From 11cb620266035ffbe17b21c4a174380cb8b6a521 Mon Sep 17 00:00:00 2001 From: Ondrej Zajicek Date: Sun, 26 Oct 2008 22:36:08 +0100 Subject: [PATCH] Implementation of 4B ASN support for BGP --- filter/config.Y | 11 +- filter/filter.c | 26 ++- nest/a-path.c | 211 +++++++++++++++++++---- nest/attrs.h | 21 ++- proto/bgp/attrs.c | 404 +++++++++++++++++++++++++++++++++++++------- proto/bgp/bgp.c | 5 + proto/bgp/bgp.h | 13 +- proto/bgp/config.Y | 8 +- proto/bgp/packets.c | 153 +++++++++++++---- 9 files changed, 695 insertions(+), 157 deletions(-) diff --git a/filter/config.Y b/filter/config.Y index d4bf44cc..fdfb2e74 100644 --- a/filter/config.Y +++ b/filter/config.Y @@ -39,7 +39,6 @@ CF_KEYWORDS(FUNCTION, PRINT, PRINTN, UNSET, RETURN, %type set_atom fprefix fprefix_s fipa %type decls declsn one_decl function_params %type bgp_path -%type bgp_one CF_GRAMMAR @@ -273,14 +272,12 @@ switch_body: /* EMPTY */ { $$ = NULL; } /* CONST '(' expr ')' { $$ = f_new_inst(); $$->code = 'c'; $$->aux = T_INT; $$->a2.i = $3; } */ -bgp_one: - NUM { $$ = $1; } - | '?' { $$ = PM_ANY; } - ; bgp_path: - bgp_one { $$ = cfg_alloc(sizeof(struct f_path_mask)); $$->next = NULL; $$->val = $1; } - | bgp_one bgp_path { $$ = cfg_alloc(sizeof(struct f_path_mask)); $$->next = $2; $$->val = $1; } + NUM { $$ = cfg_alloc(sizeof(struct f_path_mask)); $$->next = NULL; $$->val = $1; $$->any = 0; } + | '?' { $$ = cfg_alloc(sizeof(struct f_path_mask)); $$->next = NULL; $$->val = 0; $$->any = 1; } + | NUM bgp_path { $$ = cfg_alloc(sizeof(struct f_path_mask)); $$->next = $2; $$->val = $1; $$->any = 0; } + | '?' bgp_path { $$ = cfg_alloc(sizeof(struct f_path_mask)); $$->next = $2; $$->val = 0; $$->any = 1; } ; constant: diff --git a/filter/filter.c b/filter/filter.c index 9cde3d96..7893d9ae 100644 --- a/filter/filter.c +++ b/filter/filter.c @@ -69,6 +69,30 @@ pm_path_compare(struct f_path_mask *m1, struct f_path_mask *m2) } } +static void +pm_format(struct f_path_mask *p, byte *buf, unsigned int size) +{ + byte *end = buf + size - 16; + + while (p) + { + if (buf > end) + { + strcpy(buf, " ..."); + return; + } + + if (p->any) + buf += bsprintf(buf, "? "); + else + buf += bsprintf(buf, "%u ", p->val); + + p = p->next; + } + + *buf = 0; +} + /** * val_compare - compare two values * @v1: first value @@ -224,7 +248,7 @@ val_print(struct f_val v) case T_ENUM: PRINTF( "(enum %x)%d", v.type, v.val.i ); break; case T_PATH: as_path_format(v.val.ad, buf2, 1020); PRINTF( "(path %s)", buf2 ); break; case T_CLIST: int_set_format(v.val.ad, buf2, 1020); PRINTF( "(clist %s)", buf2 ); break; - case T_PATH_MASK: debug( "(pathmask " ); { struct f_path_mask *p = v.val.path_mask; while (p) { debug("%d ", p->val); p=p->next; } debug(")" ); } break; + case T_PATH_MASK: pm_format(v.val.path_mask, buf2, 1020); PRINTF( "(pathmask %s)", buf2 ); break; default: PRINTF( "[unknown type %x]", v.type ); #undef PRINTF } diff --git a/nest/a-path.c b/nest/a-path.c index 1b08f809..f4666911 100644 --- a/nest/a-path.c +++ b/nest/a-path.c @@ -14,38 +14,139 @@ #include "lib/unaligned.h" #include "lib/string.h" -struct adata * -as_path_prepend(struct linpool *pool, struct adata *olda, int as) + +/* Global AS4 support, shared by all BGP instances. + * This specifies whether BA_AS_PATH attributes contain 2 or 4 B per ASN + */ + +int bgp_as4_support = 1; + +static void +put_as(byte *data, u32 as) { + if (bgp_as4_support) + put_u32(data, as); + else if (as <= 0xFFFF) + put_u16(data, as); + else + bug("put_as: Try to put 32bit AS to 16bit AS Path"); +} + +static inline u32 +get_as(byte *data) +{ + return bgp_as4_support ? get_u32(data) : get_u16(data); +} + +struct adata * +as_path_prepend(struct linpool *pool, struct adata *olda, u32 as) +{ + int bs = bgp_as4_support ? 4 : 2; struct adata *newa; - if (olda->length && olda->data[0] == AS_PATH_SEQUENCE && - olda->data[1] < 255) /* Starting with sequence => just prepend the AS number */ + if (olda->length && olda->data[0] == AS_PATH_SEQUENCE && olda->data[1] < 255) + /* Starting with sequence => just prepend the AS number */ { - newa = lp_alloc(pool, sizeof(struct adata) + olda->length + 2); - newa->length = olda->length + 2; - newa->data[0] = 2; + int nl = olda->length + bs; + newa = lp_alloc(pool, sizeof(struct adata) + nl); + newa->length = nl; + newa->data[0] = AS_PATH_SEQUENCE; newa->data[1] = olda->data[1] + 1; - memcpy(newa->data+4, olda->data+2, olda->length-2); + memcpy(newa->data + bs + 2, olda->data + 2, olda->length - 2); } - else /* Create new path segment */ + else /* Create new path segment */ { - newa = lp_alloc(pool, sizeof(struct adata) + olda->length + 4); - newa->length = olda->length + 4; - newa->data[0] = 2; + int nl = olda->length + bs + 2; + newa = lp_alloc(pool, sizeof(struct adata) + nl); + newa->length = nl; + newa->data[0] = AS_PATH_SEQUENCE; newa->data[1] = 1; - memcpy(newa->data+4, olda->data, olda->length); + memcpy(newa->data + bs + 2, olda->data, olda->length); } - put_u16(newa->data+2, as); + put_as(newa->data + 2, as); return newa; } +int +as_path_convert_to_old(struct adata *path, byte *dst, int *new_used) +{ + byte *src = path->data; + byte *src_end = src + path->length; + byte *dst_start = dst; + u32 as; + int i, n; + *new_used = 0; + + while (src < src_end) + { + n = src[1]; + *dst++ = *src++; + *dst++ = *src++; + + for(i=0; i 0xFFFF) + { + as = AS_TRANS; + *new_used = 1; + } + put_u16(dst, as); + src += 4; + dst += 2; + } + } + + return dst - dst_start; +} + +int +as_path_convert_to_new(struct adata *path, byte *dst, int req_as) +{ + byte *src = path->data; + byte *src_end = src + path->length; + byte *dst_start = dst; + u32 as; + int i, t, n; + + + while ((src < src_end) && (req_as > 0)) + { + t = *src++; + n = *src++; + + if (t == AS_PATH_SEQUENCE) + { + if (n > req_as) + n = req_as; + + req_as -= n; + } + else // t == AS_PATH_SET + req_as--; + + *dst++ = t; + *dst++ = n; + + for(i=0; idata; byte *e = p + path->length; - byte *end = buf + size - 8; + byte *end = buf + size - 16; int sp = 1; int l, isset; @@ -69,8 +170,8 @@ as_path_format(struct adata *path, byte *buf, unsigned int size) { if (!sp) *buf++ = ' '; - buf += bsprintf(buf, "%d", get_u16(p)); - p += 2; + buf += bsprintf(buf, "%u", get_as(p)); + p += bs; sp = 0; } if (isset) @@ -86,6 +187,7 @@ as_path_format(struct adata *path, byte *buf, unsigned int size) int as_path_getlen(struct adata *path) { + int bs = bgp_as4_support ? 4 : 2; int res = 0; u8 *p = path->data; u8 *q = p+path->length; @@ -95,8 +197,8 @@ as_path_getlen(struct adata *path) { switch (*p++) { - case AS_PATH_SET: len = *p++; res++; p += 2*len; break; - case AS_PATH_SEQUENCE: len = *p++; res+=len; p += 2*len; break; + case AS_PATH_SET: len = *p++; res++; p += bs * len; break; + case AS_PATH_SEQUENCE: len = *p++; res += len; p += bs * len; break; default: bug("as_path_getlen: Invalid path segment"); } } @@ -104,9 +206,11 @@ as_path_getlen(struct adata *path) } int -as_path_get_first(struct adata *path) +as_path_get_first(struct adata *path, u32 *orig_as) { - int res = -1; + int bs = bgp_as4_support ? 4 : 2; + int found = 0; + u32 res = 0; u8 *p = path->data; u8 *q = p+path->length; int len; @@ -117,36 +221,70 @@ as_path_get_first(struct adata *path) { case AS_PATH_SET: if (len = *p++) - res = get_u16(p); - p += 2*len; + { + found = 1; + res = get_as(p); + p += bs * len; + } break; case AS_PATH_SEQUENCE: if (len = *p++) - res = get_u16(p+2*(len-1)); - p += 2*len; + { + found = 1; + res = get_as(p + bs * (len - 1)); + p += bs * len; + } break; default: bug("as_path_get_first: Invalid path segment"); } } - return res; + + *orig_as = res; + return found; } +int +as_path_is_member(struct adata *path, u32 as) +{ + int bs = bgp_as4_support ? 4 : 2; + u8 *p = path->data; + u8 *q = p+path->length; + int i, n; + + while (pnext; if (!mask) return next == q; \ - asterisk = (mask->val == PM_ANY); \ + asterisk = mask->any; \ if (asterisk) { mask = mask->next; if (!mask) { return 1; } } \ } while(0) int as_path_match(struct adata *path, struct f_path_mask *mask) { + int bs = bgp_as4_support ? 4 : 2; int i; int asterisk = 0; u8 *p = path->data; u8 *q = p+path->length; int len; u8 *next; + u32 as; - asterisk = (mask->val == PM_ANY); + asterisk = mask->any; if (asterisk) { mask = mask->next; if (!mask) return 1; } @@ -156,20 +294,21 @@ as_path_match(struct adata *path, struct f_path_mask *mask) len = *p++; { u8 *p_save = p; - next = p_save + 2*len; + next = p_save + bs * len; retry: p = p_save; for (i=0; ival)) { + as = get_as(p); + if (asterisk && (as == mask->val)) { MASK_PLUS; goto retry; } - if (!asterisk && (get_u16(p) == mask->val)) { + if (!asterisk && (as == mask->val)) { p = next; MASK_PLUS; goto okay; } - p+=2; + p += bs; } if (!asterisk) return 0; @@ -180,15 +319,15 @@ as_path_match(struct adata *path, struct f_path_mask *mask) case AS_PATH_SEQUENCE: len = *p++; for (i=0; ival)) + as = get_as(p); + if (asterisk && (as == mask->val)) MASK_PLUS; else if (!asterisk) { - if (get_u16(p) != mask->val) + if (as != mask->val) return 0; MASK_PLUS; } - p+=2; + p += bs; } break; diff --git a/nest/attrs.h b/nest/attrs.h index abd6b9e9..f63f2e45 100644 --- a/nest/attrs.h +++ b/nest/attrs.h @@ -14,16 +14,29 @@ #define AS_PATH_SET 1 /* Types of path segments */ #define AS_PATH_SEQUENCE 2 -struct adata *as_path_prepend(struct linpool *pool, struct adata *olda, int as); +#define AS_PATH_MAXLEN 10000 + +#define AS_TRANS 23456 +/* AS_TRANS is used when we need to store 32bit ASN larger than 0xFFFF + * to 16bit slot (like in 16bit AS_PATH). See RFC 4893 for details + */ + +struct adata *as_path_prepend(struct linpool *pool, struct adata *olda, u32 as); +int as_path_convert_to_old(struct adata *path, byte *dst, int *new_used); +int as_path_convert_to_new(struct adata *path, byte *dst, int req_as); void as_path_format(struct adata *path, byte *buf, unsigned int size); int as_path_getlen(struct adata *path); -int as_path_get_first(struct adata *path); +int as_path_get_first(struct adata *path, u32 *orig_as); +int as_path_is_member(struct adata *path, u32 as); + struct f_path_mask { struct f_path_mask *next; - int val; + u32 val; + int any; }; -#define PM_ANY -1 + +// #define PM_ANY -1 int as_path_match(struct adata *path, struct f_path_mask *mask); diff --git a/proto/bgp/attrs.c b/proto/bgp/attrs.c index 30699f84..48cb9dd5 100644 --- a/proto/bgp/attrs.c +++ b/proto/bgp/attrs.c @@ -55,21 +55,37 @@ bgp_format_origin(eattr *a, byte *buf) } static int -bgp_check_path(struct bgp_proto *p UNUSED, byte *a, int len) +bgp_check_path(byte *a, int len, int bs, int errcode) { while (len) { DBG("Path segment %02x %02x\n", a[0], a[1]); if (len < 2 || - a[0] != AS_PATH_SET && a[0] != AS_PATH_SEQUENCE || - 2*a[1] + 2 > len) - return 11; - len -= 2*a[1] + 2; - a += 2*a[1] + 2; + (a[0] != AS_PATH_SET && a[0] != AS_PATH_SEQUENCE) || + bs * a[1] + 2 > len) + return errcode; + len -= bs * a[1] + 2; + a += bs * a[1] + 2; } return 0; } +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); +} + +static int +bgp_check_as4_path(struct bgp_proto *p, byte *a, int len) +{ + if (bgp_as4_support && (! p->as4_support)) + return bgp_check_path(a, len, 4, 9); + else + return 0; +} + + static int bgp_check_next_hop(struct bgp_proto *p UNUSED, byte *a, int len) { @@ -87,6 +103,14 @@ bgp_check_next_hop(struct bgp_proto *p UNUSED, byte *a, int len) #endif } +static int +bgp_check_aggregator(struct bgp_proto *p UNUSED, 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_reach_nlri(struct bgp_proto *p UNUSED, byte *a UNUSED, int len UNUSED) { @@ -113,7 +137,7 @@ static struct attr_desc bgp_attr_table[] = { { "origin", 1, BAF_TRANSITIVE, EAF_TYPE_INT, 1, /* BA_ORIGIN */ bgp_check_origin, bgp_format_origin }, { "as_path", -1, BAF_TRANSITIVE, EAF_TYPE_AS_PATH, 1, /* BA_AS_PATH */ - bgp_check_path, NULL }, + bgp_check_as_path, NULL }, { "next_hop", 4, BAF_TRANSITIVE, EAF_TYPE_IP_ADDRESS, 1, /* BA_NEXT_HOP */ bgp_check_next_hop, NULL }, { "med", 4, BAF_OPTIONAL, EAF_TYPE_INT, 0, /* BA_MULTI_EXIT_DISC */ @@ -122,8 +146,8 @@ static struct attr_desc bgp_attr_table[] = { NULL, NULL }, { "atomic_aggr", 0, BAF_TRANSITIVE, EAF_TYPE_OPAQUE, 1, /* BA_ATOMIC_AGGR */ NULL, NULL }, - { "aggregator", 6, BAF_OPTIONAL | BAF_TRANSITIVE, EAF_TYPE_OPAQUE, 1, /* BA_AGGREGATOR */ - NULL, NULL }, + { "aggregator", -1, BAF_OPTIONAL | BAF_TRANSITIVE, EAF_TYPE_OPAQUE, 1, /* BA_AGGREGATOR */ + bgp_check_aggregator, NULL }, { "community", -1, BAF_OPTIONAL | BAF_TRANSITIVE, EAF_TYPE_INT_SET, 1, /* BA_COMMUNITY */ NULL, NULL }, { NULL, }, /* BA_ORIGINATOR_ID */ @@ -135,8 +159,18 @@ static struct attr_desc bgp_attr_table[] = { bgp_check_reach_nlri, NULL }, { "mp_unreach_nlri", -1, BAF_OPTIONAL, EAF_TYPE_OPAQUE, 1, /* BA_MP_UNREACH_NLRI */ bgp_check_unreach_nlri, NULL }, + { NULL, }, /* BA_EXTENDED_COMM */ + { "as4_path", -1, BAF_OPTIONAL | BAF_TRANSITIVE, EAF_TYPE_OPAQUE, 1, /* BA_AS4_PATH */ + bgp_check_as4_path, NULL }, + { "as4_aggregator", 8, BAF_OPTIONAL | BAF_TRANSITIVE, EAF_TYPE_OPAQUE, 1, /* BA_AS4_PATH */ + NULL, NULL } }; +/* BA_AS4_PATH is type EAF_TYPE_OPAQUE and not type EAF_TYPE_AS_PATH because + * EAF_TYPE_AS_PATH is supposed to have different format (2 or 4 B for each ASN) + * depending on bgp_as4_support variable. + */ + #define ATTR_KNOWN(code) ((code) < ARRAY_SIZE(bgp_attr_table) && bgp_attr_table[code].name) static byte * @@ -170,8 +204,90 @@ bgp_attach_attr(ea_list **to, struct linpool *pool, unsigned attr, unsigned val) return bgp_set_attr(a->attrs, pool, attr, val); } +static int +bgp_encode_attr_hdr(byte *dst, unsigned int flags, unsigned code, int len) +{ + int wlen; + + DBG("\tAttribute %02x (%d bytes, flags %02x)\n", code, len, flags); + + if (len < 256) + { + *dst++ = flags; + *dst++ = code; + *dst++ = len; + wlen = 3; + } + else + { + *dst++ = flags | BAF_EXT_LEN; + *dst++ = code; + put_u16(dst, len); + wlen = 4; + } + + return wlen; +} + +static void +aggregator_convert_to_old(struct adata *aggr, byte *dst, int *new_used) +{ + byte *src = aggr->data; + *new_used = 0; + + u32 as = get_u32(src); + if (as > 0xFFFF) + { + as = AS_TRANS; + *new_used = 1; + } + put_u16(dst, as); + + /* Copy IPv4 address */ + memcpy(dst + 2, src + 4, 4); +} + +static void +aggregator_convert_to_new(struct adata *aggr, byte *dst) +{ + byte *src = aggr->data; + + u32 as = get_u16(src); + put_u32(dst, as); + + /* Copy IPv4 address */ + memcpy(dst + 4, src + 2, 4); +} + +static int +bgp_get_attr_len(eattr *a) +{ + int len; + if (ATTR_KNOWN(EA_ID(a->id))) + { + int code = EA_ID(a->id); + struct attr_desc *desc = &bgp_attr_table[code]; + len = desc->expected_length; + if (len < 0) + { + ASSERT(!(a->type & EAF_EMBEDDED)); + len = a->u.ptr->length; + } + } + else + { + ASSERT((a->type & EAF_TYPE_MASK) == EAF_TYPE_OPAQUE); + len = a->u.ptr->length; + } + + return len; +} + +#define ADVANCE(w, r, l) do { r -= l; w += l; } while (0) + /** * bgp_encode_attrs - encode BGP attributes + * @p: BGP instance * @w: buffer * @attrs: a list of extended attributes * @remains: remaining space in the buffer @@ -182,11 +298,11 @@ bgp_attach_attr(ea_list **to, struct linpool *pool, unsigned attr, unsigned val) * Result: Length of the attribute block generated. */ unsigned int -bgp_encode_attrs(byte *w, ea_list *attrs, int remains) +bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int remains) { unsigned int i, code, flags; byte *start = w; - int len; + int len, rv; for(i=0; icount; i++) { @@ -198,43 +314,90 @@ bgp_encode_attrs(byte *w, ea_list *attrs, int remains) if (code == BA_NEXT_HOP) continue; #endif - flags = a->flags & (BAF_OPTIONAL | BAF_TRANSITIVE | BAF_PARTIAL); - if (ATTR_KNOWN(code)) + + /* When AS4-aware BGP speaker is talking to non-AS4-aware BGP speaker, + * 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)) { - struct attr_desc *desc = &bgp_attr_table[code]; - len = desc->expected_length; - if (len < 0) - { - ASSERT(!(a->type & EAF_EMBEDDED)); - len = a->u.ptr->length; - } - } - else - { - ASSERT((a->type & EAF_TYPE_MASK) == EAF_TYPE_OPAQUE); len = a->u.ptr->length; + + if (remains < (len + 4)) + goto err_no_buffer; + + /* Using temporary buffer because don't know a length of created attr + * and therefore a length of a header. Perhaps i should better always + * use BAF_EXT_LEN. */ + + byte buf[len]; + int new_used; + int nl = as_path_convert_to_old(a->u.ptr, buf, &new_used); + + rv = bgp_encode_attr_hdr(w, BAF_TRANSITIVE, BA_AS_PATH, nl); + ADVANCE(w, remains, rv); + memcpy(w, buf, nl); + ADVANCE(w, remains, nl); + + if (! new_used) + continue; + + if (remains < (len + 4)) + goto err_no_buffer; + + /* We should discard AS_CONFED_SEQUENCE or AS_CONFED_SET path segments + * here but we don't support confederations and such paths we already + * discarded in bgp_check_as_path(). + */ + + rv = bgp_encode_attr_hdr(w, BAF_OPTIONAL | BAF_TRANSITIVE, BA_AS4_PATH, len); + ADVANCE(w, remains, rv); + memcpy(w, a->u.ptr->data, len); + ADVANCE(w, remains, len); + + continue; } - DBG("\tAttribute %02x (type %02x, %d bytes, flags %02x)\n", code, a->type, len, flags); + + /* The same issue with AGGREGATOR attribute */ + if ((code == BA_AGGREGATOR) && bgp_as4_support && (! p->as4_support)) + { + int new_used; + + len = 6; + if (remains < (len + 3)) + goto err_no_buffer; + + rv = bgp_encode_attr_hdr(w, BAF_OPTIONAL | BAF_TRANSITIVE, BA_AGGREGATOR, len); + ADVANCE(w, remains, rv); + aggregator_convert_to_old(a->u.ptr, w, &new_used); + ADVANCE(w, remains, len); + + if (! new_used) + continue; + + len = 8; + if (remains < (len + 3)) + goto err_no_buffer; + + rv = bgp_encode_attr_hdr(w, BAF_OPTIONAL | BAF_TRANSITIVE, BA_AS4_AGGREGATOR, len); + ADVANCE(w, remains, rv); + memcpy(w, a->u.ptr->data, len); + ADVANCE(w, remains, len); + + continue; + } + + /* Standard path continues here ... */ + + flags = a->flags & (BAF_OPTIONAL | BAF_TRANSITIVE | BAF_PARTIAL); + len = bgp_get_attr_len(a); + if (remains < len + 4) - { - log(L_ERR "BGP: attribute list too long, ignoring the remaining attributes"); - break; - } - if (len < 256) - { - *w++ = flags; - *w++ = code; - *w++ = len; - remains -= 3; - } - else - { - *w++ = flags | BAF_EXT_LEN; - *w++ = code; - put_u16(w, len); - w += 2; - remains -= 4; - } + goto err_no_buffer; + + rv = bgp_encode_attr_hdr(w, flags, code, len); + ADVANCE(w, remains, rv); + switch (a->type & EAF_TYPE_MASK) { case EAF_TYPE_INT: @@ -266,10 +429,13 @@ bgp_encode_attrs(byte *w, ea_list *attrs, int remains) default: bug("bgp_encode_attrs: unknown attribute type %02x", a->type); } - remains -= len; - w += len; + ADVANCE(w, remains, len); } return w - start; + + err_no_buffer: + log(L_ERR "BGP: attribute list too long, ignoring the remaining attributes"); + return w - start; } static void @@ -566,10 +732,14 @@ bgp_create_attrs(struct bgp_proto *p, rte *e, ea_list **attrs, struct linpool *p bgp_set_attr(ea->attrs+1, pool, BA_AS_PATH, 0); else { - z = bgp_set_attr(ea->attrs+1, pool, BA_AS_PATH, 4); + z = bgp_set_attr(ea->attrs+1, pool, BA_AS_PATH, bgp_as4_support ? 6 : 4); z[0] = AS_PATH_SEQUENCE; z[1] = 1; /* 1 AS */ - put_u16(z+2, p->local_as); + + if (bgp_as4_support) + put_u32(z+2, p->local_as); + else + put_u16(z+2, p->local_as); } z = bgp_set_attr(ea->attrs+2, pool, BA_NEXT_HOP, sizeof(ip_addr)); @@ -670,8 +840,8 @@ bgp_rte_better(rte *new, rte *old) { x = ea_find(new->attrs->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)); y = ea_find(old->attrs->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)); - n = x ? as_path_getlen(x->u.ptr) : 100000; - o = y ? as_path_getlen(y->u.ptr) : 100000; + n = x ? as_path_getlen(x->u.ptr) : AS_PATH_MAXLEN; + o = y ? as_path_getlen(y->u.ptr) : AS_PATH_MAXLEN; if (n < o) return 1; if (n > o) @@ -712,23 +882,118 @@ bgp_rte_better(rte *new, rte *old) static int bgp_path_loopy(struct bgp_proto *p, eattr *a) { - byte *path = a->u.ptr->data; - int len = a->u.ptr->length; - int i, n; + return as_path_is_member(a->u.ptr, p->local_as); +} - while (len > 0) + +static struct adata * +bgp_aggregator_convert_to_new(struct adata *old, struct linpool *pool) +{ + struct adata *newa = lp_alloc(pool, sizeof(struct adata) + 8); + newa->length = 8; + aggregator_convert_to_new(old, newa->data); + return newa; +} + + +/* Take last req_as ASNs from path old2 (in 2B format), convert to 4B format + * and append path old4 (in 4B format). + */ +static struct adata * +bgp_merge_as_paths(struct adata *old2, struct adata *old4, int req_as, struct linpool *pool) +{ + byte buf[old2->length * 2]; + + int ol = as_path_convert_to_new(old2, buf, req_as); + int nl = ol + (old4 ? old4->length : 0); + + struct adata *newa = lp_alloc(pool, sizeof(struct adata) + nl); + newa->length = nl; + memcpy(newa->data, buf, ol); + if (old4) memcpy(newa->data + ol, old4->data, old4->length); + + return newa; +} + + +/* Reconstruct 4B AS_PATH and AGGREGATOR according to RFC4893 4.2.3 */ +static void +bgp_reconstruct_4b_atts(struct bgp_proto *p, rta *a, struct linpool *pool) +{ + eattr *p2 =ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_AS_PATH)); + eattr *p4 =ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_AS4_PATH)); + eattr *a2 =ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_AGGREGATOR)); + eattr *a4 =ea_find(a->eattrs, EA_CODE(EAP_BGP, BA_AS4_AGGREGATOR)); + + if (a2) { - n = path[1]; - len -= 2 + 2*n; - path += 2; - for(i=0; iu.ptr->data); + + if (a4) { - if (get_u16(path) == p->local_as) - return 1; - path += 2; + if (a2_as != AS_TRANS) + { + /* Routes were aggregated by old router and therefore AS4_PATH + * and AS4_AGGREGATOR is invalid + * + * Convert AS_PATH and AGGREGATOR to 4B format and finish. + */ + + a2->u.ptr = bgp_aggregator_convert_to_new(a2->u.ptr, pool); + p2->u.ptr = bgp_merge_as_paths(p2->u.ptr, NULL, AS_PATH_MAXLEN, pool); + + return; + } + else + { + /* Common case, use AS4_AGGREGATOR attribute */ + a2->u.ptr = a4->u.ptr; + } + } + else + { + /* Common case, use old AGGREGATOR attribute */ + a2->u.ptr = bgp_aggregator_convert_to_new(a2->u.ptr, pool); + + if (a2_as == AS_TRANS) + log(L_WARN "BGP: AGGREGATOR attribute contain AS_TRANS, but AS4_AGGREGATOR is missing"); } } - return 0; + else + if (a4) + log(L_WARN "BGP: AS4_AGGREGATOR attribute received, but AGGREGATOR attribute is missing"); + + int p2_len = as_path_getlen(p2->u.ptr); + int p4_len = p4 ? as_path_getlen(p4->u.ptr) : AS_PATH_MAXLEN; + + if (p2_len < p4_len) + p2->u.ptr = bgp_merge_as_paths(p2->u.ptr, NULL, AS_PATH_MAXLEN, pool); + else + p2->u.ptr = bgp_merge_as_paths(p2->u.ptr, p4->u.ptr, p2_len - p4_len, pool); + +} + +static void +bgp_remove_as4_attrs(struct bgp_proto *p, rta *a) +{ + unsigned id1 = EA_CODE(EAP_BGP, BA_AS4_PATH); + unsigned id2 = EA_CODE(EAP_BGP, BA_AS4_AGGREGATOR); + ea_list **el = &(a->eattrs); + + /* We know that ea_lists constructed in bgp_decode_attrs have one attribute per ea_list struct */ + while (*el != NULL) + { + unsigned fid = (*el)->attrs[0].id; + + if ((fid == id1) || (fid == id2)) + { + *el = (*el)->next; + if (p->as4_support) + log(L_WARN "BGP: Unexpected AS4_* attributes received"); + } + else + el = &((*el)->next); + } } /** @@ -883,6 +1148,15 @@ 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)) + bgp_reconstruct_4b_atts(bgp, a, pool); + + if (bgp_as4_support) + 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)); @@ -945,11 +1219,11 @@ bgp_get_route_info(rte *e, byte *buf, ea_list *attrs) { eattr *p = ea_find(attrs, EA_CODE(EAP_BGP, BA_AS_PATH)); eattr *o = ea_find(attrs, EA_CODE(EAP_BGP, BA_ORIGIN)); - int origas; + u32 origas; buf += bsprintf(buf, " (%d) [", e->pref); - if (p && (origas = as_path_get_first(p->u.ptr)) >= 0) - buf += bsprintf(buf, "AS%d", origas); + if (p && as_path_get_first(p->u.ptr, &origas)) + buf += bsprintf(buf, "AS%u", origas); if (o) buf += bsprintf(buf, "%c", "ie?"[o->u.data]); strcpy(buf, "]"); diff --git a/proto/bgp/bgp.c b/proto/bgp/bgp.c index cedd223b..ed2524c8 100644 --- a/proto/bgp/bgp.c +++ b/proto/bgp/bgp.c @@ -76,6 +76,7 @@ static void bgp_connect(struct bgp_proto *p); static void bgp_initiate(struct bgp_proto *p); static void bgp_setup_listen_sk(void); + static void bgp_close(struct bgp_proto *p UNUSED) { @@ -611,6 +612,10 @@ 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)) + cf_error("Local AS number out of range"); + if (!bgp_as4_support && (c->remote_as > 0xFFFF)) + cf_error("Neighbor AS number out of range"); } static void diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h index 6519db85..aa1bd107 100644 --- a/proto/bgp/bgp.h +++ b/proto/bgp/bgp.h @@ -16,7 +16,7 @@ struct eattr; struct bgp_config { struct proto_config c; - unsigned int local_as, remote_as; + u32 local_as, remote_as; ip_addr remote_ip; int multihop; /* Number of hops if multihop */ ip_addr multihop_via; /* Multihop: address to route to */ @@ -47,14 +47,16 @@ struct bgp_conn { byte *notify_data; int error_flag; /* Error state, ignore all input */ int primary; /* This connection is primary */ + u32 advertised_as; /* Temporary value for AS number received */ unsigned hold_time, keepalive_time; /* Times calculated from my and neighbor's requirements */ }; struct bgp_proto { struct proto p; struct bgp_config *cf; /* Shortcut to BGP configuration */ - unsigned local_as, remote_as; + u32 local_as, remote_as; int is_internal; /* Internal BGP connection (local_as == remote_as) */ + 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 */ struct bgp_conn *conn; /* Connection we have established */ @@ -100,6 +102,9 @@ struct bgp_bucket { extern struct linpool *bgp_linpool; +extern int bgp_as4_support; + + void bgp_start_timer(struct timer *t, int value); void bgp_check(struct bgp_config *c); void bgp_error(struct bgp_conn *c, unsigned code, unsigned subcode, byte *data, int len); @@ -122,7 +127,7 @@ int bgp_rte_better(struct rte *, struct rte *); void bgp_rt_notify(struct proto *, struct network *, struct rte *, struct rte *, struct ea_list *); int bgp_import_control(struct proto *, struct rte **, struct ea_list **, struct linpool *); void bgp_attr_init(struct bgp_proto *); -unsigned int bgp_encode_attrs(byte *w, struct ea_list *attrs, int remains); +unsigned int bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int remains); 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); @@ -165,6 +170,8 @@ void bgp_log_error(struct bgp_proto *p, char *msg, unsigned code, unsigned subco #define BA_MP_REACH_NLRI 0x0e /* [RFC2283] */ #define BA_MP_UNREACH_NLRI 0x0f #define BA_EXTENDED_COMM 0x10 /* draft-ramachandra-bgp-ext-communities */ +#define BA_AS4_PATH 0x11 /* [RFC4893] */ +#define BA_AS4_AGGREGATOR 0x12 /* BGP states */ diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y index 52ad731e..63dfb615 100644 --- a/proto/bgp/config.Y +++ b/proto/bgp/config.Y @@ -44,12 +44,10 @@ bgp_proto_start: proto_start BGP { bgp_proto: bgp_proto_start proto_name '{' | bgp_proto proto_item ';' - | bgp_proto LOCAL AS expr ';' { - if ($4 < 0 || $4 > 65535) cf_error("AS number out of range"); - BGP_CFG->local_as = $4; - } + | bgp_proto LOCAL AS expr ';' { BGP_CFG->local_as = $4; } | bgp_proto NEIGHBOR ipa AS expr ';' { - if ($5 < 0 || $5 > 65535) cf_error("AS number out of range"); + if (ipa_nonzero(BGP_CFG->remote_ip)) cf_error("Only one neighbor per BGP instance is allowed"); + BGP_CFG->remote_ip = $3; BGP_CFG->remote_as = $5; } 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" @@ -30,25 +31,10 @@ bgp_create_notification(struct bgp_conn *conn, byte *buf) return buf + 2 + conn->notify_size; } +#ifdef IPV6 static byte * -bgp_create_open(struct bgp_conn *conn, byte *buf) +bgp_put_cap_ipv6(struct bgp_conn *conn UNUSED, byte *buf) { - struct bgp_proto *p = conn->bgp; - - 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+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 */ @@ -56,7 +42,53 @@ bgp_create_open(struct bgp_conn *conn, byte *buf) *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 < 0xFFFF) ? p->local_as : AS_TRANS); + put_u16(buf+3, p->cf->hold_time); + put_u32(buf+5, p->local_id); + /* 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" },