = bgp_attach_attr(&ea, bgp_linpool, BA_MP_REACH_NLRI, remains-8); @@ -230,7 +262,7 @@ bgp_create_update(struct bgp_conn *conn, byte *buf) *tmp++ = 0; /* No SNPA information */ tmp += bgp_encode_prefixes(p, tmp, buck, remains - (8+3+32+1)); ea->attrs[0].u.ptr->length = tmp - tstart; - w += bgp_encode_attrs(w, ea, remains); + w += bgp_encode_attrs(p, w, ea, remains); break; } } @@ -353,9 +385,49 @@ bgp_tx(sock *sk) ; } +/* Capatibility negotiation as per RFC 2842 */ + +void +bgp_parse_capabilities(struct bgp_conn *conn, byte *opt, int len) +{ + struct bgp_proto *p = conn->bgp; + int cl; + u32 as; + + while (len > 0) + { + if (len < 2 || len < 2 + opt[1]) + goto err; + + cl = opt[1]; + + switch (opt[0]) + { + case 65: + if (cl != 4) + goto err; + p->as4_support = 1; + if (bgp_as4_support) + conn->advertised_as = get_u32(opt + 2); + break; + + /* We can safely ignore all other capabilities */ + } + len -= 2 + cl; + opt += 2 + cl; + } + return; + + err: + bgp_error(conn, 2, 0, NULL, 0); + return; +} + static int bgp_parse_options(struct bgp_conn *conn, byte *opt, int len) { + int ol; + while (len > 0) { if (len < 2 || len < 2 + opt[1]) @@ -369,12 +441,14 @@ bgp_parse_options(struct bgp_conn *conn, byte *opt, int len) DBG("\n"); } #endif + + ol = opt[1]; switch (opt[0]) { case 2: - /* Capatibility negotiation as per RFC 2842 */ - /* We can safely ignore all capabilities announced */ + bgp_parse_capabilities(conn, opt + 2, ol); break; + default: /* * BGP specs don't tell us to send which option @@ -382,11 +456,11 @@ bgp_parse_options(struct bgp_conn *conn, byte *opt, int len) * to do so. Also, capability negotiation with * Cisco routers doesn't work without that. */ - bgp_error(conn, 2, 4, opt, opt[1]); + bgp_error(conn, 2, 4, opt, ol); return 0; } - len -= 2 + opt[1]; - opt += 2 + opt[1]; + len -= 2 + ol; + opt += 2 + ol; } return 0; } @@ -397,7 +471,7 @@ bgp_rx_open(struct bgp_conn *conn, byte *pkt, int len) struct bgp_conn *other; struct bgp_proto *p = conn->bgp; struct bgp_config *cf = p->cf; - unsigned as, hold; + unsigned hold; u32 id; /* Check state */ @@ -409,20 +483,27 @@ bgp_rx_open(struct bgp_conn *conn, byte *pkt, int len) { bgp_error(conn, 1, 2, pkt+16, 2); return; } if (pkt[19] != BGP_VERSION) { bgp_error(conn, 2, 1, pkt+19, 1); return; } /* RFC 1771 says 16 bits, draft-09 tells to use 8 */ - as = get_u16(pkt+20); + conn->advertised_as = get_u16(pkt+20); hold = get_u16(pkt+22); id = get_u32(pkt+24); - BGP_TRACE(D_PACKETS, "Got OPEN(as=%d,hold=%d,id=%08x)", as, hold, id); - if (cf->remote_as && as != p->remote_as) - { bgp_error(conn, 2, 2, pkt+20, -2); return; } - if (hold > 0 && hold < 3) - { bgp_error(conn, 2, 6, pkt+22, 2); return; } - p->remote_id = id; + BGP_TRACE(D_PACKETS, "Got OPEN(as=%d,hold=%d,id=%08x)", conn->advertised_as, hold, id); + + p->remote_id = id; // ??? if (bgp_parse_options(conn, pkt+29, pkt[28])) return; + + if (hold > 0 && hold < 3) + { bgp_error(conn, 2, 6, pkt+22, 2); return; } + if (!id || id == 0xffffffff || id == p->local_id) { bgp_error(conn, 2, 3, pkt+24, -4); return; } + + if (conn->advertised_as != p->remote_as) + { + bgp_error(conn, 2, 2, (byte *) &(conn->advertised_as), -4); return; + } + /* Check the other connection */ other = (conn == &p->outgoing_conn) ? &p->incoming_conn : &p->outgoing_conn; switch (other->state) @@ -463,7 +544,7 @@ bgp_rx_open(struct bgp_conn *conn, byte *pkt, int len) else conn->hold_time = p->cf->hold_time; conn->keepalive_time = p->cf->keepalive_time ? : conn->hold_time / 3; - p->remote_as = as; + // p->remote_as = conn->advertised_as; p->remote_id = id; DBG("BGP: Hold timer set to %d, keepalive to %d, AS to %d, ID to %x\n", conn->hold_time, conn->keepalive_time, p->remote_as, p->remote_id); @@ -720,7 +801,7 @@ static struct { { 2, 4, "Unsupported optional parameter" }, { 2, 5, "Authentication failure" }, { 2, 6, "Unacceptable hold time" }, - { 2, 7, "Required capability missing" }, /* capability negotiation draft */ + { 2, 7, "Required capability missing" }, /* [RFC3392] */ { 3, 0, "Invalid UPDATE message" }, { 3, 1, "Malformed attribute list" }, { 3, 2, "Unrecognized well-known attribute" }, -- cgit v1.2.3 From 4847a894bf7d4852325c3f1ea4bb4890054a1f66 Mon Sep 17 00:00:00 2001 From: Ondrej Zajicek Date: Sun, 26 Oct 2008 22:45:09 +0100 Subject: Implementation of route reflection for BGP --- filter/filter.h | 1 + nest/a-set.c | 8 +- nest/attrs.h | 3 + proto/bgp/attrs.c | 223 +++++++++++++++++++++++++++++++++++++--------------- proto/bgp/bgp.c | 9 +++ proto/bgp/bgp.h | 10 ++- proto/bgp/config.Y | 6 +- proto/bgp/packets.c | 6 +- 8 files changed, 195 insertions(+), 71 deletions(-) (limited to 'proto/bgp/packets.c') diff --git a/filter/filter.h b/filter/filter.h index 04a26236..f71e54d3 100644 --- a/filter/filter.h +++ b/filter/filter.h @@ -11,6 +11,7 @@ #include "lib/resource.h" #include "lib/ip.h" +#include "nest/route.h" #include "nest/attrs.h" struct f_inst { /* Instruction */ diff --git a/nest/a-set.c b/nest/a-set.c index 44407141..69c090b7 100644 --- a/nest/a-set.c +++ b/nest/a-set.c @@ -40,10 +40,12 @@ int_set_format(struct adata *set, byte *buf, unsigned int size) struct adata * int_set_add(struct linpool *pool, struct adata *list, u32 val) { - struct adata *res = lp_alloc(pool, list->length + sizeof(struct adata) + 4); - res->length = list->length+4; + int len = list ? list->length : 0; + struct adata *res = lp_alloc(pool, len + sizeof(struct adata) + 4); + res->length = len + 4; * (u32 *) res->data = val; - memcpy((char *) res->data + 4, list->data, list->length); + if (list) + memcpy((char *) res->data + 4, list->data, list->length); return res; } diff --git a/nest/attrs.h b/nest/attrs.h index f63f2e45..aaa5f4a2 100644 --- a/nest/attrs.h +++ b/nest/attrs.h @@ -47,4 +47,7 @@ struct adata *int_set_add(struct linpool *pool, struct adata *list, u32 val); int int_set_contains(struct adata *list, u32 val); struct adata *int_set_del(struct linpool *pool, struct adata *list, u32 val); +static inline int int_set_get_size(struct adata *list) +{ return list->length / 4; } + #endif diff --git a/proto/bgp/attrs.c b/proto/bgp/attrs.c index 48cb9dd5..a42a4880 100644 --- a/proto/bgp/attrs.c +++ b/proto/bgp/attrs.c @@ -104,13 +104,19 @@ bgp_check_next_hop(struct bgp_proto *p UNUSED, byte *a, int len) } static int -bgp_check_aggregator(struct bgp_proto *p UNUSED, UNUSED byte *a, int len) +bgp_check_aggregator(struct bgp_proto *p, UNUSED byte *a, int len) { int exp_len = (bgp_as4_support && p->as4_support) ? 8 : 6; return (len == exp_len) ? 0 : 5; } +static int +bgp_check_cluster_list(struct bgp_proto *p UNUSED, UNUSED byte *a, int len) +{ + return ((len % 4) == 0) ? 0 : 5; +} + static int bgp_check_reach_nlri(struct bgp_proto *p UNUSED, byte *a UNUSED, int len UNUSED) { @@ -150,8 +156,10 @@ static struct attr_desc bgp_attr_table[] = { bgp_check_aggregator, NULL }, { "community", -1, BAF_OPTIONAL | BAF_TRANSITIVE, EAF_TYPE_INT_SET, 1, /* BA_COMMUNITY */ NULL, NULL }, - { NULL, }, /* BA_ORIGINATOR_ID */ - { NULL, }, /* BA_CLUSTER_LIST */ + { "originator_id", 4, BAF_OPTIONAL, EAF_TYPE_INT, 0, /* BA_ORIGINATOR_ID */ + NULL, NULL }, + { "cluster_list", -1, BAF_OPTIONAL, EAF_TYPE_INT_SET, 0, /* BA_CLUSTER_LIST */ + bgp_check_cluster_list, NULL }, { NULL, }, /* BA_DPA */ { NULL, }, /* BA_ADVERTISER */ { NULL, }, /* BA_RCID_PATH */ @@ -173,35 +181,52 @@ static struct attr_desc bgp_attr_table[] = { #define ATTR_KNOWN(code) ((code) < ARRAY_SIZE(bgp_attr_table) && bgp_attr_table[code].name) -static byte * -bgp_set_attr(eattr *e, struct linpool *pool, unsigned attr, unsigned val) +static inline struct adata * +bgp_alloc_adata(struct linpool *pool, unsigned len) +{ + struct adata *ad = lp_alloc(pool, sizeof(struct adata) + len); + ad->length = len; + return ad; +} + +static void +bgp_set_attr(eattr *e, unsigned attr, uintptr_t val) { ASSERT(ATTR_KNOWN(attr)); e->id = EA_CODE(EAP_BGP, attr); e->type = bgp_attr_table[attr].type; e->flags = bgp_attr_table[attr].expected_flags; if (e->type & EAF_EMBEDDED) - { - e->u.data = val; - return NULL; - } + e->u.data = val; else - { - e->u.ptr = lp_alloc(pool, sizeof(struct adata) + val); - e->u.ptr->length = val; - return e->u.ptr->data; - } + e->u.ptr = (struct adata *) val; } -byte * -bgp_attach_attr(ea_list **to, struct linpool *pool, unsigned attr, unsigned val) +static byte * +bgp_set_attr_wa(eattr *e, struct linpool *pool, unsigned attr, unsigned len) +{ + struct adata *ad = bgp_alloc_adata(pool, len); + bgp_set_attr(e, attr, (uintptr_t) ad); + return ad->data; +} + +void +bgp_attach_attr(ea_list **to, struct linpool *pool, unsigned attr, uintptr_t val) { ea_list *a = lp_alloc(pool, sizeof(ea_list) + sizeof(eattr)); a->next = *to; *to = a; a->flags = EALF_SORTED; a->count = 1; - return bgp_set_attr(a->attrs, pool, attr, val); + bgp_set_attr(a->attrs, attr, val); +} + +byte * +bgp_attach_attr_wa(ea_list **to, struct linpool *pool, unsigned attr, unsigned len) +{ + struct adata *ad = bgp_alloc_adata(pool, len); + bgp_attach_attr(to, pool, attr, (uintptr_t) ad); + return ad->data; } static int @@ -713,6 +738,7 @@ bgp_rt_notify(struct proto *P, net *n, rte *new, rte *old UNUSED, ea_list *attrs bgp_schedule_packet(p->conn, PKT_UPDATE); } + static int bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *pool) { @@ -725,14 +751,14 @@ bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p ea->flags = EALF_SORTED; ea->count = 4; - bgp_set_attr(ea->attrs, pool, BA_ORIGIN, + bgp_set_attr(ea->attrs, BA_ORIGIN, ((rta->source == RTS_OSPF_EXT1) || (rta->source == RTS_OSPF_EXT2)) ? ORIGIN_INCOMPLETE : ORIGIN_IGP); if (p->is_internal) - bgp_set_attr(ea->attrs+1, pool, BA_AS_PATH, 0); + bgp_set_attr_wa(ea->attrs+1, pool, BA_AS_PATH, 0); else { - z = bgp_set_attr(ea->attrs+1, pool, BA_AS_PATH, bgp_as4_support ? 6 : 4); + z = bgp_set_attr_wa(ea->attrs+1, pool, BA_AS_PATH, bgp_as4_support ? 6 : 4); z[0] = AS_PATH_SEQUENCE; z[1] = 1; /* 1 AS */ @@ -742,7 +768,7 @@ bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p put_u16(z+2, p->local_as); } - z = bgp_set_attr(ea->attrs+2, pool, BA_NEXT_HOP, sizeof(ip_addr)); + z = bgp_set_attr_wa(ea->attrs+2, pool, BA_NEXT_HOP, sizeof(ip_addr)); if (p->cf->next_hop_self || !p->is_internal || rta->dest != RTD_ROUTER) @@ -755,34 +781,55 @@ bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p else *(ip_addr *)z = e->attrs->gw; - bgp_set_attr(ea->attrs+3, pool, BA_LOCAL_PREF, 0); + bgp_set_attr(ea->attrs+3, BA_LOCAL_PREF, 0); return 0; /* Leave decision to the filters */ } -static ea_list * -bgp_path_prepend(struct linpool *pool, eattr *a, ea_list *old, int as) + +static inline int +bgp_as_path_loopy(struct bgp_proto *p, rta *a) +{ + eattr *e = ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)); + return (e && as_path_is_member(e->u.ptr, p->local_as)); +} + +static inline int +bgp_originator_id_loopy(struct bgp_proto *p, rta *a) { - struct ea_list *e = lp_alloc(pool, sizeof(ea_list) + sizeof(eattr)); - struct adata *olda = a->u.ptr; - - e->next = old; - e->flags = EALF_SORTED; - e->count = 1; - e->attrs[0].id = EA_CODE(EAP_BGP, BA_AS_PATH); - e->attrs[0].flags = BAF_TRANSITIVE; - e->attrs[0].type = EAF_TYPE_AS_PATH; - e->attrs[0].u.ptr = as_path_prepend(pool, olda, as); - return e; + eattr *e = ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_ORIGINATOR_ID)); + return (e && (e->u.data == p->local_id)); +} + +static inline int +bgp_cluster_list_loopy(struct bgp_proto *p, rta *a) +{ + eattr *e = ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_CLUSTER_LIST)); + return (e && p->rr_client && int_set_contains(e->u.ptr, p->rr_cluster_id)); +} + + +static inline void +bgp_path_prepend(rte *e, ea_list **attrs, struct linpool *pool, u32 as) +{ + eattr *a = ea_find(e->attrs->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)); + bgp_attach_attr(attrs, pool, BA_AS_PATH, (uintptr_t) as_path_prepend(pool, a->u.ptr, as)); +} + +static inline void +bgp_cluster_list_prepend(rte *e, ea_list **attrs, struct linpool *pool, u32 cid) +{ + eattr *a = ea_find(e->attrs->eattrs, EA_CODE(EAP_BGP, BA_CLUSTER_LIST)); + bgp_attach_attr(attrs, pool, BA_CLUSTER_LIST, (uintptr_t) int_set_add(pool, a ? a->u.ptr : NULL, cid)); } static int -bgp_update_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *pool) +bgp_update_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *pool, int rr) { eattr *a; - if (!p->is_internal && (a = ea_find(e->attrs->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)))) - *attrs = bgp_path_prepend(pool, a, *attrs, p->local_as); + if (!p->is_internal) + bgp_path_prepend(e, attrs, pool, p->local_as); a = ea_find(e->attrs->eattrs, EA_CODE(EAP_BGP, BA_NEXT_HOP)); if (a && (p->is_internal || (!p->is_internal && e->attrs->iface == p->neigh->iface))) @@ -792,7 +839,24 @@ bgp_update_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p else { /* Need to create new one */ - *(ip_addr *) bgp_attach_attr(attrs, pool, BA_NEXT_HOP, sizeof(ip_addr)) = p->local_addr; + bgp_attach_attr_ip(attrs, pool, BA_NEXT_HOP, p->local_addr); + } + + if (rr) + { + /* Handling route reflection, RFC 4456 */ + struct bgp_proto *src = (struct bgp_proto *) e->attrs->proto; + + a = ea_find(e->attrs->eattrs, EA_CODE(EAP_BGP, BA_ORIGINATOR_ID)); + if (!a) + bgp_attach_attr(attrs, pool, BA_ORIGINATOR_ID, src->remote_id); + + /* We attach proper cluster ID according to whether the route is entering or leaving the cluster */ + bgp_cluster_list_prepend(e, attrs, pool, src->rr_client ? src->rr_cluster_id : p->rr_cluster_id); + + /* Two RR clients with different cluster ID, hmmm */ + if (src->rr_client && p->rr_client && (src->rr_cluster_id != p->rr_cluster_id)) + bgp_cluster_list_prepend(e, attrs, pool, p->rr_cluster_id); } return 0; /* Leave decision to the filters */ @@ -809,9 +873,22 @@ bgp_import_control(struct proto *P, rte **new, ea_list **attrs, struct linpool * return -1; if (new_bgp) { + /* We should check here for cluster list loop, because the receiving BGP instance + might have different cluster ID */ + if (bgp_cluster_list_loopy(p, e->attrs)) + return -1; + if (p->local_as == new_bgp->local_as && p->is_internal && new_bgp->is_internal) - return -1; /* Don't redistribute internal routes with IBGP */ - return bgp_update_attrs(p, e, attrs, pool); + { + /* Redistribution of internal routes with IBGP */ + if (p->rr_client || new_bgp->rr_client) + /* Route reflection, RFC 4456 */ + return bgp_update_attrs(p, e, attrs, pool, 1); + else + return -1; + } + else + return bgp_update_attrs(p, e, attrs, pool, 0); } else return bgp_create_attrs(p, e, attrs, pool); @@ -835,7 +912,7 @@ bgp_rte_better(rte *new, rte *old) if (n < o) return 0; - /* Use AS path lengths */ + /* RFC 4271 9.1.2.2. a) Use AS path lengths */ if (new_bgp->cf->compare_path_lengths || old_bgp->cf->compare_path_lengths) { x = ea_find(new->attrs->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)); @@ -848,7 +925,7 @@ bgp_rte_better(rte *new, rte *old) return 0; } - /* Use origins */ + /* RFC 4271 9.1.2.2. b) Use origins */ x = ea_find(new->attrs->eattrs, EA_CODE(EAP_BGP, BA_ORIGIN)); y = ea_find(old->attrs->eattrs, EA_CODE(EAP_BGP, BA_ORIGIN)); n = x ? x->u.data : ORIGIN_INCOMPLETE; @@ -858,7 +935,7 @@ bgp_rte_better(rte *new, rte *old) if (n > o) return 0; - /* Compare MED's */ + /* RFC 4271 9.1.2.2. c) Compare MED's */ x = ea_find(new->attrs->eattrs, EA_CODE(EAP_BGP, BA_MULTI_EXIT_DISC)); y = ea_find(old->attrs->eattrs, EA_CODE(EAP_BGP, BA_MULTI_EXIT_DISC)); n = x ? x->u.data : new_bgp->cf->default_med; @@ -868,24 +945,41 @@ bgp_rte_better(rte *new, rte *old) if (n > o) return 0; - /* A tie breaking procedure according to RFC 1771, section 9.1.2.1 */ - /* We don't have interior distances */ - /* We prefer external peers */ + /* RFC 4271 9.1.2.2. d) Prefer external peers */ if (new_bgp->is_internal > old_bgp->is_internal) return 0; if (new_bgp->is_internal < old_bgp->is_internal) return 1; - /* Finally we compare BGP identifiers */ - return (new_bgp->remote_id < old_bgp->remote_id); -} -static int -bgp_path_loopy(struct bgp_proto *p, eattr *a) -{ - return as_path_is_member(a->u.ptr, p->local_as); -} + /* Skipping RFC 4271 9.1.2.2. e) */ + /* We don't have interior distances */ + + /* RFC 4456 9. b) Compare cluster list lengths */ + x = ea_find(new->attrs->eattrs, EA_CODE(EAP_BGP, BA_CLUSTER_LIST)); + y = ea_find(old->attrs->eattrs, EA_CODE(EAP_BGP, BA_CLUSTER_LIST)); + n = x ? int_set_get_size(x->u.ptr) : 0; + o = y ? int_set_get_size(y->u.ptr) : 0; + if (n < o) + return 1; + if (n > o) + return 0; + + /* RFC 4271 9.1.2.2. f) Compare BGP identifiers */ + /* RFC 4456 9. a) Use ORIGINATOR_ID instead of local neighor ID */ + x = ea_find(new->attrs->eattrs, EA_CODE(EAP_BGP, BA_ORIGINATOR_ID)); + y = ea_find(old->attrs->eattrs, EA_CODE(EAP_BGP, BA_ORIGINATOR_ID)); + n = x ? x->u.data : new_bgp->remote_id; + o = y ? y->u.data : old_bgp->remote_id; + if (n < o) + return 1; + if (n > o) + return 0; + /* RFC 4271 9.1.2.2. g) Compare peer IP adresses */ + return (ipa_compare(new_bgp->cf->remote_ip, old_bgp->cf->remote_ip) < 0); +} + static struct adata * bgp_aggregator_convert_to_new(struct adata *old, struct linpool *pool) { @@ -916,7 +1010,7 @@ bgp_merge_as_paths(struct adata *old2, struct adata *old4, int req_as, struct li } -/* Reconstruct 4B AS_PATH and AGGREGATOR according to RFC4893 4.2.3 */ +/* Reconstruct 4B AS_PATH and AGGREGATOR according to RFC 4893 4.2.3 */ static void bgp_reconstruct_4b_atts(struct bgp_proto *p, rta *a, struct linpool *pool) { @@ -1159,18 +1253,23 @@ bgp_decode_attrs(struct bgp_conn *conn, byte *attr, unsigned int len, struct lin bgp_remove_as4_attrs(bgp, a); /* If the AS path attribute contains our AS, reject the routes */ - e = ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)); - if (e && bgp_path_loopy(bgp, e)) - { - DBG("BGP: Path loop!\n"); - return NULL; - } + if (bgp_as_path_loopy(bgp, a)) + goto loop; + + /* Two checks for IBGP loops caused by route reflection, RFC 4456 */ + if (bgp_originator_id_loopy(bgp, a) || + bgp_cluster_list_loopy(bgp, a)) + goto loop; /* If there's no local preference, define one */ if (!(seen[0] && (1 << BA_LOCAL_PREF))) bgp_attach_attr(&a->eattrs, pool, BA_LOCAL_PREF, 0); return a; +loop: + DBG("BGP: Path loop!\n"); + return NULL; + malformed: bgp_error(conn, 3, 1, NULL, 0); return NULL; diff --git a/proto/bgp/bgp.c b/proto/bgp/bgp.c index e1f5ec02..5fa12492 100644 --- a/proto/bgp/bgp.c +++ b/proto/bgp/bgp.c @@ -485,6 +485,13 @@ bgp_start_locked(struct object_lock *lock) p->local_id = cf->c.global->router_id; p->next_hop = cf->multihop ? cf->multihop_via : cf->remote_ip; p->neigh = neigh_find(&p->p, &p->next_hop, NEF_STICKY); + + if (cf->rr_client) + { + p->rr_cluster_id = cf->rr_cluster_id ? cf->rr_cluster_id : p->local_id; + p->rr_client = cf->rr_client; + } + if (!p->neigh) { log(L_ERR "%s: Invalid next hop %I", p->p.name, p->next_hop); @@ -633,6 +640,8 @@ bgp_check(struct bgp_config *c) cf_error("Local AS number out of range"); if (!bgp_as4_support && (c->remote_as > 0xFFFF)) cf_error("Neighbor AS number out of range"); + if ((c->local_as != c->remote_as) && (c->rr_client)) + cf_error("Only internal neighbor can be RR client"); } static void diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h index 93383244..af3c5c5a 100644 --- a/proto/bgp/bgp.h +++ b/proto/bgp/bgp.h @@ -25,6 +25,8 @@ struct bgp_config { int compare_path_lengths; /* Use path lengths when selecting best route */ u32 default_local_pref; /* Default value for LOCAL_PREF attribute */ u32 default_med; /* Default value for MULTI_EXIT_DISC attribute */ + u32 rr_cluster_id; /* Route reflector cluster ID, if different from local ID */ + int rr_client; /* Whether neighbor is RR client of me */ unsigned connect_retry_time; unsigned hold_time, initial_hold_time; unsigned keepalive_time; @@ -60,6 +62,8 @@ struct bgp_proto { int as4_support; /* Peer supports 4B AS numbers [RFC4893] */ u32 local_id; /* BGP identifier of this router */ u32 remote_id; /* BGP identifier of the neighbor */ + u32 rr_cluster_id; /* Route reflector cluster ID */ + int rr_client; /* Whether neighbor is RR client of me */ struct bgp_conn *conn; /* Connection we have established */ struct bgp_conn outgoing_conn; /* Outgoing connection we're working with */ struct bgp_conn incoming_conn; /* Incoming connection we have neither accepted nor rejected yet */ @@ -121,7 +125,8 @@ void bgp_close_conn(struct bgp_conn *c); /* attrs.c */ -byte *bgp_attach_attr(struct ea_list **to, struct linpool *, unsigned attr, unsigned val); +void bgp_attach_attr(struct ea_list **to, struct linpool *pool, unsigned attr, uintptr_t val); +byte *bgp_attach_attr_wa(struct ea_list **to, struct linpool *pool, unsigned attr, unsigned len); struct rta *bgp_decode_attrs(struct bgp_conn *conn, byte *a, unsigned int len, struct linpool *pool, int mandatory); int bgp_get_attr(struct eattr *e, byte *buf); int bgp_rte_better(struct rte *, struct rte *); @@ -132,6 +137,9 @@ unsigned int bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int void bgp_free_bucket(struct bgp_proto *p, struct bgp_bucket *buck); void bgp_get_route_info(struct rte *, byte *buf, struct ea_list *attrs); +inline static void bgp_attach_attr_ip(struct ea_list **to, struct linpool *pool, unsigned attr, ip_addr a) +{ *(ip_addr *) bgp_attach_attr_wa(to, pool, attr, sizeof(ip_addr)) = a; } + /* packets.c */ void bgp_schedule_packet(struct bgp_conn *conn, int type); diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y index 580c008f..b23b66cf 100644 --- a/proto/bgp/config.Y +++ b/proto/bgp/config.Y @@ -20,7 +20,8 @@ CF_KEYWORDS(BGP, LOCAL, NEIGHBOR, AS, HOLD, TIME, CONNECT, RETRY, KEEPALIVE, MULTIHOP, STARTUP, VIA, NEXT, HOP, SELF, DEFAULT, PATH, METRIC, ERROR, START, DELAY, FORGET, WAIT, DISABLE, AFTER, BGP_PATH, BGP_LOCAL_PREF, BGP_MED, BGP_ORIGIN, BGP_NEXT_HOP, - BGP_ATOMIC_AGGR, BGP_AGGREGATOR, BGP_COMMUNITY, SOURCE, ADDRESS, PASSWORD) + BGP_ATOMIC_AGGR, BGP_AGGREGATOR, BGP_COMMUNITY, SOURCE, ADDRESS, + PASSWORD, RR, CLIENT, CLUSTER, ID) CF_GRAMMAR @@ -38,7 +39,6 @@ bgp_proto_start: proto_start BGP { BGP_CFG->error_amnesia_time = 300; BGP_CFG->error_delay_time_min = 60; BGP_CFG->error_delay_time_max = 300; - BGP_CFG->password = NULL; } ; @@ -52,6 +52,8 @@ bgp_proto: BGP_CFG->remote_ip = $3; BGP_CFG->remote_as = $5; } + | bgp_proto RR CLUSTER ID expr ';' { BGP_CFG->rr_cluster_id = $5; } + | bgp_proto RR CLIENT ';' { BGP_CFG->rr_client = 1; } | bgp_proto HOLD TIME expr ';' { BGP_CFG->hold_time = $4; } | bgp_proto STARTUP HOLD TIME expr ';' { BGP_CFG->initial_hold_time = $5; } | bgp_proto CONNECT RETRY TIME expr ';' { BGP_CFG->connect_retry_time = $5; } diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c index 0dd920e4..8a352c68 100644 --- a/proto/bgp/packets.c +++ b/proto/bgp/packets.c @@ -193,7 +193,7 @@ bgp_create_update(struct bgp_conn *conn, byte *buf) if ((buck = p->withdraw_bucket) && !EMPTY_LIST(buck->prefixes)) { DBG("Withdrawn routes:\n"); - tmp = bgp_attach_attr(&ea, bgp_linpool, BA_MP_UNREACH_NLRI, remains-8); + tmp = bgp_attach_attr_wa(&ea, bgp_linpool, BA_MP_UNREACH_NLRI, remains-8); *tmp++ = 0; *tmp++ = BGP_AF_IPV6; *tmp++ = 1; @@ -218,7 +218,7 @@ bgp_create_update(struct bgp_conn *conn, byte *buf) size = bgp_encode_attrs(p, w, buck->eattrs, 1024); w += size; remains -= size; - tstart = tmp = bgp_attach_attr(&ea, bgp_linpool, BA_MP_REACH_NLRI, remains-8); + tstart = tmp = bgp_attach_attr_wa(&ea, bgp_linpool, BA_MP_REACH_NLRI, remains-8); *tmp++ = 0; *tmp++ = BGP_AF_IPV6; *tmp++ = 1; @@ -702,7 +702,7 @@ bgp_do_rx_update(struct bgp_conn *conn, /* Create fake NEXT_HOP attribute */ if (len < 1 || (*x != 16 && *x != 32) || len < *x + 2) goto bad; - memcpy(bgp_attach_attr(&a0->eattrs, bgp_linpool, BA_NEXT_HOP, 16), x+1, 16); + bgp_attach_attr_ip(&a0->eattrs, bgp_linpool, BA_NEXT_HOP, x[1]); len -= *x + 2; x += *x + 1; -- cgit v1.2.3 From ba5ed6f3e4eb4b2899cdad08e2edb99063bfbcee Mon Sep 17 00:00:00 2001 From: Ondrej Zajicek Date: Sun, 26 Oct 2008 22:48:02 +0100 Subject: Implementation of an option for disabling AS4 support per BGP instance. --- proto/bgp/attrs.c | 14 +++++++------- proto/bgp/bgp.c | 6 ++++-- proto/bgp/bgp.h | 2 ++ proto/bgp/config.Y | 6 ++++-- proto/bgp/packets.c | 5 +++-- 5 files changed, 20 insertions(+), 13 deletions(-) (limited to 'proto/bgp/packets.c') diff --git a/proto/bgp/attrs.c b/proto/bgp/attrs.c index a42a4880..0fcd1ce1 100644 --- a/proto/bgp/attrs.c +++ b/proto/bgp/attrs.c @@ -73,13 +73,13 @@ bgp_check_path(byte *a, int len, int bs, int errcode) static int bgp_check_as_path(struct bgp_proto *p, byte *a, int len) { - return bgp_check_path(a, len, (bgp_as4_support && p->as4_support) ? 4 : 2, 11); + return bgp_check_path(a, len, p->as4_session ? 4 : 2, 11); } static int bgp_check_as4_path(struct bgp_proto *p, byte *a, int len) { - if (bgp_as4_support && (! p->as4_support)) + if (bgp_as4_support && (! p->as4_session)) return bgp_check_path(a, len, 4, 9); else return 0; @@ -106,7 +106,7 @@ bgp_check_next_hop(struct bgp_proto *p UNUSED, byte *a, int len) static int bgp_check_aggregator(struct bgp_proto *p, UNUSED byte *a, int len) { - int exp_len = (bgp_as4_support && p->as4_support) ? 8 : 6; + int exp_len = p->as4_session ? 8 : 6; return (len == exp_len) ? 0 : 5; } @@ -344,7 +344,7 @@ bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int remains) * we have to convert our 4B AS_PATH to 2B AS_PATH and send our AS_PATH * as optional AS4_PATH attribute. */ - if ((code == BA_AS_PATH) && bgp_as4_support && (! p->as4_support)) + if ((code == BA_AS_PATH) && bgp_as4_support && (! p->as4_session)) { len = a->u.ptr->length; @@ -384,7 +384,7 @@ bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int remains) } /* The same issue with AGGREGATOR attribute */ - if ((code == BA_AGGREGATOR) && bgp_as4_support && (! p->as4_support)) + if ((code == BA_AGGREGATOR) && bgp_as4_support && (! p->as4_session)) { int new_used; @@ -1082,7 +1082,7 @@ bgp_remove_as4_attrs(struct bgp_proto *p, rta *a) if ((fid == id1) || (fid == id2)) { *el = (*el)->next; - if (p->as4_support) + if (p->as4_session) log(L_WARN "BGP: Unexpected AS4_* attributes received"); } else @@ -1246,7 +1246,7 @@ bgp_decode_attrs(struct bgp_conn *conn, byte *attr, unsigned int len, struct lin /* When receiving attributes from non-AS4-aware BGP speaker, * we have to reconstruct 4B AS_PATH and AGGREGATOR attributes */ - if (bgp_as4_support && (! bgp->as4_support)) + if (bgp_as4_support && (! bgp->as4_session)) bgp_reconstruct_4b_atts(bgp, a, pool); if (bgp_as4_support) diff --git a/proto/bgp/bgp.c b/proto/bgp/bgp.c index 5fa12492..0d580be1 100644 --- a/proto/bgp/bgp.c +++ b/proto/bgp/bgp.c @@ -636,9 +636,11 @@ bgp_check(struct bgp_config *c) cf_error("Local AS number must be set"); if (!c->remote_as) cf_error("Neighbor must be configured"); - if (!bgp_as4_support && (c->local_as > 0xFFFF)) + if (!bgp_as4_support && c->enable_as4) + cf_error("AS4 support disabled globbaly"); + if (!c->enable_as4 && (c->local_as > 0xFFFF)) cf_error("Local AS number out of range"); - if (!bgp_as4_support && (c->remote_as > 0xFFFF)) + if (!c->enable_as4 && (c->remote_as > 0xFFFF)) cf_error("Neighbor AS number out of range"); if ((c->local_as != c->remote_as) && (c->rr_client)) cf_error("Only internal neighbor can be RR client"); diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h index af3c5c5a..1d67e336 100644 --- a/proto/bgp/bgp.h +++ b/proto/bgp/bgp.h @@ -25,6 +25,7 @@ struct bgp_config { int compare_path_lengths; /* Use path lengths when selecting best route */ u32 default_local_pref; /* Default value for LOCAL_PREF attribute */ u32 default_med; /* Default value for MULTI_EXIT_DISC attribute */ + int enable_as4; /* Enable local support for 4B AS numbers [RFC4893] */ u32 rr_cluster_id; /* Route reflector cluster ID, if different from local ID */ int rr_client; /* Whether neighbor is RR client of me */ unsigned connect_retry_time; @@ -60,6 +61,7 @@ struct bgp_proto { u32 local_as, remote_as; int is_internal; /* Internal BGP connection (local_as == remote_as) */ int as4_support; /* Peer supports 4B AS numbers [RFC4893] */ + int as4_session; /* Session uses 4B AS numbers in AS_PATH (both sides support it) */ u32 local_id; /* BGP identifier of this router */ u32 remote_id; /* BGP identifier of the neighbor */ u32 rr_cluster_id; /* Route reflector cluster ID */ diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y index b23b66cf..d7bba575 100644 --- a/proto/bgp/config.Y +++ b/proto/bgp/config.Y @@ -18,10 +18,10 @@ CF_DECLS CF_KEYWORDS(BGP, LOCAL, NEIGHBOR, AS, HOLD, TIME, CONNECT, RETRY, KEEPALIVE, MULTIHOP, STARTUP, VIA, NEXT, HOP, SELF, DEFAULT, PATH, METRIC, - ERROR, START, DELAY, FORGET, WAIT, DISABLE, AFTER, + ERROR, START, DELAY, FORGET, WAIT, ENABLE, DISABLE, AFTER, BGP_PATH, BGP_LOCAL_PREF, BGP_MED, BGP_ORIGIN, BGP_NEXT_HOP, BGP_ATOMIC_AGGR, BGP_AGGREGATOR, BGP_COMMUNITY, SOURCE, ADDRESS, - PASSWORD, RR, CLIENT, CLUSTER, ID) + PASSWORD, RR, CLIENT, CLUSTER, ID, AS4) CF_GRAMMAR @@ -39,6 +39,7 @@ bgp_proto_start: proto_start BGP { BGP_CFG->error_amnesia_time = 300; BGP_CFG->error_delay_time_min = 60; BGP_CFG->error_delay_time_max = 300; + BGP_CFG->enable_as4 = bgp_as4_support; } ; @@ -68,6 +69,7 @@ bgp_proto: | bgp_proto ERROR FORGET TIME expr ';' { BGP_CFG->error_amnesia_time = $5; } | bgp_proto ERROR WAIT TIME expr ',' expr ';' { BGP_CFG->error_delay_time_min = $5; BGP_CFG->error_delay_time_max = $7; } | bgp_proto DISABLE AFTER ERROR bool ';' { BGP_CFG->disable_after_error = $5; } + | bgp_proto ENABLE AS4 bool ';' { BGP_CFG->enable_as4 = $4; } | bgp_proto PASSWORD TEXT ';' { BGP_CFG->password = $3; } ; diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c index 8a352c68..c18c6e42 100644 --- a/proto/bgp/packets.c +++ b/proto/bgp/packets.c @@ -73,7 +73,7 @@ bgp_create_open(struct bgp_conn *conn, byte *buf) #ifdef IPV6 cap = bgp_put_cap_ipv6(conn, cap); #endif - if (bgp_as4_support) + if (p->cf->enable_as4) cap = bgp_put_cap_as4(conn, cap); cap_len = cap - buf - 12; @@ -407,7 +407,8 @@ bgp_parse_capabilities(struct bgp_conn *conn, byte *opt, int len) if (cl != 4) goto err; p->as4_support = 1; - if (bgp_as4_support) + p->as4_session = p->cf->enable_as4; + if (p->as4_session) conn->advertised_as = get_u32(opt + 2); break; -- cgit v1.2.3