2 * BIRD -- BGP Attributes
4 * (c) 2000 Martin Mares <mj@ucw.cz>
5 * (c) 2008--2016 Ondrej Zajicek <santiago@crfreenet.org>
6 * (c) 2008--2016 CZ.NIC z.s.p.o.
8 * Can be freely distributed and used under the terms of the GNU GPL.
15 #include "nest/bird.h"
16 #include "nest/iface.h"
17 #include "nest/protocol.h"
18 #include "nest/route.h"
19 #include "nest/attrs.h"
20 #include "conf/conf.h"
21 #include "lib/resource.h"
22 #include "lib/string.h"
23 #include "lib/unaligned.h"
28 * UPDATE message error handling
30 * All checks from RFC 4271 6.3 are done as specified with these exceptions:
31 * - The semantic check of an IP address from NEXT_HOP attribute is missing.
32 * - Checks of some optional attribute values are missing.
33 * - Syntactic and semantic checks of NLRIs (done in DECODE_PREFIX())
34 * are probably inadequate.
36 * Loop detection based on AS_PATH causes updates to be withdrawn. RFC
37 * 4271 does not explicitly specify the behavior in that case.
39 * Loop detection related to route reflection (based on ORIGINATOR_ID
40 * and CLUSTER_LIST) causes updates to be withdrawn. RFC 4456 8
41 * specifies that such updates should be ignored, but that is generally
44 * BGP attribute table has several hooks:
46 * export - Hook that validates and normalizes attribute during export phase.
47 * Receives eattr, may modify it (e.g., sort community lists for canonical
48 * representation), UNSET() it (e.g., skip empty lists), or WITHDRAW() it if
49 * necessary. May assume that eattr has value valid w.r.t. its type, but may be
50 * invalid w.r.t. BGP constraints. Optional.
52 * encode - Hook that converts internal representation to external one during
53 * packet writing. Receives eattr and puts it in the buffer (including attribute
54 * header). Returns number of bytes, or -1 if not enough space. May assume that
55 * eattr has value valid w.r.t. its type and validated by export hook. Mandatory
56 * for all known attributes that exist internally after export phase (i.e., all
57 * except pseudoattributes MP_(UN)REACH_NLRI).
59 * decode - Hook that converts external representation to internal one during
60 * packet parsing. Receives attribute data in buffer, validates it and adds
61 * attribute to ea_list. If data are invalid, steps DISCARD(), WITHDRAW() or
62 * bgp_parse_error() may be used to escape. Mandatory for all known attributes.
64 * format - Optional hook that converts eattr to textual representation.
68 struct bgp_attr_desc
{
72 void (*export
)(struct bgp_export_state
*s
, eattr
*a
);
73 int (*encode
)(struct bgp_write_state
*s
, eattr
*a
, byte
*buf
, uint size
);
74 void (*decode
)(struct bgp_parse_state
*s
, uint code
, uint flags
, byte
*data
, uint len
, ea_list
**to
);
75 void (*format
)(eattr
*ea
, byte
*buf
, uint size
);
78 static const struct bgp_attr_desc bgp_attr_table
[];
80 static inline int bgp_attr_known(uint code
);
83 bgp_set_attr(ea_list
**attrs
, struct linpool
*pool
, uint code
, uint flags
, uintptr_t val
)
85 ASSERT(bgp_attr_known(code
));
90 EA_CODE(PROTOCOL_BGP
, code
),
92 bgp_attr_table
[code
].type
,
99 #define REPORT(msg, args...) \
100 ({ log(L_REMOTE "%s: " msg, s->proto->p.name, ## args); })
102 #define DISCARD(msg, args...) \
103 ({ REPORT(msg, ## args); return; })
105 #define WITHDRAW(msg, args...) \
106 ({ REPORT(msg, ## args); s->err_withdraw = 1; return; })
109 ({ a->type = EAF_TYPE_UNDEF; return; })
111 #define NEW_BGP "Discarding %s attribute received from AS4-aware neighbor"
112 #define BAD_EBGP "Discarding %s attribute received from EBGP neighbor"
113 #define BAD_LENGTH "Malformed %s attribute - invalid length (%u)"
114 #define BAD_VALUE "Malformed %s attribute - invalid value (%u)"
115 #define NO_MANDATORY "Missing mandatory %s attribute"
119 bgp_put_attr_hdr3(byte
*buf
, uint code
, uint flags
, uint len
)
128 bgp_put_attr_hdr4(byte
*buf
, uint code
, uint flags
, uint len
)
130 *buf
++ = flags
| BAF_EXT_LEN
;
137 bgp_put_attr_hdr(byte
*buf
, uint code
, uint flags
, uint len
)
140 return bgp_put_attr_hdr3(buf
, code
, flags
, len
);
142 return bgp_put_attr_hdr4(buf
, code
, flags
, len
);
146 bgp_encode_u8(struct bgp_write_state
*s UNUSED
, eattr
*a
, byte
*buf
, uint size
)
151 bgp_put_attr_hdr3(buf
, EA_ID(a
->id
), a
->flags
, 1);
158 bgp_encode_u32(struct bgp_write_state
*s UNUSED
, eattr
*a
, byte
*buf
, uint size
)
163 bgp_put_attr_hdr3(buf
, EA_ID(a
->id
), a
->flags
, 4);
164 put_u32(buf
+3, a
->u
.data
);
170 bgp_encode_u32s(struct bgp_write_state
*s UNUSED
, eattr
*a
, byte
*buf
, uint size
)
172 uint len
= a
->u
.ptr
->length
;
177 uint hdr
= bgp_put_attr_hdr(buf
, EA_ID(a
->id
), a
->flags
, len
);
178 put_u32s(buf
+ hdr
, (u32
*) a
->u
.ptr
->data
, len
/ 4);
184 bgp_put_attr(byte
*buf
, uint size
, uint code
, uint flags
, const byte
*data
, uint len
)
189 uint hdr
= bgp_put_attr_hdr(buf
, code
, flags
, len
);
190 memcpy(buf
+ hdr
, data
, len
);
196 bgp_encode_raw(struct bgp_write_state
*s UNUSED
, eattr
*a
, byte
*buf
, uint size
)
198 return bgp_put_attr(buf
, size
, EA_ID(a
->id
), a
->flags
, a
->u
.ptr
->data
, a
->u
.ptr
->length
);
207 bgp_aigp_valid(byte
*data
, uint len
, char *err
, uint elen
)
210 char *err_dsc
= NULL
;
213 #define BAD(DSC,VAL) ({ err_dsc = DSC; err_val = VAL; goto bad; })
217 BAD("TLV framing error", len
);
219 /* Process one TLV */
221 uint plen
= get_u16(pos
+ 1);
224 BAD("TLV framing error", plen
);
227 BAD("Bad TLV length", plen
);
229 if ((ptype
== BGP_AIGP_METRIC
) && (plen
!= 11))
230 BAD("Bad AIGP TLV length", plen
);
232 ADVANCE(pos
, len
, plen
);
240 if (bsnprintf(err
, elen
, "%s (%u) at %d", err_dsc
, err_val
, (int) (pos
- data
)) < 0)
247 bgp_aigp_get_tlv(const struct adata
*ad
, uint type
)
252 uint len
= ad
->length
;
253 const byte
*pos
= ad
->data
;
258 uint plen
= get_u16(pos
+ 1);
263 ADVANCE(pos
, len
, plen
);
269 static const struct adata
*
270 bgp_aigp_set_tlv(struct linpool
*pool
, const struct adata
*ad
, uint type
, byte
*data
, uint dlen
)
272 uint len
= ad
? ad
->length
: 0;
273 const byte
*pos
= ad
? ad
->data
: NULL
;
274 struct adata
*res
= lp_alloc_adata(pool
, len
+ 3 + dlen
);
275 byte
*dst
= res
->data
;
282 uint plen
= get_u16(pos
+ 1);
284 /* Find position for new TLV */
285 if ((ptype
>= type
) && !tlv
)
291 /* Skip first matching TLV, copy others */
292 if ((ptype
== type
) && !del
)
296 memcpy(dst
, pos
, plen
);
300 ADVANCE(pos
, len
, plen
);
310 put_u8(tlv
+ 0, type
);
311 put_u16(tlv
+ 1, 3 + dlen
);
312 memcpy(tlv
+ 3, data
, dlen
);
315 res
->length
= dst
- res
->data
;
321 bgp_aigp_get_metric(const struct adata
*ad
, u64 def
)
323 const byte
*b
= bgp_aigp_get_tlv(ad
, BGP_AIGP_METRIC
);
324 return b
? get_u64(b
+ 3) : def
;
327 static const struct adata
*
328 bgp_aigp_set_metric(struct linpool
*pool
, const struct adata
*ad
, u64 metric
)
331 put_u64(data
, metric
);
332 return bgp_aigp_set_tlv(pool
, ad
, BGP_AIGP_METRIC
, data
, 8);
336 bgp_total_aigp_metric_(rte
*e
, u64
*metric
, const struct adata
**ad
)
338 eattr
*a
= ea_find(e
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_AIGP
));
342 const byte
*b
= bgp_aigp_get_tlv(a
->u
.ptr
, BGP_AIGP_METRIC
);
346 u64 aigp
= get_u64(b
+ 3);
347 u64 step
= e
->attrs
->igp_metric
;
349 if (!rte_resolvable(e
) || (step
>= IGP_METRIC_UNKNOWN
))
356 *metric
= aigp
+ step
;
358 *metric
= BGP_AIGP_MAX
;
364 bgp_init_aigp_metric(rte
*e
, u64
*metric
, const struct adata
**ad
)
366 if (e
->attrs
->source
== RTS_BGP
)
369 *metric
= rt_get_igp_metric(e
);
371 return *metric
< IGP_METRIC_UNKNOWN
;
380 bgp_export_origin(struct bgp_export_state
*s
, eattr
*a
)
383 WITHDRAW(BAD_VALUE
, "ORIGIN", a
->u
.data
);
387 bgp_decode_origin(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
390 WITHDRAW(BAD_LENGTH
, "ORIGIN", len
);
393 WITHDRAW(BAD_VALUE
, "ORIGIN", data
[0]);
395 bgp_set_attr_u32(to
, s
->pool
, BA_ORIGIN
, flags
, data
[0]);
399 bgp_format_origin(eattr
*a
, byte
*buf
, uint size UNUSED
)
401 static const char *bgp_origin_names
[] = { "IGP", "EGP", "Incomplete" };
403 bsprintf(buf
, (a
->u
.data
<= 2) ? bgp_origin_names
[a
->u
.data
] : "?");
408 bgp_encode_as_path(struct bgp_write_state
*s
, eattr
*a
, byte
*buf
, uint size
)
410 const byte
*data
= a
->u
.ptr
->data
;
411 uint len
= a
->u
.ptr
->length
;
415 /* Prepare 16-bit AS_PATH (from 32-bit one) in a temporary buffer */
416 byte
*dst
= alloca(len
);
417 len
= as_path_32to16(dst
, data
, len
);
421 return bgp_put_attr(buf
, size
, BA_AS_PATH
, a
->flags
, data
, len
);
425 bgp_decode_as_path(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
427 struct bgp_proto
*p
= s
->proto
;
428 int as_length
= s
->as4_session
? 4 : 2;
429 int as_sets
= p
->cf
->allow_as_sets
;
430 int as_confed
= p
->cf
->confederation
&& p
->is_interior
;
433 if (!as_path_valid(data
, len
, as_length
, as_sets
, as_confed
, err
, sizeof(err
)))
434 WITHDRAW("Malformed AS_PATH attribute - %s", err
);
436 /* In some circumstances check for initial AS_CONFED_SEQUENCE; RFC 5065 5.0 */
437 if (p
->is_interior
&& !p
->is_internal
&&
438 ((len
< 2) || (data
[0] != AS_PATH_CONFED_SEQUENCE
)))
439 WITHDRAW("Malformed AS_PATH attribute - %s", "missing initial AS_CONFED_SEQUENCE");
443 /* Prepare 32-bit AS_PATH (from 16-bit one) in a temporary buffer */
445 data
= alloca(2*len
);
446 len
= as_path_16to32(data
, src
, len
);
449 bgp_set_attr_data(to
, s
->pool
, BA_AS_PATH
, flags
, data
, len
);
454 bgp_encode_next_hop(struct bgp_write_state
*s
, eattr
*a
, byte
*buf
, uint size
)
457 * The NEXT_HOP attribute is used only in traditional (IPv4) BGP. In MP-BGP,
458 * the next hop is encoded as a part of the MP_REACH_NLRI attribute, so we
459 * store it and encode it later by AFI-specific hooks.
464 // ASSERT(a->u.ptr->length == sizeof(ip_addr));
466 /* FIXME: skip IPv6 next hops for IPv4 routes during MRT dump */
467 ip_addr
*addr
= (void *) a
->u
.ptr
->data
;
468 if ((a
->u
.ptr
->length
!= sizeof(ip_addr
)) || !ipa_is_ip4(*addr
))
474 bgp_put_attr_hdr3(buf
, BA_NEXT_HOP
, a
->flags
, 4);
475 put_ip4(buf
+3, ipa_to_ip4(*addr
));
487 bgp_decode_next_hop(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags UNUSED
, byte
*data
, uint len
, ea_list
**to UNUSED
)
490 WITHDRAW(BAD_LENGTH
, "NEXT_HOP", len
);
492 /* Semantic checks are done later */
493 s
->ip_next_hop_len
= len
;
494 s
->ip_next_hop_data
= data
;
497 /* TODO: This function should use AF-specific hook */
499 bgp_format_next_hop(eattr
*a
, byte
*buf
, uint size UNUSED
)
501 ip_addr
*nh
= (void *) a
->u
.ptr
->data
;
502 uint len
= a
->u
.ptr
->length
;
504 ASSERT((len
== 16) || (len
== 32));
506 /* in IPv6, we may have two addresses in NEXT HOP */
507 if ((len
== 16) || ipa_zero(nh
[1]))
508 bsprintf(buf
, "%I", nh
[0]);
510 bsprintf(buf
, "%I %I", nh
[0], nh
[1]);
515 bgp_decode_med(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
518 WITHDRAW(BAD_LENGTH
, "MULTI_EXIT_DISC", len
);
520 u32 val
= get_u32(data
);
521 bgp_set_attr_u32(to
, s
->pool
, BA_MULTI_EXIT_DISC
, flags
, val
);
526 bgp_export_local_pref(struct bgp_export_state
*s
, eattr
*a
)
528 if (!s
->proto
->is_interior
&& !s
->proto
->cf
->allow_local_pref
)
533 bgp_decode_local_pref(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
535 if (!s
->proto
->is_interior
&& !s
->proto
->cf
->allow_local_pref
)
536 DISCARD(BAD_EBGP
, "LOCAL_PREF");
539 WITHDRAW(BAD_LENGTH
, "LOCAL_PREF", len
);
541 u32 val
= get_u32(data
);
542 bgp_set_attr_u32(to
, s
->pool
, BA_LOCAL_PREF
, flags
, val
);
547 bgp_decode_atomic_aggr(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data UNUSED
, uint len
, ea_list
**to
)
550 DISCARD(BAD_LENGTH
, "ATOMIC_AGGR", len
);
552 bgp_set_attr_data(to
, s
->pool
, BA_ATOMIC_AGGR
, flags
, NULL
, 0);
556 bgp_encode_aggregator(struct bgp_write_state
*s
, eattr
*a
, byte
*buf
, uint size
)
558 const byte
*data
= a
->u
.ptr
->data
;
559 uint len
= a
->u
.ptr
->length
;
563 /* Prepare 16-bit AGGREGATOR (from 32-bit one) in a temporary buffer */
564 byte
*dst
= alloca(6);
565 len
= aggregator_32to16(dst
, data
);
568 return bgp_put_attr(buf
, size
, BA_AGGREGATOR
, a
->flags
, data
, len
);
572 bgp_decode_aggregator(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
574 if (len
!= (s
->as4_session
? 8 : 6))
575 DISCARD(BAD_LENGTH
, "AGGREGATOR", len
);
579 /* Prepare 32-bit AGGREGATOR (from 16-bit one) in a temporary buffer */
582 len
= aggregator_16to32(data
, src
);
585 bgp_set_attr_data(to
, s
->pool
, BA_AGGREGATOR
, flags
, data
, len
);
589 bgp_format_aggregator(eattr
*a
, byte
*buf
, uint size UNUSED
)
591 const byte
*data
= a
->u
.ptr
->data
;
593 bsprintf(buf
, "%I4 AS%u", get_ip4(data
+4), get_u32(data
+0));
598 bgp_export_community(struct bgp_export_state
*s
, eattr
*a
)
600 if (a
->u
.ptr
->length
== 0)
603 a
->u
.ptr
= int_set_sort(s
->pool
, a
->u
.ptr
);
607 bgp_decode_community(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
609 if (!len
|| (len
% 4))
610 WITHDRAW(BAD_LENGTH
, "COMMUNITY", len
);
612 struct adata
*ad
= lp_alloc_adata(s
->pool
, len
);
613 get_u32s(data
, (u32
*) ad
->data
, len
/ 4);
614 bgp_set_attr_ptr(to
, s
->pool
, BA_COMMUNITY
, flags
, ad
);
619 bgp_export_originator_id(struct bgp_export_state
*s
, eattr
*a
)
621 if (!s
->proto
->is_internal
)
626 bgp_decode_originator_id(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
628 if (!s
->proto
->is_internal
)
629 DISCARD(BAD_EBGP
, "ORIGINATOR_ID");
632 WITHDRAW(BAD_LENGTH
, "ORIGINATOR_ID", len
);
634 u32 val
= get_u32(data
);
635 bgp_set_attr_u32(to
, s
->pool
, BA_ORIGINATOR_ID
, flags
, val
);
640 bgp_export_cluster_list(struct bgp_export_state
*s UNUSED
, eattr
*a
)
642 if (!s
->proto
->is_internal
)
645 if (a
->u
.ptr
->length
== 0)
650 bgp_decode_cluster_list(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
652 if (!s
->proto
->is_internal
)
653 DISCARD(BAD_EBGP
, "CLUSTER_LIST");
655 if (!len
|| (len
% 4))
656 WITHDRAW(BAD_LENGTH
, "CLUSTER_LIST", len
);
658 struct adata
*ad
= lp_alloc_adata(s
->pool
, len
);
659 get_u32s(data
, (u32
*) ad
->data
, len
/ 4);
660 bgp_set_attr_ptr(to
, s
->pool
, BA_CLUSTER_LIST
, flags
, ad
);
664 bgp_format_cluster_list(eattr
*a
, byte
*buf
, uint size
)
666 /* Truncates cluster lists larger than buflen, probably not a problem */
667 int_set_format(a
->u
.ptr
, 0, -1, buf
, size
);
674 return (get_u16(buf
) << 16) | buf
[2];
678 bgp_decode_mp_reach_nlri(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags UNUSED
, byte
*data
, uint len
, ea_list
**to UNUSED
)
681 * 2 B MP_REACH_NLRI data - Address Family Identifier
682 * 1 B MP_REACH_NLRI data - Subsequent Address Family Identifier
683 * 1 B MP_REACH_NLRI data - Length of Next Hop Network Address
684 * var MP_REACH_NLRI data - Network Address of Next Hop
685 * 1 B MP_REACH_NLRI data - Reserved (zero)
686 * var MP_REACH_NLRI data - Network Layer Reachability Information
689 if ((len
< 5) || (len
< (5 + (uint
) data
[3])))
690 bgp_parse_error(s
, 9);
692 s
->mp_reach_af
= get_af3(data
);
693 s
->mp_next_hop_len
= data
[3];
694 s
->mp_next_hop_data
= data
+ 4;
695 s
->mp_reach_len
= len
- 5 - s
->mp_next_hop_len
;
696 s
->mp_reach_nlri
= data
+ 5 + s
->mp_next_hop_len
;
701 bgp_decode_mp_unreach_nlri(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags UNUSED
, byte
*data
, uint len
, ea_list
**to UNUSED
)
704 * 2 B MP_UNREACH_NLRI data - Address Family Identifier
705 * 1 B MP_UNREACH_NLRI data - Subsequent Address Family Identifier
706 * var MP_UNREACH_NLRI data - Network Layer Reachability Information
710 bgp_parse_error(s
, 9);
712 s
->mp_unreach_af
= get_af3(data
);
713 s
->mp_unreach_len
= len
- 3;
714 s
->mp_unreach_nlri
= data
+ 3;
719 bgp_export_ext_community(struct bgp_export_state
*s
, eattr
*a
)
721 if (!s
->proto
->is_interior
)
723 struct adata
*ad
= ec_set_del_nontrans(s
->pool
, a
->u
.ptr
);
733 if (a
->u
.ptr
->length
== 0)
736 a
->u
.ptr
= ec_set_sort(s
->pool
, a
->u
.ptr
);
741 bgp_decode_ext_community(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
743 if (!len
|| (len
% 8))
744 WITHDRAW(BAD_LENGTH
, "EXT_COMMUNITY", len
);
746 struct adata
*ad
= lp_alloc_adata(s
->pool
, len
);
747 get_u32s(data
, (u32
*) ad
->data
, len
/ 4);
748 bgp_set_attr_ptr(to
, s
->pool
, BA_EXT_COMMUNITY
, flags
, ad
);
753 bgp_decode_as4_aggregator(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
756 DISCARD(NEW_BGP
, "AS4_AGGREGATOR");
759 DISCARD(BAD_LENGTH
, "AS4_AGGREGATOR", len
);
761 bgp_set_attr_data(to
, s
->pool
, BA_AS4_AGGREGATOR
, flags
, data
, len
);
765 bgp_decode_as4_path(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
767 struct bgp_proto
*p
= s
->proto
;
768 int sets
= p
->cf
->allow_as_sets
;
773 DISCARD(NEW_BGP
, "AS4_PATH");
776 DISCARD(BAD_LENGTH
, "AS4_PATH", len
);
778 if (!as_path_valid(data
, len
, 4, sets
, 1, err
, sizeof(err
)))
779 DISCARD("Malformed AS4_PATH attribute - %s", err
);
781 struct adata
*a
= lp_alloc_adata(s
->pool
, len
);
782 memcpy(a
->data
, data
, len
);
784 /* AS_CONFED* segments are invalid in AS4_PATH; RFC 6793 6 */
785 if (as_path_contains_confed(a
))
787 REPORT("Discarding AS_CONFED* segment from AS4_PATH attribute");
788 a
= as_path_strip_confed(s
->pool
, a
);
791 bgp_set_attr_ptr(to
, s
->pool
, BA_AS4_PATH
, flags
, a
);
796 bgp_export_aigp(struct bgp_export_state
*s
, eattr
*a
)
798 if (!s
->channel
->cf
->aigp
)
803 bgp_decode_aigp(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
807 /* Acceptability test postponed to bgp_finish_attrs() */
809 if ((flags
^ bgp_attr_table
[BA_AIGP
].flags
) & (BAF_OPTIONAL
| BAF_TRANSITIVE
))
810 DISCARD("Malformed AIGP attribute - conflicting flags (%02x)", flags
);
812 if (!bgp_aigp_valid(data
, len
, err
, sizeof(err
)))
813 DISCARD("Malformed AIGP attribute - %s", err
);
815 bgp_set_attr_data(to
, s
->pool
, BA_AIGP
, flags
, data
, len
);
819 bgp_format_aigp(eattr
*a
, byte
*buf
, uint size UNUSED
)
821 const byte
*b
= bgp_aigp_get_tlv(a
->u
.ptr
, BGP_AIGP_METRIC
);
826 bsprintf(buf
, "%lu", get_u64(b
+ 3));
831 bgp_export_large_community(struct bgp_export_state
*s
, eattr
*a
)
833 if (a
->u
.ptr
->length
== 0)
836 a
->u
.ptr
= lc_set_sort(s
->pool
, a
->u
.ptr
);
840 bgp_decode_large_community(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
842 if (!len
|| (len
% 12))
843 WITHDRAW(BAD_LENGTH
, "LARGE_COMMUNITY", len
);
845 struct adata
*ad
= lp_alloc_adata(s
->pool
, len
);
846 get_u32s(data
, (u32
*) ad
->data
, len
/ 4);
847 bgp_set_attr_ptr(to
, s
->pool
, BA_LARGE_COMMUNITY
, flags
, ad
);
851 bgp_export_mpls_label_stack(struct bgp_export_state
*s
, eattr
*a
)
853 net_addr
*n
= s
->route
->net
->n
.addr
;
854 u32
*labels
= (u32
*) a
->u
.ptr
->data
;
855 uint lnum
= a
->u
.ptr
->length
/ 4;
857 /* Perhaps we should just ignore it? */
859 WITHDRAW("Unexpected MPLS stack");
861 /* Empty MPLS stack is not allowed */
863 WITHDRAW("Malformed MPLS stack - empty");
865 /* This is ugly, but we must ensure that labels fit into NLRI field */
866 if ((24*lnum
+ (net_is_vpn(n
) ? 64 : 0) + net_pxlen(n
)) > 255)
867 WITHDRAW("Malformed MPLS stack - too many labels (%u)", lnum
);
869 for (uint i
= 0; i
< lnum
; i
++)
871 if (labels
[i
] > 0xfffff)
872 WITHDRAW("Malformed MPLS stack - invalid label (%u)", labels
[i
]);
874 /* TODO: Check for special-purpose label values? */
879 bgp_encode_mpls_label_stack(struct bgp_write_state
*s
, eattr
*a
, byte
*buf UNUSED
, uint size UNUSED
)
882 * MPLS labels are encoded as a part of the NLRI in MP_REACH_NLRI attribute,
883 * so we store MPLS_LABEL_STACK and encode it later by AFI-specific hooks.
886 s
->mpls_labels
= a
->u
.ptr
;
891 bgp_decode_mpls_label_stack(struct bgp_parse_state
*s
, uint code UNUSED
, uint flags UNUSED
, byte
*data UNUSED
, uint len UNUSED
, ea_list
**to UNUSED
)
893 DISCARD("Discarding received attribute #0");
897 bgp_format_mpls_label_stack(eattr
*a
, byte
*buf
, uint size
)
899 u32
*labels
= (u32
*) a
->u
.ptr
->data
;
900 uint lnum
= a
->u
.ptr
->length
/ 4;
903 for (uint i
= 0; i
< lnum
; i
++)
907 bsprintf(pos
, "...");
911 uint l
= bsprintf(pos
, "%d/", labels
[i
]);
912 ADVANCE(pos
, size
, l
);
915 /* Clear last slash or terminate empty string */
916 pos
[lnum
? -1 : 0] = 0;
920 bgp_decode_unknown(struct bgp_parse_state
*s
, uint code
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
922 /* Cannot use bgp_set_attr_data() as it works on known attributes only */
923 ea_set_attr_data(to
, s
->pool
, EA_CODE(PROTOCOL_BGP
, code
), flags
, EAF_TYPE_OPAQUE
, data
, len
);
931 static const struct bgp_attr_desc bgp_attr_table
[] = {
934 .type
= EAF_TYPE_INT
,
935 .flags
= BAF_TRANSITIVE
,
936 .export
= bgp_export_origin
,
937 .encode
= bgp_encode_u8
,
938 .decode
= bgp_decode_origin
,
939 .format
= bgp_format_origin
,
943 .type
= EAF_TYPE_AS_PATH
,
944 .flags
= BAF_TRANSITIVE
,
945 .encode
= bgp_encode_as_path
,
946 .decode
= bgp_decode_as_path
,
950 .type
= EAF_TYPE_IP_ADDRESS
,
951 .flags
= BAF_TRANSITIVE
,
952 .encode
= bgp_encode_next_hop
,
953 .decode
= bgp_decode_next_hop
,
954 .format
= bgp_format_next_hop
,
956 [BA_MULTI_EXIT_DISC
] = {
958 .type
= EAF_TYPE_INT
,
959 .flags
= BAF_OPTIONAL
,
960 .encode
= bgp_encode_u32
,
961 .decode
= bgp_decode_med
,
964 .name
= "local_pref",
965 .type
= EAF_TYPE_INT
,
966 .flags
= BAF_TRANSITIVE
,
967 .export
= bgp_export_local_pref
,
968 .encode
= bgp_encode_u32
,
969 .decode
= bgp_decode_local_pref
,
972 .name
= "atomic_aggr",
973 .type
= EAF_TYPE_OPAQUE
,
974 .flags
= BAF_TRANSITIVE
,
975 .encode
= bgp_encode_raw
,
976 .decode
= bgp_decode_atomic_aggr
,
979 .name
= "aggregator",
980 .type
= EAF_TYPE_OPAQUE
,
981 .flags
= BAF_OPTIONAL
| BAF_TRANSITIVE
,
982 .encode
= bgp_encode_aggregator
,
983 .decode
= bgp_decode_aggregator
,
984 .format
= bgp_format_aggregator
,
988 .type
= EAF_TYPE_INT_SET
,
989 .flags
= BAF_OPTIONAL
| BAF_TRANSITIVE
,
990 .export
= bgp_export_community
,
991 .encode
= bgp_encode_u32s
,
992 .decode
= bgp_decode_community
,
994 [BA_ORIGINATOR_ID
] = {
995 .name
= "originator_id",
996 .type
= EAF_TYPE_ROUTER_ID
,
997 .flags
= BAF_OPTIONAL
,
998 .export
= bgp_export_originator_id
,
999 .encode
= bgp_encode_u32
,
1000 .decode
= bgp_decode_originator_id
,
1002 [BA_CLUSTER_LIST
] = {
1003 .name
= "cluster_list",
1004 .type
= EAF_TYPE_INT_SET
,
1005 .flags
= BAF_OPTIONAL
,
1006 .export
= bgp_export_cluster_list
,
1007 .encode
= bgp_encode_u32s
,
1008 .decode
= bgp_decode_cluster_list
,
1009 .format
= bgp_format_cluster_list
,
1011 [BA_MP_REACH_NLRI
] = {
1012 .name
= "mp_reach_nlri",
1013 .type
= EAF_TYPE_OPAQUE
,
1014 .flags
= BAF_OPTIONAL
,
1015 .decode
= bgp_decode_mp_reach_nlri
,
1017 [BA_MP_UNREACH_NLRI
] = {
1018 .name
= "mp_unreach_nlri",
1019 .type
= EAF_TYPE_OPAQUE
,
1020 .flags
= BAF_OPTIONAL
,
1021 .decode
= bgp_decode_mp_unreach_nlri
,
1023 [BA_EXT_COMMUNITY
] = {
1024 .name
= "ext_community",
1025 .type
= EAF_TYPE_EC_SET
,
1026 .flags
= BAF_OPTIONAL
| BAF_TRANSITIVE
,
1027 .export
= bgp_export_ext_community
,
1028 .encode
= bgp_encode_u32s
,
1029 .decode
= bgp_decode_ext_community
,
1033 .type
= EAF_TYPE_AS_PATH
,
1034 .flags
= BAF_OPTIONAL
| BAF_TRANSITIVE
,
1035 .encode
= bgp_encode_raw
,
1036 .decode
= bgp_decode_as4_path
,
1038 [BA_AS4_AGGREGATOR
] = {
1039 .name
= "as4_aggregator",
1040 .type
= EAF_TYPE_OPAQUE
,
1041 .flags
= BAF_OPTIONAL
| BAF_TRANSITIVE
,
1042 .encode
= bgp_encode_raw
,
1043 .decode
= bgp_decode_as4_aggregator
,
1044 .format
= bgp_format_aggregator
,
1048 .type
= EAF_TYPE_OPAQUE
,
1049 .flags
= BAF_OPTIONAL
| BAF_DECODE_FLAGS
,
1050 .export
= bgp_export_aigp
,
1051 .encode
= bgp_encode_raw
,
1052 .decode
= bgp_decode_aigp
,
1053 .format
= bgp_format_aigp
,
1055 [BA_LARGE_COMMUNITY
] = {
1056 .name
= "large_community",
1057 .type
= EAF_TYPE_LC_SET
,
1058 .flags
= BAF_OPTIONAL
| BAF_TRANSITIVE
,
1059 .export
= bgp_export_large_community
,
1060 .encode
= bgp_encode_u32s
,
1061 .decode
= bgp_decode_large_community
,
1063 [BA_MPLS_LABEL_STACK
] = {
1064 .name
= "mpls_label_stack",
1065 .type
= EAF_TYPE_INT_SET
,
1066 .export
= bgp_export_mpls_label_stack
,
1067 .encode
= bgp_encode_mpls_label_stack
,
1068 .decode
= bgp_decode_mpls_label_stack
,
1069 .format
= bgp_format_mpls_label_stack
,
1074 bgp_attr_known(uint code
)
1076 return (code
< ARRAY_SIZE(bgp_attr_table
)) && bgp_attr_table
[code
].name
;
1085 bgp_export_attr(struct bgp_export_state
*s
, eattr
*a
, ea_list
*to
)
1087 if (EA_PROTO(a
->id
) != PROTOCOL_BGP
)
1090 uint code
= EA_ID(a
->id
);
1092 if (bgp_attr_known(code
))
1094 const struct bgp_attr_desc
*desc
= &bgp_attr_table
[code
];
1096 /* The flags might have been zero if the attr was added by filters */
1097 a
->flags
= (a
->flags
& BAF_PARTIAL
) | desc
->flags
;
1099 /* Set partial bit if new opt-trans attribute is attached to non-local route */
1100 if ((s
->src
!= NULL
) && (a
->type
& EAF_ORIGINATED
) &&
1101 (a
->flags
& BAF_OPTIONAL
) && (a
->flags
& BAF_TRANSITIVE
))
1102 a
->flags
|= BAF_PARTIAL
;
1104 /* Call specific hook */
1105 CALL(desc
->export
, s
, a
);
1107 /* Attribute might become undefined in hook */
1108 if ((a
->type
& EAF_TYPE_MASK
) == EAF_TYPE_UNDEF
)
1113 /* Don't re-export unknown non-transitive attributes */
1114 if (!(a
->flags
& BAF_TRANSITIVE
))
1117 a
->flags
|= BAF_PARTIAL
;
1120 /* Append updated attribute */
1121 to
->attrs
[to
->count
++] = *a
;
1125 * bgp_export_attrs - export BGP attributes
1126 * @s: BGP export state
1127 * @attrs: a list of extended attributes
1129 * The bgp_export_attrs() function takes a list of attributes and merges it to
1130 * one newly allocated and sorted segment. Attributes are validated and
1131 * normalized by type-specific export hooks and attribute flags are updated.
1132 * Some attributes may be eliminated (e.g. unknown non-tranitive attributes, or
1133 * empty community sets).
1135 * Result: one sorted attribute list segment, or NULL if attributes are unsuitable.
1137 static inline ea_list
*
1138 bgp_export_attrs(struct bgp_export_state
*s
, ea_list
*attrs
)
1140 /* Merge the attribute list */
1141 ea_list
*new = lp_alloc(s
->pool
, ea_scan(attrs
));
1142 ea_merge(attrs
, new);
1149 /* Export each attribute */
1150 for (i
= 0; i
< count
; i
++)
1151 bgp_export_attr(s
, &new->attrs
[i
], new);
1153 if (s
->err_withdraw
)
1161 * Attribute encoding
1165 bgp_encode_attr(struct bgp_write_state
*s
, eattr
*a
, byte
*buf
, uint size
)
1167 ASSERT(EA_PROTO(a
->id
) == PROTOCOL_BGP
);
1169 uint code
= EA_ID(a
->id
);
1171 if (bgp_attr_known(code
))
1172 return bgp_attr_table
[code
].encode(s
, a
, buf
, size
);
1174 return bgp_encode_raw(s
, a
, buf
, size
);
1178 * bgp_encode_attrs - encode BGP attributes
1179 * @s: BGP write state
1180 * @attrs: a list of extended attributes
1184 * The bgp_encode_attrs() function takes a list of extended attributes
1185 * and converts it to its BGP representation (a part of an Update message).
1186 * BGP write state may be fake when called from MRT protocol.
1188 * Result: Length of the attribute block generated or -1 if not enough space.
1191 bgp_encode_attrs(struct bgp_write_state
*s
, ea_list
*attrs
, byte
*buf
, byte
*end
)
1196 for (i
= 0; i
< attrs
->count
; i
++)
1198 len
= bgp_encode_attr(s
, &attrs
->attrs
[i
], pos
, end
- pos
);
1211 * Attribute decoding
1214 static void bgp_process_as4_attrs(ea_list
**attrs
, struct linpool
*pool
);
1217 bgp_as_path_loopy(struct bgp_proto
*p
, ea_list
*attrs
, u32 asn
)
1219 eattr
*e
= bgp_find_attr(attrs
, BA_AS_PATH
);
1220 int num
= p
->cf
->allow_local_as
+ 1;
1221 return (e
&& (num
> 0) && as_path_contains(e
->u
.ptr
, asn
, num
));
1225 bgp_originator_id_loopy(struct bgp_proto
*p
, ea_list
*attrs
)
1227 eattr
*e
= bgp_find_attr(attrs
, BA_ORIGINATOR_ID
);
1228 return (e
&& (e
->u
.data
== p
->local_id
));
1232 bgp_cluster_list_loopy(struct bgp_proto
*p
, ea_list
*attrs
)
1234 eattr
*e
= bgp_find_attr(attrs
, BA_CLUSTER_LIST
);
1235 return (e
&& int_set_contains(e
->u
.ptr
, p
->rr_cluster_id
));
1239 bgp_decode_attr(struct bgp_parse_state
*s
, uint code
, uint flags
, byte
*data
, uint len
, ea_list
**to
)
1241 /* Handle duplicate attributes; RFC 7606 3 (g) */
1242 if (BIT32_TEST(s
->attrs_seen
, code
))
1244 if ((code
== BA_MP_REACH_NLRI
) || (code
== BA_MP_UNREACH_NLRI
))
1245 bgp_parse_error(s
, 1);
1247 DISCARD("Discarding duplicate attribute (code %u)", code
);
1249 BIT32_SET(s
->attrs_seen
, code
);
1251 if (bgp_attr_known(code
))
1253 const struct bgp_attr_desc
*desc
= &bgp_attr_table
[code
];
1255 /* Handle conflicting flags; RFC 7606 3 (c) */
1256 if (((flags
^ desc
->flags
) & (BAF_OPTIONAL
| BAF_TRANSITIVE
)) &&
1257 !(desc
->flags
& BAF_DECODE_FLAGS
))
1258 WITHDRAW("Malformed %s attribute - conflicting flags (%02x)", desc
->name
, flags
);
1260 desc
->decode(s
, code
, flags
, data
, len
, to
);
1262 else /* Unknown attribute */
1264 if (!(flags
& BAF_OPTIONAL
))
1265 WITHDRAW("Unknown attribute (code %u) - conflicting flags (%02x)", code
, flags
);
1267 bgp_decode_unknown(s
, code
, flags
, data
, len
, to
);
1272 * bgp_decode_attrs - check and decode BGP attributes
1273 * @s: BGP parse state
1274 * @data: start of attribute block
1275 * @len: length of attribute block
1277 * This function takes a BGP attribute block (a part of an Update message), checks
1278 * its consistency and converts it to a list of BIRD route attributes represented
1279 * by an (uncached) &rta.
1282 bgp_decode_attrs(struct bgp_parse_state
*s
, byte
*data
, uint len
)
1284 struct bgp_proto
*p
= s
->proto
;
1285 ea_list
*attrs
= NULL
;
1286 uint code
, flags
, alen
;
1289 /* Parse the attributes */
1294 /* Read attribute type */
1299 ADVANCE(pos
, len
, 2);
1301 /* Read attribute length */
1302 if (flags
& BAF_EXT_LEN
)
1306 alen
= get_u16(pos
);
1307 ADVANCE(pos
, len
, 2);
1314 ADVANCE(pos
, len
, 1);
1320 DBG("Attr %02x %02x %u\n", code
, flags
, alen
);
1322 bgp_decode_attr(s
, code
, flags
, pos
, alen
, &attrs
);
1323 ADVANCE(pos
, len
, alen
);
1326 if (s
->err_withdraw
)
1329 /* If there is no reachability NLRI, we are finished */
1330 if (!s
->ip_reach_len
&& !s
->mp_reach_len
)
1334 /* Handle missing mandatory attributes; RFC 7606 3 (d) */
1335 if (!BIT32_TEST(s
->attrs_seen
, BA_ORIGIN
))
1336 { REPORT(NO_MANDATORY
, "ORIGIN"); goto withdraw
; }
1338 if (!BIT32_TEST(s
->attrs_seen
, BA_AS_PATH
))
1339 { REPORT(NO_MANDATORY
, "AS_PATH"); goto withdraw
; }
1341 if (s
->ip_reach_len
&& !BIT32_TEST(s
->attrs_seen
, BA_NEXT_HOP
))
1342 { REPORT(NO_MANDATORY
, "NEXT_HOP"); goto withdraw
; }
1344 /* When receiving attributes from non-AS4-aware BGP speaker, we have to
1345 reconstruct AS_PATH and AGGREGATOR attributes; RFC 6793 4.2.3 */
1346 if (!p
->as4_session
)
1347 bgp_process_as4_attrs(&attrs
, s
->pool
);
1349 /* Reject routes with our ASN in AS_PATH attribute */
1350 if (bgp_as_path_loopy(p
, attrs
, p
->local_as
))
1353 /* Reject routes with our Confederation ID in AS_PATH attribute; RFC 5065 4.0 */
1354 if ((p
->public_as
!= p
->local_as
) && bgp_as_path_loopy(p
, attrs
, p
->public_as
))
1357 /* Reject routes with our Router ID in ORIGINATOR_ID attribute; RFC 4456 8 */
1358 if (p
->is_internal
&& bgp_originator_id_loopy(p
, attrs
))
1361 /* Reject routes with our Cluster ID in CLUSTER_LIST attribute; RFC 4456 8 */
1362 if (p
->rr_client
&& bgp_cluster_list_loopy(p
, attrs
))
1365 /* If there is no local preference, define one */
1366 if (!BIT32_TEST(s
->attrs_seen
, BA_LOCAL_PREF
))
1367 bgp_set_attr_u32(&attrs
, s
->pool
, BA_LOCAL_PREF
, 0, p
->cf
->default_local_pref
);
1373 /* RFC 7606 4 - handle attribute framing errors */
1374 REPORT("Malformed attribute list - framing error (%u/%u) at %d",
1375 alen
, len
, (int) (pos
- s
->attrs
));
1378 /* RFC 7606 5.2 - handle missing NLRI during errors */
1379 if (!s
->ip_reach_len
&& !s
->mp_reach_len
)
1380 bgp_parse_error(s
, 1);
1382 s
->err_withdraw
= 1;
1387 bgp_finish_attrs(struct bgp_parse_state
*s
, rta
*a
)
1389 /* AIGP test here instead of in bgp_decode_aigp() - we need to know channel */
1390 if (BIT32_TEST(s
->attrs_seen
, BA_AIGP
) && !s
->channel
->cf
->aigp
)
1392 REPORT("Discarding AIGP attribute received on non-AIGP session");
1393 bgp_unset_attr(&a
->eattrs
, s
->pool
, BA_AIGP
);
1399 * Route bucket hash table
1402 #define RBH_KEY(b) b->eattrs, b->hash
1403 #define RBH_NEXT(b) b->next
1404 #define RBH_EQ(a1,h1,a2,h2) h1 == h2 && ea_same(a1, a2)
1405 #define RBH_FN(a,h) h
1407 #define RBH_REHASH bgp_rbh_rehash
1408 #define RBH_PARAMS /8, *2, 2, 2, 8, 20
1411 HASH_DEFINE_REHASH_FN(RBH
, struct bgp_bucket
)
1414 bgp_init_bucket_table(struct bgp_channel
*c
)
1416 HASH_INIT(c
->bucket_hash
, c
->pool
, 8);
1418 init_list(&c
->bucket_queue
);
1419 c
->withdraw_bucket
= NULL
;
1423 bgp_free_bucket_table(struct bgp_channel
*c
)
1425 HASH_FREE(c
->bucket_hash
);
1427 struct bgp_bucket
*b
;
1428 WALK_LIST_FIRST(b
, c
->bucket_queue
)
1430 rem_node(&b
->send_node
);
1434 mb_free(c
->withdraw_bucket
);
1435 c
->withdraw_bucket
= NULL
;
1438 static struct bgp_bucket
*
1439 bgp_get_bucket(struct bgp_channel
*c
, ea_list
*new)
1441 /* Hash and lookup */
1442 u32 hash
= ea_hash(new);
1443 struct bgp_bucket
*b
= HASH_FIND(c
->bucket_hash
, RBH
, new, hash
);
1448 uint ea_size
= sizeof(ea_list
) + new->count
* sizeof(eattr
);
1449 uint ea_size_aligned
= BIRD_ALIGN(ea_size
, CPU_STRUCT_ALIGN
);
1450 uint size
= sizeof(struct bgp_bucket
) + ea_size_aligned
;
1454 /* Gather total size of non-inline attributes */
1455 for (i
= 0; i
< new->count
; i
++)
1457 eattr
*a
= &new->attrs
[i
];
1459 if (!(a
->type
& EAF_EMBEDDED
))
1460 size
+= BIRD_ALIGN(sizeof(struct adata
) + a
->u
.ptr
->length
, CPU_STRUCT_ALIGN
);
1463 /* Create the bucket */
1464 b
= mb_alloc(c
->pool
, size
);
1465 init_list(&b
->prefixes
);
1468 /* Copy list of extended attributes */
1469 memcpy(b
->eattrs
, new, ea_size
);
1470 dest
= ((byte
*) b
->eattrs
) + ea_size_aligned
;
1472 /* Copy values of non-inline attributes */
1473 for (i
= 0; i
< new->count
; i
++)
1475 eattr
*a
= &b
->eattrs
->attrs
[i
];
1477 if (!(a
->type
& EAF_EMBEDDED
))
1479 const struct adata
*oa
= a
->u
.ptr
;
1480 struct adata
*na
= (struct adata
*) dest
;
1481 memcpy(na
, oa
, sizeof(struct adata
) + oa
->length
);
1483 dest
+= BIRD_ALIGN(sizeof(struct adata
) + na
->length
, CPU_STRUCT_ALIGN
);
1487 /* Insert the bucket to send queue and bucket hash */
1488 add_tail(&c
->bucket_queue
, &b
->send_node
);
1489 HASH_INSERT2(c
->bucket_hash
, RBH
, c
->pool
, b
);
1494 static struct bgp_bucket
*
1495 bgp_get_withdraw_bucket(struct bgp_channel
*c
)
1497 if (!c
->withdraw_bucket
)
1499 c
->withdraw_bucket
= mb_allocz(c
->pool
, sizeof(struct bgp_bucket
));
1500 init_list(&c
->withdraw_bucket
->prefixes
);
1503 return c
->withdraw_bucket
;
1507 bgp_free_bucket(struct bgp_channel
*c
, struct bgp_bucket
*b
)
1509 rem_node(&b
->send_node
);
1510 HASH_REMOVE2(c
->bucket_hash
, RBH
, c
->pool
, b
);
1515 bgp_defer_bucket(struct bgp_channel
*c
, struct bgp_bucket
*b
)
1517 rem_node(&b
->send_node
);
1518 add_tail(&c
->bucket_queue
, &b
->send_node
);
1522 bgp_withdraw_bucket(struct bgp_channel
*c
, struct bgp_bucket
*b
)
1524 struct bgp_proto
*p
= (void *) c
->c
.proto
;
1525 struct bgp_bucket
*wb
= bgp_get_withdraw_bucket(c
);
1527 log(L_ERR
"%s: Attribute list too long", p
->p
.name
);
1528 while (!EMPTY_LIST(b
->prefixes
))
1530 struct bgp_prefix
*px
= HEAD(b
->prefixes
);
1532 log(L_ERR
"%s: - withdrawing %N", p
->p
.name
, &px
->net
);
1533 rem_node(&px
->buck_node
);
1534 add_tail(&wb
->prefixes
, &px
->buck_node
);
1543 #define PXH_KEY(px) px->net, px->path_id, px->hash
1544 #define PXH_NEXT(px) px->next
1545 #define PXH_EQ(n1,i1,h1,n2,i2,h2) h1 == h2 && i1 == i2 && net_equal(n1, n2)
1546 #define PXH_FN(n,i,h) h
1548 #define PXH_REHASH bgp_pxh_rehash
1549 #define PXH_PARAMS /8, *2, 2, 2, 8, 24
1552 HASH_DEFINE_REHASH_FN(PXH
, struct bgp_prefix
)
1555 bgp_init_prefix_table(struct bgp_channel
*c
)
1557 HASH_INIT(c
->prefix_hash
, c
->pool
, 8);
1559 uint alen
= net_addr_length
[c
->c
.net_type
];
1560 c
->prefix_slab
= alen
? sl_new(c
->pool
, sizeof(struct bgp_prefix
) + alen
) : NULL
;
1564 bgp_free_prefix_table(struct bgp_channel
*c
)
1566 HASH_FREE(c
->prefix_hash
);
1568 rfree(c
->prefix_slab
);
1569 c
->prefix_slab
= NULL
;
1572 static struct bgp_prefix
*
1573 bgp_get_prefix(struct bgp_channel
*c
, net_addr
*net
, u32 path_id
)
1575 u32 hash
= net_hash(net
) ^ u32_hash(path_id
);
1576 struct bgp_prefix
*px
= HASH_FIND(c
->prefix_hash
, PXH
, net
, path_id
, hash
);
1580 rem_node(&px
->buck_node
);
1585 px
= sl_alloc(c
->prefix_slab
);
1587 px
= mb_alloc(c
->pool
, sizeof(struct bgp_prefix
) + net
->length
);
1589 px
->buck_node
.next
= NULL
;
1590 px
->buck_node
.prev
= NULL
;
1592 px
->path_id
= path_id
;
1593 net_copy(px
->net
, net
);
1595 HASH_INSERT2(c
->prefix_hash
, PXH
, c
->pool
, px
);
1601 bgp_free_prefix(struct bgp_channel
*c
, struct bgp_prefix
*px
)
1603 rem_node(&px
->buck_node
);
1604 HASH_REMOVE2(c
->prefix_hash
, PXH
, c
->pool
, px
);
1607 sl_free(c
->prefix_slab
, px
);
1618 bgp_preexport(struct proto
*P
, rte
**new, struct linpool
*pool UNUSED
)
1621 struct proto
*SRC
= e
->attrs
->src
->proto
;
1622 struct bgp_proto
*p
= (struct bgp_proto
*) P
;
1623 struct bgp_proto
*src
= (SRC
->proto
== &proto_bgp
) ? (struct bgp_proto
*) SRC
: NULL
;
1625 /* Reject our routes */
1629 /* Accept non-BGP routes */
1633 /* IBGP route reflection, RFC 4456 */
1634 if (p
->is_internal
&& src
->is_internal
&& (p
->local_as
== src
->local_as
))
1636 /* Rejected unless configured as route reflector */
1637 if (!p
->rr_client
&& !src
->rr_client
)
1640 /* Generally, this should be handled when path is received, but we check it
1641 also here as rr_cluster_id may be undefined or different in src. */
1642 if (p
->rr_cluster_id
&& bgp_cluster_list_loopy(p
, e
->attrs
->eattrs
))
1646 /* Handle well-known communities, RFC 1997 */
1648 if (p
->cf
->interpret_communities
&&
1649 (c
= ea_find(e
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_COMMUNITY
))))
1651 const struct adata
*d
= c
->u
.ptr
;
1653 /* Do not export anywhere */
1654 if (int_set_contains(d
, BGP_COMM_NO_ADVERTISE
))
1657 /* Do not export outside of AS (or member-AS) */
1658 if (!p
->is_internal
&& int_set_contains(d
, BGP_COMM_NO_EXPORT_SUBCONFED
))
1661 /* Do not export outside of AS (or confederation) */
1662 if (!p
->is_interior
&& int_set_contains(d
, BGP_COMM_NO_EXPORT
))
1665 /* Do not export LLGR_STALE routes to LLGR-ignorant peers */
1666 if (!p
->conn
->remote_caps
->llgr_aware
&& int_set_contains(d
, BGP_COMM_LLGR_STALE
))
1674 bgp_update_attrs(struct bgp_proto
*p
, struct bgp_channel
*c
, rte
*e
, ea_list
*attrs0
, struct linpool
*pool
)
1676 struct proto
*SRC
= e
->attrs
->src
->proto
;
1677 struct bgp_proto
*src
= (SRC
->proto
== &proto_bgp
) ? (void *) SRC
: NULL
;
1678 struct bgp_export_state s
= { .proto
= p
, .channel
= c
, .pool
= pool
, .src
= src
, .route
= e
, .mpls
= c
->desc
->mpls
};
1679 ea_list
*attrs
= attrs0
;
1683 /* ORIGIN attribute - mandatory, attach if missing */
1684 if (! bgp_find_attr(attrs0
, BA_ORIGIN
))
1685 bgp_set_attr_u32(&attrs
, pool
, BA_ORIGIN
, 0, src
? ORIGIN_INCOMPLETE
: ORIGIN_IGP
);
1687 /* AS_PATH attribute - mandatory */
1688 a
= bgp_find_attr(attrs0
, BA_AS_PATH
);
1689 ad
= a
? a
->u
.ptr
: &null_adata
;
1691 /* AS_PATH attribute - strip AS_CONFED* segments outside confederation */
1692 if ((!p
->cf
->confederation
|| !p
->is_interior
) && as_path_contains_confed(ad
))
1693 ad
= as_path_strip_confed(pool
, ad
);
1695 /* AS_PATH attribute - keep or prepend ASN */
1696 if (p
->is_internal
|| p
->rs_client
)
1698 /* IBGP or route server -> just ensure there is one */
1700 bgp_set_attr_ptr(&attrs
, pool
, BA_AS_PATH
, 0, &null_adata
);
1702 else if (p
->is_interior
)
1704 /* Confederation -> prepend ASN as AS_CONFED_SEQUENCE */
1705 ad
= as_path_prepend2(pool
, ad
, AS_PATH_CONFED_SEQUENCE
, p
->public_as
);
1706 bgp_set_attr_ptr(&attrs
, pool
, BA_AS_PATH
, 0, ad
);
1708 else /* Regular EBGP (no RS, no confederation) */
1710 /* Regular EBGP -> prepend ASN as regular sequence */
1711 ad
= as_path_prepend2(pool
, ad
, AS_PATH_SEQUENCE
, p
->public_as
);
1712 bgp_set_attr_ptr(&attrs
, pool
, BA_AS_PATH
, 0, ad
);
1714 /* MULTI_EXIT_DESC attribute - accept only if set in export filter */
1715 a
= bgp_find_attr(attrs0
, BA_MULTI_EXIT_DISC
);
1716 if (a
&& !(a
->type
& EAF_FRESH
))
1717 bgp_unset_attr(&attrs
, pool
, BA_MULTI_EXIT_DISC
);
1720 /* NEXT_HOP attribute - delegated to AF-specific hook */
1721 a
= bgp_find_attr(attrs0
, BA_NEXT_HOP
);
1722 bgp_update_next_hop(&s
, a
, &attrs
);
1724 /* LOCAL_PREF attribute - required for IBGP, attach if missing */
1725 if (p
->is_interior
&& ! bgp_find_attr(attrs0
, BA_LOCAL_PREF
))
1726 bgp_set_attr_u32(&attrs
, pool
, BA_LOCAL_PREF
, 0, p
->cf
->default_local_pref
);
1728 /* AIGP attribute - accumulate local metric or originate new one */
1730 if (s
.local_next_hop
&&
1731 (bgp_total_aigp_metric_(e
, &metric
, &ad
) ||
1732 (c
->cf
->aigp_originate
&& bgp_init_aigp_metric(e
, &metric
, &ad
))))
1734 ad
= bgp_aigp_set_metric(pool
, ad
, metric
);
1735 bgp_set_attr_ptr(&attrs
, pool
, BA_AIGP
, 0, ad
);
1738 /* IBGP route reflection, RFC 4456 */
1739 if (src
&& src
->is_internal
&& p
->is_internal
&& (src
->local_as
== p
->local_as
))
1741 /* ORIGINATOR_ID attribute - attach if not already set */
1742 if (! bgp_find_attr(attrs0
, BA_ORIGINATOR_ID
))
1743 bgp_set_attr_u32(&attrs
, pool
, BA_ORIGINATOR_ID
, 0, src
->remote_id
);
1745 /* CLUSTER_LIST attribute - prepend cluster ID */
1746 a
= bgp_find_attr(attrs0
, BA_CLUSTER_LIST
);
1747 ad
= a
? a
->u
.ptr
: NULL
;
1749 /* Prepend src cluster ID */
1750 if (src
->rr_cluster_id
)
1751 ad
= int_set_prepend(pool
, ad
, src
->rr_cluster_id
);
1753 /* Prepend dst cluster ID if src and dst clusters are different */
1754 if (p
->rr_cluster_id
&& (src
->rr_cluster_id
!= p
->rr_cluster_id
))
1755 ad
= int_set_prepend(pool
, ad
, p
->rr_cluster_id
);
1757 /* Should be at least one prepended cluster ID */
1758 bgp_set_attr_ptr(&attrs
, pool
, BA_CLUSTER_LIST
, 0, ad
);
1761 /* AS4_* transition attributes, RFC 6793 4.2.2 */
1762 if (! p
->as4_session
)
1764 a
= bgp_find_attr(attrs
, BA_AS_PATH
);
1765 if (a
&& as_path_contains_as4(a
->u
.ptr
))
1767 bgp_set_attr_ptr(&attrs
, pool
, BA_AS_PATH
, 0, as_path_to_old(pool
, a
->u
.ptr
));
1768 bgp_set_attr_ptr(&attrs
, pool
, BA_AS4_PATH
, 0, as_path_strip_confed(pool
, a
->u
.ptr
));
1771 a
= bgp_find_attr(attrs
, BA_AGGREGATOR
);
1772 if (a
&& aggregator_contains_as4(a
->u
.ptr
))
1774 bgp_set_attr_ptr(&attrs
, pool
, BA_AGGREGATOR
, 0, aggregator_to_old(pool
, a
->u
.ptr
));
1775 bgp_set_attr_ptr(&attrs
, pool
, BA_AS4_AGGREGATOR
, 0, a
->u
.ptr
);
1780 * Presence of mandatory attributes ORIGIN and AS_PATH is ensured by above
1781 * conditions. Presence and validity of quasi-mandatory NEXT_HOP attribute
1782 * should be checked in AF-specific hooks.
1785 /* Apply per-attribute export hooks for validatation and normalization */
1786 return bgp_export_attrs(&s
, attrs
);
1790 bgp_rt_notify(struct proto
*P
, struct channel
*C
, net
*n
, rte
*new, rte
*old
)
1792 struct bgp_proto
*p
= (void *) P
;
1793 struct bgp_channel
*c
= (void *) C
;
1794 struct bgp_bucket
*buck
;
1795 struct bgp_prefix
*px
;
1800 struct ea_list
*attrs
= bgp_update_attrs(p
, c
, new, new->attrs
->eattrs
, bgp_linpool2
);
1802 /* If attributes are invalid, we fail back to withdraw */
1803 buck
= attrs
? bgp_get_bucket(c
, attrs
) : bgp_get_withdraw_bucket(c
);
1804 path
= new->attrs
->src
->global_id
;
1806 lp_flush(bgp_linpool2
);
1810 buck
= bgp_get_withdraw_bucket(c
);
1811 path
= old
->attrs
->src
->global_id
;
1814 px
= bgp_get_prefix(c
, n
->n
.addr
, c
->add_path_tx
? path
: 0);
1815 add_tail(&buck
->prefixes
, &px
->buck_node
);
1817 bgp_schedule_packet(p
->conn
, c
, PKT_UPDATE
);
1822 bgp_get_neighbor(rte
*r
)
1824 eattr
*e
= ea_find(r
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_AS_PATH
));
1827 if (e
&& as_path_get_first_regular(e
->u
.ptr
, &as
))
1830 /* If AS_PATH is not defined, we treat rte as locally originated */
1831 struct bgp_proto
*p
= (void *) r
->attrs
->src
->proto
;
1832 return p
->cf
->confederation
?: p
->local_as
;
1838 if (r
->u
.bgp
.stale
< 0)
1840 /* If staleness is unknown, compute and cache it */
1841 eattr
*a
= ea_find(r
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_COMMUNITY
));
1842 r
->u
.bgp
.stale
= a
&& int_set_contains(a
->u
.ptr
, BGP_COMM_LLGR_STALE
);
1845 return r
->u
.bgp
.stale
;
1849 bgp_rte_better(rte
*new, rte
*old
)
1851 struct bgp_proto
*new_bgp
= (struct bgp_proto
*) new->attrs
->src
->proto
;
1852 struct bgp_proto
*old_bgp
= (struct bgp_proto
*) old
->attrs
->src
->proto
;
1856 /* Skip suppressed routes (see bgp_rte_recalculate()) */
1857 n
= new->u
.bgp
.suppressed
;
1858 o
= old
->u
.bgp
.suppressed
;
1864 /* RFC 4271 9.1.2.1. Route resolvability test */
1865 n
= rte_resolvable(new);
1866 o
= rte_resolvable(old
);
1872 /* LLGR draft - depreference stale routes */
1880 /* Start with local preferences */
1881 x
= ea_find(new->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_LOCAL_PREF
));
1882 y
= ea_find(old
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_LOCAL_PREF
));
1883 n
= x
? x
->u
.data
: new_bgp
->cf
->default_local_pref
;
1884 o
= y
? y
->u
.data
: old_bgp
->cf
->default_local_pref
;
1890 /* RFC 7311 4.1 - Apply AIGP metric */
1891 u64 n2
= bgp_total_aigp_metric(new);
1892 u64 o2
= bgp_total_aigp_metric(old
);
1898 /* RFC 4271 9.1.2.2. a) Use AS path lengths */
1899 if (new_bgp
->cf
->compare_path_lengths
|| old_bgp
->cf
->compare_path_lengths
)
1901 x
= ea_find(new->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_AS_PATH
));
1902 y
= ea_find(old
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_AS_PATH
));
1903 n
= x
? as_path_getlen(x
->u
.ptr
) : AS_PATH_MAXLEN
;
1904 o
= y
? as_path_getlen(y
->u
.ptr
) : AS_PATH_MAXLEN
;
1911 /* RFC 4271 9.1.2.2. b) Use origins */
1912 x
= ea_find(new->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_ORIGIN
));
1913 y
= ea_find(old
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_ORIGIN
));
1914 n
= x
? x
->u
.data
: ORIGIN_INCOMPLETE
;
1915 o
= y
? y
->u
.data
: ORIGIN_INCOMPLETE
;
1921 /* RFC 4271 9.1.2.2. c) Compare MED's */
1922 /* Proper RFC 4271 path selection cannot be interpreted as finding
1923 * the best path in some ordering. It is implemented partially in
1924 * bgp_rte_recalculate() when deterministic_med option is
1925 * active. Without that option, the behavior is just an
1926 * approximation, which in specific situations may lead to
1927 * persistent routing loops, because it is nondeterministic - it
1928 * depends on the order in which routes appeared. But it is also the
1929 * same behavior as used by default in Cisco routers, so it is
1930 * probably not a big issue.
1932 if (new_bgp
->cf
->med_metric
|| old_bgp
->cf
->med_metric
||
1933 (bgp_get_neighbor(new) == bgp_get_neighbor(old
)))
1935 x
= ea_find(new->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_MULTI_EXIT_DISC
));
1936 y
= ea_find(old
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_MULTI_EXIT_DISC
));
1937 n
= x
? x
->u
.data
: new_bgp
->cf
->default_med
;
1938 o
= y
? y
->u
.data
: old_bgp
->cf
->default_med
;
1945 /* RFC 4271 9.1.2.2. d) Prefer external peers */
1946 if (new_bgp
->is_interior
> old_bgp
->is_interior
)
1948 if (new_bgp
->is_interior
< old_bgp
->is_interior
)
1951 /* RFC 4271 9.1.2.2. e) Compare IGP metrics */
1952 n
= new_bgp
->cf
->igp_metric
? new->attrs
->igp_metric
: 0;
1953 o
= old_bgp
->cf
->igp_metric
? old
->attrs
->igp_metric
: 0;
1959 /* RFC 4271 9.1.2.2. f) Compare BGP identifiers */
1960 /* RFC 4456 9. a) Use ORIGINATOR_ID instead of local neighbor ID */
1961 x
= ea_find(new->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_ORIGINATOR_ID
));
1962 y
= ea_find(old
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_ORIGINATOR_ID
));
1963 n
= x
? x
->u
.data
: new_bgp
->remote_id
;
1964 o
= y
? y
->u
.data
: old_bgp
->remote_id
;
1966 /* RFC 5004 - prefer older routes */
1967 /* (if both are external and from different peer) */
1968 if ((new_bgp
->cf
->prefer_older
|| old_bgp
->cf
->prefer_older
) &&
1969 !new_bgp
->is_internal
&& n
!= o
)
1972 /* rest of RFC 4271 9.1.2.2. f) */
1978 /* RFC 4456 9. b) Compare cluster list lengths */
1979 x
= ea_find(new->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_CLUSTER_LIST
));
1980 y
= ea_find(old
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_CLUSTER_LIST
));
1981 n
= x
? int_set_get_size(x
->u
.ptr
) : 0;
1982 o
= y
? int_set_get_size(y
->u
.ptr
) : 0;
1988 /* RFC 4271 9.1.2.2. g) Compare peer IP adresses */
1989 return ipa_compare(new_bgp
->remote_ip
, old_bgp
->remote_ip
) < 0;
1994 bgp_rte_mergable(rte
*pri
, rte
*sec
)
1996 struct bgp_proto
*pri_bgp
= (struct bgp_proto
*) pri
->attrs
->src
->proto
;
1997 struct bgp_proto
*sec_bgp
= (struct bgp_proto
*) sec
->attrs
->src
->proto
;
2001 /* Skip suppressed routes (see bgp_rte_recalculate()) */
2002 if (pri
->u
.bgp
.suppressed
!= sec
->u
.bgp
.suppressed
)
2005 /* RFC 4271 9.1.2.1. Route resolvability test */
2006 if (rte_resolvable(pri
) != rte_resolvable(sec
))
2009 /* LLGR draft - depreference stale routes */
2010 if (rte_stale(pri
) != rte_stale(sec
))
2013 /* Start with local preferences */
2014 x
= ea_find(pri
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_LOCAL_PREF
));
2015 y
= ea_find(sec
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_LOCAL_PREF
));
2016 p
= x
? x
->u
.data
: pri_bgp
->cf
->default_local_pref
;
2017 s
= y
? y
->u
.data
: sec_bgp
->cf
->default_local_pref
;
2021 /* RFC 4271 9.1.2.2. a) Use AS path lengths */
2022 if (pri_bgp
->cf
->compare_path_lengths
|| sec_bgp
->cf
->compare_path_lengths
)
2024 x
= ea_find(pri
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_AS_PATH
));
2025 y
= ea_find(sec
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_AS_PATH
));
2026 p
= x
? as_path_getlen(x
->u
.ptr
) : AS_PATH_MAXLEN
;
2027 s
= y
? as_path_getlen(y
->u
.ptr
) : AS_PATH_MAXLEN
;
2032 // if (DELTA(p, s) > pri_bgp->cf->relax_multipath)
2036 /* RFC 4271 9.1.2.2. b) Use origins */
2037 x
= ea_find(pri
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_ORIGIN
));
2038 y
= ea_find(sec
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_ORIGIN
));
2039 p
= x
? x
->u
.data
: ORIGIN_INCOMPLETE
;
2040 s
= y
? y
->u
.data
: ORIGIN_INCOMPLETE
;
2044 /* RFC 4271 9.1.2.2. c) Compare MED's */
2045 if (pri_bgp
->cf
->med_metric
|| sec_bgp
->cf
->med_metric
||
2046 (bgp_get_neighbor(pri
) == bgp_get_neighbor(sec
)))
2048 x
= ea_find(pri
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_MULTI_EXIT_DISC
));
2049 y
= ea_find(sec
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_MULTI_EXIT_DISC
));
2050 p
= x
? x
->u
.data
: pri_bgp
->cf
->default_med
;
2051 s
= y
? y
->u
.data
: sec_bgp
->cf
->default_med
;
2056 /* RFC 4271 9.1.2.2. d) Prefer external peers */
2057 if (pri_bgp
->is_interior
!= sec_bgp
->is_interior
)
2060 /* RFC 4271 9.1.2.2. e) Compare IGP metrics */
2061 p
= pri_bgp
->cf
->igp_metric
? pri
->attrs
->igp_metric
: 0;
2062 s
= sec_bgp
->cf
->igp_metric
? sec
->attrs
->igp_metric
: 0;
2066 /* Remaining criteria are ignored */
2073 same_group(rte
*r
, u32 lpref
, u32 lasn
)
2075 return (r
->pref
== lpref
) && (bgp_get_neighbor(r
) == lasn
);
2079 use_deterministic_med(rte
*r
)
2081 struct proto
*P
= r
->attrs
->src
->proto
;
2082 return (P
->proto
== &proto_bgp
) && ((struct bgp_proto
*) P
)->cf
->deterministic_med
;
2086 bgp_rte_recalculate(rtable
*table
, net
*net
, rte
*new, rte
*old
, rte
*old_best
)
2089 rte
*key
= new ? new : old
;
2090 u32 lpref
= key
->pref
;
2091 u32 lasn
= bgp_get_neighbor(key
);
2092 int old_suppressed
= old
? old
->u
.bgp
.suppressed
: 0;
2095 * Proper RFC 4271 path selection is a bit complicated, it cannot be
2096 * implemented just by rte_better(), because it is not a linear
2097 * ordering. But it can be splitted to two levels, where the lower
2098 * level chooses the best routes in each group of routes from the
2099 * same neighboring AS and higher level chooses the best route (with
2100 * a slightly different ordering) between the best-in-group routes.
2102 * When deterministic_med is disabled, we just ignore this issue and
2103 * choose the best route by bgp_rte_better() alone. If enabled, the
2104 * lower level of the route selection is done here (for the group
2105 * to which the changed route belongs), all routes in group are
2106 * marked as suppressed, just chosen best-in-group is not.
2108 * Global best route selection then implements higher level by
2109 * choosing between non-suppressed routes (as they are always
2110 * preferred over suppressed routes). Routes from BGP protocols
2111 * that do not set deterministic_med are just never suppressed. As
2112 * they do not participate in the lower level selection, it is OK
2113 * that this fn is not called for them.
2115 * The idea is simple, the implementation is more problematic,
2116 * mostly because of optimizations in rte_recalculate() that
2117 * avoids full recalculation in most cases.
2119 * We can assume that at least one of new, old is non-NULL and both
2120 * are from the same protocol with enabled deterministic_med. We
2121 * group routes by both neighbor AS (lasn) and preference (lpref),
2122 * because bgp_rte_better() does not handle preference itself.
2125 /* If new and old are from different groups, we just process that
2126 as two independent events */
2127 if (new && old
&& !same_group(old
, lpref
, lasn
))
2130 i1
= bgp_rte_recalculate(table
, net
, NULL
, old
, old_best
);
2131 i2
= bgp_rte_recalculate(table
, net
, new, NULL
, old_best
);
2136 * We could find the best-in-group and then make some shortcuts like
2137 * in rte_recalculate, but as we would have to walk through all
2138 * net->routes just to find it, it is probably not worth. So we
2139 * just have one simple fast case that use just the old route.
2140 * We also set suppressed flag to avoid using it in bgp_rte_better().
2144 new->u
.bgp
.suppressed
= 1;
2148 old
->u
.bgp
.suppressed
= 1;
2150 /* The fast case - replace not best with worse (or remove not best) */
2151 if (old_suppressed
&& !(new && bgp_rte_better(new, old
)))
2155 /* The default case - find a new best-in-group route */
2156 r
= new; /* new may not be in the list */
2157 for (s
=net
->routes
; rte_is_valid(s
); s
=s
->next
)
2158 if (use_deterministic_med(s
) && same_group(s
, lpref
, lasn
))
2160 s
->u
.bgp
.suppressed
= 1;
2161 if (!r
|| bgp_rte_better(s
, r
))
2165 /* Simple case - the last route in group disappears */
2169 /* Found if new is mergable with best-in-group */
2170 if (new && (new != r
) && bgp_rte_mergable(r
, new))
2171 new->u
.bgp
.suppressed
= 0;
2173 /* Found all existing routes mergable with best-in-group */
2174 for (s
=net
->routes
; rte_is_valid(s
); s
=s
->next
)
2175 if (use_deterministic_med(s
) && same_group(s
, lpref
, lasn
))
2176 if ((s
!= r
) && bgp_rte_mergable(r
, s
))
2177 s
->u
.bgp
.suppressed
= 0;
2179 /* Found best-in-group */
2180 r
->u
.bgp
.suppressed
= 0;
2183 * There are generally two reasons why we have to force
2184 * recalculation (return 1): First, the new route may be wrongfully
2185 * chosen to be the best in the first case check in
2186 * rte_recalculate(), this may happen only if old_best is from the
2187 * same group. Second, another (different than new route)
2188 * best-in-group is chosen and that may be the proper best (although
2189 * rte_recalculate() without ignore that possibility).
2191 * There are three possible cases according to whether the old route
2192 * was the best in group (OBG, i.e. !old_suppressed) and whether the
2193 * new route is the best in group (NBG, tested by r == new). These
2194 * cases work even if old or new is NULL.
2196 * NBG -> new is a possible candidate for the best route, so we just
2197 * check for the first reason using same_group().
2199 * !NBG && OBG -> Second reason applies, return 1
2201 * !NBG && !OBG -> Best in group does not change, old != old_best,
2202 * rte_better(new, old_best) is false and therefore
2203 * the first reason does not apply, return 0
2207 return old_best
&& same_group(old_best
, lpref
, lasn
);
2209 return !old_suppressed
;
2213 bgp_rte_modify_stale(struct rte
*r
, struct linpool
*pool
)
2215 eattr
*a
= ea_find(r
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_COMMUNITY
));
2216 const struct adata
*ad
= a
? a
->u
.ptr
: NULL
;
2217 uint flags
= a
? a
->flags
: BAF_PARTIAL
;
2219 if (ad
&& int_set_contains(ad
, BGP_COMM_NO_LLGR
))
2222 if (ad
&& int_set_contains(ad
, BGP_COMM_LLGR_STALE
))
2225 r
= rte_cow_rta(r
, pool
);
2226 bgp_set_attr_ptr(&(r
->attrs
->eattrs
), pool
, BA_COMMUNITY
, flags
,
2227 int_set_add(pool
, ad
, BGP_COMM_LLGR_STALE
));
2235 * Reconstruct AS_PATH and AGGREGATOR according to RFC 6793 4.2.3
2238 bgp_process_as4_attrs(ea_list
**attrs
, struct linpool
*pool
)
2240 eattr
*p2
= bgp_find_attr(*attrs
, BA_AS_PATH
);
2241 eattr
*p4
= bgp_find_attr(*attrs
, BA_AS4_PATH
);
2242 eattr
*a2
= bgp_find_attr(*attrs
, BA_AGGREGATOR
);
2243 eattr
*a4
= bgp_find_attr(*attrs
, BA_AS4_AGGREGATOR
);
2245 /* First, unset AS4_* attributes */
2246 if (p4
) bgp_unset_attr(attrs
, pool
, BA_AS4_PATH
);
2247 if (a4
) bgp_unset_attr(attrs
, pool
, BA_AS4_AGGREGATOR
);
2249 /* Handle AGGREGATOR attribute */
2252 u32 a2_asn
= get_u32(a2
->u
.ptr
->data
);
2254 /* If routes were aggregated by an old router, then AS4_PATH and
2255 AS4_AGGREGATOR are invalid. In that case we give up. */
2256 if (a2_asn
!= AS_TRANS
)
2259 /* Use AS4_AGGREGATOR instead of AGGREGATOR */
2260 a2
->u
.ptr
= a4
->u
.ptr
;
2263 /* Handle AS_PATH attribute */
2266 /* Both as_path_getlen() and as_path_cut() take AS_CONFED* as zero length */
2267 int p2_len
= as_path_getlen(p2
->u
.ptr
);
2268 int p4_len
= as_path_getlen(p4
->u
.ptr
);
2270 /* AS_PATH is too short, give up */
2271 if (p2_len
< p4_len
)
2274 /* Merge AS_PATH and AS4_PATH */
2275 struct adata
*apc
= as_path_cut(pool
, p2
->u
.ptr
, p2_len
- p4_len
);
2276 p2
->u
.ptr
= as_path_merge(pool
, apc
, p4
->u
.ptr
);
2281 bgp_get_attr(eattr
*a
, byte
*buf
, int buflen
)
2283 uint i
= EA_ID(a
->id
);
2284 const struct bgp_attr_desc
*d
;
2287 if (bgp_attr_known(i
))
2289 d
= &bgp_attr_table
[i
];
2290 len
= bsprintf(buf
, "%s", d
->name
);
2296 d
->format(a
, buf
, buflen
- len
- 2);
2302 bsprintf(buf
, "%02x%s", i
, (a
->flags
& BAF_TRANSITIVE
) ? " [t]" : "");
2307 bgp_get_route_info(rte
*e
, byte
*buf
)
2309 eattr
*p
= ea_find(e
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_AS_PATH
));
2310 eattr
*o
= ea_find(e
->attrs
->eattrs
, EA_CODE(PROTOCOL_BGP
, BA_ORIGIN
));
2313 buf
+= bsprintf(buf
, " (%d", e
->pref
);
2315 if (e
->u
.bgp
.suppressed
)
2316 buf
+= bsprintf(buf
, "-");
2319 buf
+= bsprintf(buf
, "s");
2321 u64 metric
= bgp_total_aigp_metric(e
);
2322 if (metric
< BGP_AIGP_MAX
)
2324 buf
+= bsprintf(buf
, "/%lu", metric
);
2326 else if (e
->attrs
->igp_metric
)
2328 if (!rte_resolvable(e
))
2329 buf
+= bsprintf(buf
, "/-");
2330 else if (e
->attrs
->igp_metric
>= IGP_METRIC_UNKNOWN
)
2331 buf
+= bsprintf(buf
, "/?");
2333 buf
+= bsprintf(buf
, "/%d", e
->attrs
->igp_metric
);
2335 buf
+= bsprintf(buf
, ") [");
2337 if (p
&& as_path_get_last(p
->u
.ptr
, &origas
))
2338 buf
+= bsprintf(buf
, "AS%u", origas
);
2340 buf
+= bsprintf(buf
, "%c", "ie?"[o
->u
.data
]);