BGP redesign

Integrated and extensible BGP with generalized AFI handling,
support for IPv4+IPv6 AFI and unicast+multicast SAFI.

configure.in

@@ -170,7 +170,7 @@ fi
 AC_SUBST(iproutedir)
 
 # all_protocols="$proto_bfd babel bgp ospf pipe radv rip static"
-all_protocols="$proto_bfd ospf pipe radv rip static"
+all_protocols="$proto_bfd bgp ospf pipe radv rip static"
 
 all_protocols=`echo $all_protocols | sed 's/ /,/g'`
 

filter/filter.c

@@ -1065,7 +1065,7 @@ interpret(struct f_inst *what)
       l->count = 1;
       l->attrs[0].id = code;
       l->attrs[0].flags = 0;
-      l->attrs[0].type = what->aux | EAF_ORIGINATED;
+      l->attrs[0].type = what->aux | EAF_ORIGINATED | EAF_FRESH;
 
       switch (what->aux & EAF_TYPE_MASK) {
       case EAF_TYPE_INT:

lib/birdlib.h

@@ -35,6 +35,7 @@
 #define DELTA(a,b) (((a)>=(b))?(a)-(b):(b)-(a))
 #define ARRAY_SIZE(a) (sizeof(a)/sizeof(*(a)))
 #define CALL(fn, args...) ({ if (fn) fn(args); })
+#define ADVANCE(w, r, l) ({ r -= l; w += l; })
 
 static inline int uint_cmp(uint i1, uint i2)
 { return (int)(i1 > i2) - (int)(i1 < i2); }

lib/ip.h

@@ -91,6 +91,7 @@ typedef ip6_addr ip_addr;
 #define ipa_to_u32(x) ip4_to_u32(ipa_to_ip4(x))
 
 #define ipa_is_ip4(a) ip6_is_v4mapped(a)
+#define ipa_is_ip6(a) (! ip6_is_v4mapped(a))
 
 #define IPA_NONE4 ipa_from_ip4(IP4_NONE)
 #define IPA_NONE6 ipa_from_ip6(IP6_NONE)

lib/lists.c

@@ -158,3 +158,15 @@ add_tail_list(list *to, list *l)
   q->next = &to->tail_node;
   to->tail = q;
 }
+
+LIST_INLINE uint
+list_length(list *l)
+{
+  uint len = 0;
+  node *n;
+
+  WALK_LIST(n, *l)
+    len++;
+
+  return len;
+}

lib/lists.h

@@ -80,6 +80,7 @@ void rem_node(node *);
 void add_tail_list(list *, list *);
 void init_list(list *);
 void insert_node(node *, node *);
+uint list_length(list *);
 #endif
 
 #endif

lib/net.c

@@ -109,6 +109,24 @@ net_compare(const net_addr *a, const net_addr *b)
   return 0;
 }
 
+#define NET_HASH(a,t) net_hash_##t((const net_addr_##t *) a)
+
+u32
+net_hash(const net_addr *n)
+{
+  switch (n->type)
+  {
+  case NET_IP4: return NET_HASH(n, ip4);
+  case NET_IP6: return NET_HASH(n, ip6);
+  case NET_VPN4: return NET_HASH(n, vpn4);
+  case NET_VPN6: return NET_HASH(n, vpn6);
+  case NET_ROA4: return NET_HASH(n, roa4);
+  case NET_ROA6: return NET_HASH(n, roa6);
+  default: bug("invalid type");
+  }
+}
+
+
 int
 net_validate(const net_addr *N)
 {

lib/net.h

@@ -321,6 +321,8 @@ static inline u32 net_hash_roa4(const net_addr_roa4 *n)
 static inline u32 net_hash_roa6(const net_addr_roa6 *n)
 { return ip6_hash(n->prefix) ^ ((u32) n->pxlen << 26); }
 
+u32 net_hash(const net_addr *a);
+
 
 static inline int net_validate_ip4(const net_addr_ip4 *n)
 {

lib/unaligned.h

@@ -69,4 +69,22 @@ put_u64(void *p, u64 x)
   memcpy(p+4, &xl, 4);
 }
 
+static inline void
+get_u32s(const void *p, u32 *x, int n)
+{
+  int i;
+  memcpy(x, p, 4*n);
+  for (i = 0; i < n; i++)
+    x[i] = ntohl(x[i]);
+}
+
+static inline void
+put_u32s(void *p, const u32 *x, int n)
+{
+  int i;
+  for (i = 0; i < n; i++)
+    put_u32((byte *) p + 4*i, x[i]);
+}
+
+
 #endif

nest/a-path.c

@@ -22,45 +22,56 @@
 #define get_as get_u32
 #define BS  4	/* Default block size of ASN (autonomous system number) */
 
-struct adata *
+#define BAD(DSC, VAL) ({ err_dsc = DSC; err_val = VAL; goto bad; })
-as_path_prepend(struct linpool *pool, struct adata *olda, u32 as)
-{
-  struct adata *newa;
 
-  if (olda->length && olda->data[0] == AS_PATH_SEQUENCE && olda->data[1] < 255)
+int
-    /* Starting with sequence => just prepend the AS number */
+as_path_valid(byte *data, uint len, int bs, char *err, uint elen)
 {
-      int nl = olda->length + BS;
+  byte *pos = data;
-      newa = lp_alloc(pool, sizeof(struct adata) + nl);
+  char *err_dsc = NULL;
-      newa->length = nl;
+  uint err_val = 0;
-      newa->data[0] = AS_PATH_SEQUENCE;
+
-      newa->data[1] = olda->data[1] + 1;
+  while (len)
-      memcpy(newa->data + BS + 2, olda->data + 2, olda->length - 2);
-    }
-  else /* Create new path segment */
   {
-      int nl = olda->length + BS + 2;
+    if (len < 2)
-      newa = lp_alloc(pool, sizeof(struct adata) + nl);
+      BAD("segment framing error", 0);
-      newa->length = nl;
+
-      newa->data[0] = AS_PATH_SEQUENCE;
+    /* Process one AS path segment */
-      newa->data[1] = 1;
+    uint type = pos[0];
-      memcpy(newa->data + BS + 2, olda->data, olda->length);
+    uint slen = 2 + bs * pos[1];
+
+    if (len < slen)
+      BAD("segment framing error", len);
+
+    /* XXXX handle CONFED segments */
+    if ((type != AS_PATH_SET) && (type != AS_PATH_SEQUENCE))
+      BAD("unknown segment", type);
+
+    if (pos[1] == 0)
+      BAD("zero-length segment", type);
+
+    pos += slen;
+    len -= slen;
   }
-  put_as(newa->data + 2, as);
+
-  return newa;
+  return 1;
+
+bad:
+  if (err)
+    if (bsnprintf(err, elen, "%s (%u) at %d", err_dsc, err_val, (int) (pos - data)) < 0)
+      err[0] = 0;
+
+  return 0;
 }
 
 int
-as_path_convert_to_old(struct adata *path, byte *dst, int *new_used)
+as_path_16to32(byte *dst, byte *src, uint len)
 {
-  byte *src = path->data;
+  byte *dst0 = dst;
-  byte *src_end = src + path->length;
+  byte *end = src + len;
-  byte *dst_start = dst;
+  uint i, n;
-  u32 as;
-  int i, n;
-  *new_used = 0;
 
-  while (src < src_end)
+  while (src < end)
   {
     n = src[1];
     *dst++ = *src++;
@@ -68,66 +79,256 @@ as_path_convert_to_old(struct adata *path, byte *dst, int *new_used)
 
     for (i = 0; i < n; i++)
     {
-	  as = get_u32(src);
+      put_u32(dst, get_u16(src));
-	  if (as > 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; i<n; i++)
-	{
-	  as = get_u16(src);
-	  put_u32(dst, as);
       src += 2;
       dst += 4;
     }
   }
 
-  return dst - dst_start;
+  return dst - dst0;
+}
+
+int
+as_path_32to16(byte *dst, byte *src, uint len)
+{
+  byte *dst0 = dst;
+  byte *end = src + len;
+  uint i, n;
+
+  while (src < end)
+  {
+    n = src[1];
+    *dst++ = *src++;
+    *dst++ = *src++;
+
+    for (i = 0; i < n; i++)
+    {
+      put_u16(dst, get_u32(src));
+      src += 4;
+      dst += 2;
+    }
+  }
+
+  return dst - dst0;
+}
+
+int
+as_path_contains_as4(const struct adata *path)
+{
+  const byte *pos = path->data;
+  const byte *end = pos + path->length;
+  uint i, n;
+
+  while (pos < end)
+  {
+    n = pos[1];
+    pos += 2;
+
+    for (i = 0; i < n; i++)
+    {
+      if (get_as(pos) > 0xFFFF)
+	return 1;
+
+      pos += BS;
+    }
+  }
+
+  return 0;
+}
+
+int
+as_path_contains_confed(const struct adata *path)
+{
+  const byte *pos = path->data;
+  const byte *end = pos + path->length;
+
+  while (pos < end)
+  {
+    uint type = pos[0];
+    uint slen = 2 + BS * pos[1];
+
+    if ((type == AS_PATH_CONFED_SEQUENCE) ||
+	(type == AS_PATH_CONFED_SET))
+      return 1;
+
+    pos += slen;
+  }
+
+  return 0;
+}
+
+static void
+as_path_strip_confed_(byte *dst, const byte *src, uint len)
+{
+  const byte *end = src + len;
+
+  while (src < end)
+  {
+    uint type = src[0];
+    uint slen = 2 + BS * src[1];
+
+    /* Copy regular segments */
+    if ((type == AS_PATH_SET) || (type == AS_PATH_SEQUENCE))
+    {
+      memcpy(dst, src, slen);
+      dst += slen;
+    }
+
+    src += slen;
+  }
+}
+
+struct adata *
+as_path_strip_confed(struct linpool *pool, const struct adata *op)
+{
+  struct adata *np = lp_alloc_adata(pool, op->length);
+  as_path_strip_confed_(np->data, op->data, op->length);
+  return np;
+}
+
+struct adata *
+as_path_prepend2(struct linpool *pool, const struct adata *op, int seq, u32 as, int strip)
+{
+  struct adata *np;
+  const byte *pos = op->data;
+  uint len = op->length;
+
+  if (len && (pos[0] == seq) && (pos[1] < 255))
+  {
+    /* Starting with matching segment => just prepend the AS number */
+    np = lp_alloc_adata(pool, len + BS);
+    np->data[0] = seq;
+    np->data[1] = pos[1] + 1;
+    put_as(np->data + 2, as);
+
+    uint dlen = BS * pos[1];
+    memcpy(np->data + 2 + BS, pos + 2, dlen);
+    ADVANCE(pos, len, 2 + dlen);
+  }
+  else
+  {
+    /* Create a new path segment */
+    np = lp_alloc_adata(pool, len + 2 + BS);
+    np->data[0] = seq;
+    np->data[1] = 1;
+    put_as(np->data + 2, as);
+  }
+
+  if (len)
+  {
+    byte *dst = np->data + 2 + BS * np->data[1];
+
+    if (strip)
+      as_path_strip_confed_(dst, pos, len);
+    else
+      memcpy(dst, pos, len);
+  }
+
+  return np;
+}
+
+
+struct adata *
+as_path_to_old(struct linpool *pool, const struct adata *path)
+{
+  struct adata *res = lp_alloc_adata(pool, path->length);
+  byte *pos = res->data;
+  byte *end = pos + res->length;
+  uint i, n;
+  u32 as;
+
+  /* Copy the whole path */
+  memcpy(res->data, path->data, path->length);
+
+  /* Replace 32-bit AS numbers with AS_TRANS */
+  while (pos < end)
+  {
+    n = pos[1];
+    pos += 2;
+
+    for (i = 0; i < n; i++)
+    {
+      as = get_as(pos);
+      if (as > 0xFFFF)
+	put_as(pos, AS_TRANS);
+
+      pos += BS;
+    }
+  }
+
+  return res;
+}
+
+/*
+ * Cut the path to the length @num, measured to the usual path metric. Note that
+ * AS_CONFED_* segments have zero length and must be added if they are on edge.
+ * In contrast to other as_path_* functions, @path is modified in place.
+ */
+void
+as_path_cut(struct adata *path, uint num)
+{
+  byte *pos = path->data;
+  byte *end = pos + path->length;
+
+  while (pos < end)
+  {
+    uint t = pos[0];
+    uint l = pos[1];
+    uint n = 0;
+
+    switch (t)
+    {
+    case AS_PATH_SET:			n = 1; break;
+    case AS_PATH_SEQUENCE:		n = l; break;
+    case AS_PATH_CONFED_SEQUENCE:	n = 0; break;
+    case AS_PATH_CONFED_SET:		n = 0; break;
+    default: bug("as_path_cut: Invalid path segment");
+    }
+
+    /* Cannot add whole segment, so try partial one and finish */
+    if (num < n)
+    {
+      if (num)
+      {
+	pos[1] = num;
+	pos += 2 + BS * num;
+      }
+
+      break;
+    }
+
+    num -= n;
+    pos += 2 + BS * l;
+  }
+
+  path->length = pos - path->data;
+}
+
+/*
+ * Merge (concatenate) paths @p1 and @p2 and return the result.
+ * In contrast to other as_path_* functions, @p1 and @p2 may be reused.
+ */
+struct adata *
+as_path_merge(struct linpool *pool, struct adata *p1, struct adata *p2)
+{
+  if (p1->length == 0)
+    return p2;
+
+  if (p2->length == 0)
+    return p1;
+
+  struct adata *res = lp_alloc_adata(pool, p1->length + p2->length);
+  memcpy(res->data, p1->data, p1->length);
+  memcpy(res->data + p1->length, p2->data, p2->length);
+
+  return res;
 }
 
 void
-as_path_format(struct adata *path, byte *buf, uint size)
+as_path_format(const struct adata *path, byte *buf, uint size)
 {
-  byte *p = path->data;
+  const byte *p = path->data;
-  byte *e = p + path->length;
+  const byte *e = p + path->length;
   byte *end = buf + size - 16;
   int sp = 1;
   int l, isset;
@@ -167,38 +368,41 @@ as_path_format(struct adata *path, byte *buf, uint size)
 }
 
 int
-as_path_getlen(struct adata *path)
+as_path_getlen(const struct adata *path)
 {
-  return as_path_getlen_int(path, BS);
+  const byte *pos = path->data;
-}
+  const byte *end = pos + path->length;
+  uint res = 0;
 
-int
+  while (pos < end)
-as_path_getlen_int(struct adata *path, int bs)
   {
-  int res = 0;
+    uint t = pos[0];
-  u8 *p = path->data;
+    uint l = pos[1];
-  u8 *q = p+path->length;
+    uint n = 0;
-  int len;
 
-  while (p<q)
+    switch (t)
     {
-      switch (*p++)
+    case AS_PATH_SET:			n = 1; break;
-	{
+    case AS_PATH_SEQUENCE:		n = l; break;
-	case AS_PATH_SET:      len = *p++; res++;      p += bs * len; break;
+    case AS_PATH_CONFED_SEQUENCE:	n = 0; break;
-	case AS_PATH_SEQUENCE: len = *p++; res += len; p += bs * len; break;
+    case AS_PATH_CONFED_SET:		n = 0; break;
     default: bug("as_path_getlen: Invalid path segment");
     }
+
+    res += n;
+    pos += 2 + BS * l;
   }
+
   return res;
 }
 
 int
-as_path_get_last(struct adata *path, u32 *orig_as)
+as_path_get_last(const struct adata *path, u32 *orig_as)
 {
   int found = 0;
   u32 res = 0;
-  u8 *p = path->data;
+  const u8 *p = path->data;
-  u8 *q = p+path->length;
+  const u8 *q = p+path->length;
   int len;
 
   while (p<q)
@@ -230,10 +434,10 @@ as_path_get_last(struct adata *path, u32 *orig_as)
 }
 
 u32
-as_path_get_last_nonaggregated(struct adata *path)
+as_path_get_last_nonaggregated(const struct adata *path)
 {
-  u8 *p = path->data;
+  const u8 *p = path->data;
-  u8 *q = p+path->length;
+  const u8 *q = p+path->length;
   u32 res = 0;
   int len;
 
@@ -257,11 +461,10 @@ as_path_get_last_nonaggregated(struct adata *path)
   return res;
 }
 
-
 int
-as_path_get_first(struct adata *path, u32 *last_as)
+as_path_get_first(const struct adata *path, u32 *last_as)
 {
-  u8 *p = path->data;
+  const u8 *p = path->data;
 
   if ((path->length == 0) || (p[0] != AS_PATH_SEQUENCE) || (p[1] == 0))
     return 0;
@@ -273,10 +476,10 @@ as_path_get_first(struct adata *path, u32 *last_as)
 }
 
 int
-as_path_contains(struct adata *path, u32 as, int min)
+as_path_contains(const struct adata *path, u32 as, int min)
 {
-  u8 *p = path->data;
+  const u8 *p = path->data;
-  u8 *q = p+path->length;
+  const u8 *q = p+path->length;
   int num = 0;
   int i, n;
 
@@ -296,10 +499,10 @@ as_path_contains(struct adata *path, u32 as, int min)
 }
 
 int
-as_path_match_set(struct adata *path, struct f_tree *set)
+as_path_match_set(const struct adata *path, struct f_tree *set)
 {
-  u8 *p = path->data;
+  const u8 *p = path->data;
-  u8 *q = p+path->length;
+  const u8 *q = p+path->length;
   int i, n;
 
   while (p<q)
@@ -325,8 +528,8 @@ as_path_filter(struct linpool *pool, struct adata *path, struct f_tree *set, u32
     return NULL;
 
   int len = path->length;
-  u8 *p = path->data;
+  const u8 *p = path->data;
-  u8 *q = path->data + len;
+  const u8 *q = path->data + len;
   u8 *d, *d2;
   int i, bt, sn, dn;
   u8 buf[len];
@@ -388,16 +591,16 @@ struct pm_pos
   u8 mark;
   union
   {
-    char *sp;
+    const char *sp;
     u32 asn;
   } val;
 };
 
 static int
-parse_path(struct adata *path, struct pm_pos *pos)
+parse_path(const struct adata *path, struct pm_pos *pos)
 {
-  u8 *p = path->data;
+  const u8 *p = path->data;
-  u8 *q = p + path->length;
+  const u8 *q = p + path->length;
   struct pm_pos *opos = pos;
   int i, len;
 
@@ -433,7 +636,6 @@ parse_path(struct adata *path, struct pm_pos *pos)
   return pos - opos;
 }
 
-
 static int
 pm_match(struct pm_pos *pos, u32 asn, u32 asn2)
 {
@@ -441,7 +643,7 @@ pm_match(struct pm_pos *pos, u32 asn, u32 asn2)
   if (! pos->set)
     return ((pos->val.asn >= asn) && (pos->val.asn <= asn2));
 
-  u8 *p = pos->val.sp;
+  const u8 *p = pos->val.sp;
   int len = *p++;
   int i;
 
@@ -500,7 +702,7 @@ pm_mark(struct pm_pos *pos, int i, int plen, int *nl, int *nh)
  * is marked.
  */
 int
-as_path_match(struct adata *path, struct f_path_mask *mask)
+as_path_match(const struct adata *path, struct f_path_mask *mask)
 {
   struct pm_pos pos[2048 + 1];
   int plen = parse_path(path, pos);

nest/a-set.c

@@ -7,6 +7,8 @@
  *	Can be freely distributed and used under the terms of the GNU GPL.
  */
 
+#include <stdlib.h>
+
 #include "nest/bird.h"
 #include "nest/route.h"
 #include "nest/attrs.h"
@@ -455,3 +457,85 @@ lc_set_union(struct linpool *pool, struct adata *l1, struct adata *l2)
   memcpy(res->data + l1->length, tmp, len);
   return res;
 }
+
+
+struct adata *
+ec_set_del_nontrans(struct linpool *pool, struct adata *set)
+{
+  adata *res = lp_alloc_adata(pool, set->length);
+  u32 *src = int_set_get_data(set);
+  u32 *dst = int_set_get_data(res);
+  int len = int_set_get_size(set);
+  int i;
+
+  /* Remove non-transitive communities (EC_TBIT set) */
+  for (i = 0; i < len; i += 2)
+  {
+    if (src[i] & EC_TBIT)
+      continue;
+
+    *dst++ = src[i];
+    *dst++ = src[i+1];
+  }
+
+  res->length = ((byte *) dst) - res->data;
+
+  return res;
+}
+
+static int
+int_set_cmp(const void *X, const void *Y)
+{
+  const u32 *x = X, *y = Y;
+  return (*x < *y) ? -1 : (*x > *y) ? 1 : 0;
+}
+
+struct adata *
+int_set_sort(struct linpool *pool, struct adata *src)
+{
+  struct adata *dst = lp_alloc_adata(pool, src->length);
+  memcpy(dst->data, src->data, src->length);
+  qsort(dst->data, dst->length / 4, 4, int_set_cmp);
+  return dst;
+}
+
+
+static int
+ec_set_cmp(const void *X, const void *Y)
+{
+  u64 x = ec_get(X, 0);
+  u64 y = ec_get(Y, 0);
+  return (x < y) ? -1 : (x > y) ? 1 : 0;
+}
+
+struct adata *
+ec_set_sort(struct linpool *pool, struct adata *src)
+{
+  struct adata *dst = lp_alloc_adata(pool, src->length);
+  memcpy(dst->data, src->data, src->length);
+  qsort(dst->data, dst->length / 8, 8, ec_set_cmp);
+  return dst;
+}
+
+
+static int
+lc_set_cmp(const void *X, const void *Y)
+{
+  const u32 *x = X, *y = Y;
+  if (x[0] != y[0])
+    return (x[0] > y[0]) ? 1 : -1;
+  if (x[1] != y[1])
+    return (x[1] > y[1]) ? 1 : -1;
+  if (x[2] != y[2])
+    return (x[2] > y[2]) ? 1 : -1;
+  return 0;
+}
+
+struct adata *
+lc_set_sort(struct linpool *pool, struct adata *src)
+{
+  struct adata *dst = lp_alloc_adata(pool, src->length);
+  memcpy(dst->data, src->data, src->length);
+  qsort(dst->data, dst->length / LCOMM_LENGTH, LCOMM_LENGTH, lc_set_cmp);
+  return dst;
+}

nest/attrs.h

@@ -10,6 +10,9 @@
 #define _BIRD_ATTRS_H_
 
 #include <stdint.h>
+#include "lib/unaligned.h"
+#include "nest/route.h"
+
 
 /* a-path.c */
 
@@ -27,19 +30,29 @@
 
 struct f_tree;
 
-struct adata *as_path_prepend(struct linpool *pool, struct adata *olda, u32 as);
+int as_path_valid(byte *data, uint len, int bs, char *err, uint elen);
-int as_path_convert_to_old(struct adata *path, byte *dst, int *new_used);
+int as_path_16to32(byte *dst, byte *src, uint len);
-int as_path_convert_to_new(struct adata *path, byte *dst, int req_as);
+int as_path_32to16(byte *dst, byte *src, uint len);
-void as_path_format(struct adata *path, byte *buf, uint size);
+int as_path_contains_as4(const struct adata *path);
-int as_path_getlen(struct adata *path);
+int as_path_contains_confed(const struct adata *path);
-int as_path_getlen_int(struct adata *path, int bs);
+struct adata *as_path_strip_confed(struct linpool *pool, const struct adata *op);
-int as_path_get_first(struct adata *path, u32 *orig_as);
+struct adata *as_path_prepend2(struct linpool *pool, const struct adata *op, int seq, u32 as, int strip);
-int as_path_get_last(struct adata *path, u32 *last_as);
+struct adata *as_path_to_old(struct linpool *pool, const struct adata *path);
-u32 as_path_get_last_nonaggregated(struct adata *path);
+void as_path_cut(struct adata *path, uint num);
-int as_path_contains(struct adata *path, u32 as, int min);
+struct adata *as_path_merge(struct linpool *pool, struct adata *p1, struct adata *p2);
-int as_path_match_set(struct adata *path, struct f_tree *set);
+void as_path_format(const struct adata *path, byte *buf, uint size);
+int as_path_getlen(const struct adata *path);
+int as_path_getlen_int(const struct adata *path, int bs);
+int as_path_get_first(const struct adata *path, u32 *orig_as);
+int as_path_get_last(const struct adata *path, u32 *last_as);
+u32 as_path_get_last_nonaggregated(const struct adata *path);
+int as_path_contains(const struct adata *path, u32 as, int min);
+int as_path_match_set(const struct adata *path, struct f_tree *set);
 struct adata *as_path_filter(struct linpool *pool, struct adata *path, struct f_tree *set, u32 key, int pos);
 
+static inline struct adata *as_path_prepend(struct linpool *pool, const struct adata *path, u32 as)
+{ return as_path_prepend2(pool, path, AS_PATH_SEQUENCE, as, 0); }
+
 
 #define PM_ASN		0
 #define PM_QUESTION	1
@@ -54,7 +67,42 @@ struct f_path_mask {
   uintptr_t val2;
 };
 
-int as_path_match(struct adata *path, struct f_path_mask *mask);
+int as_path_match(const struct adata *path, struct f_path_mask *mask);
+
+
+/* Counterparts to appropriate as_path_* functions */
+
+static inline int
+aggregator_16to32(byte *dst, byte *src)
+{
+  put_u32(dst, get_u16(src));
+  memcpy(dst+4, src+2, 4);
+  return 8;
+}
+
+static inline int
+aggregator_32to16(byte *dst, byte *src)
+{
+  put_u16(dst, get_u32(src));
+  memcpy(dst+2, src+4, 4);
+  return 6;
+}
+
+static inline int
+aggregator_contains_as4(struct adata *a)
+{
+  return get_u32(a->data) > 0xFFFF;
+}
+
+static inline struct adata *
+aggregator_to_old(struct linpool *pool, struct adata *a)
+{
+  struct adata *d = lp_alloc_adata(pool, 8);
+  put_u32(d->data, 0xFFFF);
+  memcpy(d->data + 4, a->data + 4, 4);
+  return d;
+}
+
 
 /* a-set.c */
 
@@ -142,5 +190,9 @@ struct adata *int_set_union(struct linpool *pool, struct adata *l1, struct adata
 struct adata *ec_set_union(struct linpool *pool, struct adata *l1, struct adata *l2);
 struct adata *lc_set_union(struct linpool *pool, struct adata *l1, struct adata *l2);
 
+struct adata *ec_set_del_nontrans(struct linpool *pool, struct adata *set);
+struct adata *int_set_sort(struct linpool *pool, struct adata *src);
+struct adata *ec_set_sort(struct linpool *pool, struct adata *src);
+struct adata *lc_set_sort(struct linpool *pool, struct adata *src);
 
 #endif

nest/cmds.c

@@ -80,7 +80,6 @@ print_size(char *dsc, size_t val)
 
 extern pool *rt_table_pool;
 extern pool *rta_pool;
-extern pool *proto_pool;
 
 void
 cmd_show_memory(void)

nest/proto.c

@@ -39,7 +39,7 @@ static int graceful_restart_state;
 static u32 graceful_restart_locks;
 
 static char *p_states[] = { "DOWN", "START", "UP", "STOP" };
-static char *c_states[] UNUSED = { "DOWN", "START", "UP", "FLUSHING" };
+static char *c_states[] = { "DOWN", "START", "UP", "FLUSHING" };
 
 extern struct protocol proto_unix_iface;
 
@@ -304,6 +304,8 @@ channel_do_down(struct channel *c)
 
   memset(&c->stats, 0, sizeof(struct proto_stats));
 
+  CALL(c->channel->cleanup, c);
+
   /* Schedule protocol shutddown */
   if (proto_is_done(c->proto))
     ev_schedule(c->proto->event);
@@ -514,7 +516,9 @@ channel_reconfigure(struct channel *c, struct channel_config *cf)
 
   channel_verify_limits(c);
 
-  CALL(c->channel->reconfigure, c, cf);
+  /* Execute channel-specific reconfigure hook */
+  if (c->channel->reconfigure && !c->channel->reconfigure(c, cf))
+    return 0;
 
   /* If the channel is not open, it has no routes and we cannot reload it anyways */
   if (c->channel_state != CS_UP)
@@ -797,7 +801,6 @@ proto_reconfigure(struct proto *p, struct proto_config *oc, struct proto_config
   if ((nc->protocol != oc->protocol) ||
       (nc->net_type != oc->net_type) ||
       (nc->disabled != p->disabled))
-
     return 0;
 
   p->name = nc->name;
@@ -1575,6 +1578,7 @@ void
 channel_show_info(struct channel *c)
 {
   cli_msg(-1006, "  Channel %s", c->name);
+  cli_msg(-1006, "    State:          %s", c_states[c->channel_state]);
   cli_msg(-1006, "    Table:          %s", c->table->name);
   cli_msg(-1006, "    Preference:     %d", c->preference);
   cli_msg(-1006, "    Input filter:   %s", filter_name(c->in_filter));

nest/protocol.h

@@ -273,6 +273,7 @@ proto_get_router_id(struct proto_config *pc)
 /* Moved from route.h to avoid dependency conflicts */
 static inline void rte_update(struct proto *p, net_addr *n, rte *new) { rte_update2(p->main_channel, n, new, p->main_source); }
 
+extern pool *proto_pool;
 extern list proto_list;
 
 /*
@@ -418,20 +419,22 @@ struct channel_class {
   uint channel_size;			/* Size of channel data structure */
   uint config_size;			/* Size of channel config data structure */
 
-  struct channel * (*init)(struct channel *, struct channel_config *);	/* Create new instance */
+  void (*init)(struct channel *, struct channel_config *);	/* Create new instance */
   int (*reconfigure)(struct channel *, struct channel_config *);	/* Try to reconfigure instance, returns success */
   int (*start)(struct channel *);	/* Start the instance */
-  int (*shutdown)(struct channel *);	/* Stop the instance */
+  void (*shutdown)(struct channel *);	/* Stop the instance */
+  void (*cleanup)(struct channel *);	/* Channel finished flush */
 
   void (*copy_config)(struct channel_config *, struct channel_config *); /* Copy config from given channel instance */
 #if 0
+  XXXX;
   void (*preconfig)(struct protocol *, struct config *);	/* Just before configuring */
   void (*postconfig)(struct proto_config *);			/* After configuring each instance */
 
 
   void (*dump)(struct proto *);			/* Debugging dump */
   void (*dump_attrs)(struct rte *);		/* Dump protocol-dependent attributes */
-  void (*cleanup)(struct proto *);		/* Called after shutdown when protocol became hungry/down */
+
   void (*get_status)(struct proto *, byte *buf); /* Get instance status (for `show protocols' command) */
   void (*get_route_info)(struct rte *, byte *buf, struct ea_list *attrs); /* Get route information (for `show route' command) */
   int (*get_attr)(struct eattr *, byte *buf, int buflen);	/* ASCIIfy dynamic attribute (returns GA_*) */
@@ -440,6 +443,8 @@ struct channel_class {
 #endif
 };
 
+extern struct channel_class channel_bgp;
+
 struct channel_config {
   node n;
   const char *name;
@@ -484,6 +489,7 @@ struct channel {
   u8 merge_limit;			/* Maximal number of nexthops for RA_MERGED */
   u8 in_keep_filtered;			/* Routes rejected in import filter are kept */
   u8 disabled;
+  u8 stale;				/* Used in reconfiguration */
 
   u8 channel_state;
   u8 export_state;			/* Route export state (ES_*, see below) */

nest/route.h

@@ -457,13 +457,22 @@ typedef struct eattr {
 #define EAF_TYPE_UNDEF 0x1f		/* `force undefined' entry */
 #define EAF_EMBEDDED 0x01		/* Data stored in eattr.u.data (part of type spec) */
 #define EAF_VAR_LENGTH 0x02		/* Attribute length is variable (part of type spec) */
-#define EAF_ORIGINATED 0x40		/* The attribute has originated locally */
+#define EAF_ORIGINATED 0x20		/* The attribute has originated locally */
+#define EAF_FRESH 0x40			/* An uncached attribute (e.g. modified in export filter) */
 #define EAF_TEMP 0x80			/* A temporary attribute (the one stored in the tmp attr list) */
 
-struct adata {
+typedef struct adata {
   uint length;				/* Length of data */
   byte data[0];
-};
+} adata;
+
+static inline struct adata *
+lp_alloc_adata(struct linpool *pool, uint len)
+{
+  struct adata *ad = lp_alloc(pool, sizeof(struct adata) + len);
+  ad->length = len;
+  return ad;
+}
 
 static inline int adata_same(struct adata *a, struct adata *b)
 { return (a->length == b->length && !memcmp(a->data, b->data, a->length)); }
@@ -523,7 +532,7 @@ static inline rta * rta_cow(rta *r, linpool *lp) { return rta_is_cached(r) ? rta
 void rta_dump(rta *);
 void rta_dump_all(void);
 void rta_show(struct cli *, rta *, ea_list *);
-void rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr *gw, ip_addr *ll);
+void rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr gw, ip_addr ll);
 
 /*
  * rta_set_recursive_next_hop() acquires hostentry from hostcache and fills

nest/rt-attr.c

@@ -528,7 +528,7 @@ ea_do_prune(ea_list *e)
       if ((s0->type & EAF_TYPE_MASK) != EAF_TYPE_UNDEF)
 	{
 	  *d = *s0;
-	  d->type = (d->type & ~EAF_ORIGINATED) | (s[-1].type & EAF_ORIGINATED);
+	  d->type = (d->type & ~(EAF_ORIGINATED|EAF_FRESH)) | (s[-1].type & EAF_ORIGINATED);
 	  d++;
 	  i++;
 	}

nest/rt-table.c

@@ -2426,9 +2426,9 @@ rt_get_hostentry(rtable *tab, ip_addr a, ip_addr ll, rtable *dep)
 }
 
 void
-rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr *gw, ip_addr *ll)
+rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr gw, ip_addr ll)
 {
-  rta_apply_hostentry(a, rt_get_hostentry(tab, *gw, *ll, dep));
+  rta_apply_hostentry(a, rt_get_hostentry(tab, gw, ipa_zero(ll) ? gw : ll, dep));
 }
 
 

proto/bgp/bgp.h

@@ -2,6 +2,8 @@
  *	BIRD -- The Border Gateway Protocol
  *
  *	(c) 2000 Martin Mares <mj@ucw.cz>
+ *	(c) 2008--2016 Ondrej Zajicek <santiago@crfreenet.org>
+ *	(c) 2008--2016 CZ.NIC z.s.p.o.
  *
  *	Can be freely distributed and used under the terms of the GNU GPL.
  */
@@ -10,26 +12,66 @@
 #define _BIRD_BGP_H_
 
 #include <stdint.h>
+#include <setjmp.h>
+#include "nest/bird.h"
 #include "nest/route.h"
 #include "nest/bfd.h"
+//#include "lib/lists.h"
 #include "lib/hash.h"
+#include "lib/socket.h"
 
 struct linpool;
 struct eattr;
 
+
+/* Address families */
+
+#define BGP_AFI_IPV4		1
+#define BGP_AFI_IPV6		2
+
+#define BGP_SAFI_UNICAST	1
+#define BGP_SAFI_MULTICAST	2
+
+/* Internal AF codes */
+
+#define BGP_AF(A, B)		(((u32)(A) << 16) | (u32)(B))
+#define BGP_AFI(A)		((u32)(A) >> 16)
+#define BGP_SAFI(A)		((u32)(A) & 0xFFFF)
+
+#define BGP_AF_IPV4		BGP_AF( BGP_AFI_IPV4, BGP_SAFI_UNICAST )
+#define BGP_AF_IPV6		BGP_AF( BGP_AFI_IPV6, BGP_SAFI_UNICAST )
+#define BGP_AF_IPV4_MC		BGP_AF( BGP_AFI_IPV4, BGP_SAFI_MULTICAST )
+#define BGP_AF_IPV6_MC		BGP_AF( BGP_AFI_IPV6, BGP_SAFI_MULTICAST )
+
+
+struct bgp_write_state;
+struct bgp_parse_state;
+struct bgp_export_state;
+struct bgp_bucket;
+
+struct bgp_af_desc {
+  u32 afi;
+  u32 net;
+  const char *name;
+  uint (*encode_nlri)(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, uint size);
+  void (*decode_nlri)(struct bgp_parse_state *s, byte *pos, uint len, rta *a);
+  void (*update_next_hop)(struct bgp_export_state *s, eattr *nh, ea_list **to);
+  uint (*encode_next_hop)(struct bgp_write_state *s, eattr *nh, byte *buf, uint size);
+  void (*decode_next_hop)(struct bgp_parse_state *s, byte *pos, uint len, rta *a);
+};
+
+
 struct bgp_config {
   struct proto_config c;
   u32 local_as, remote_as;
+  ip_addr local_ip;			/* Source address to use */
   ip_addr remote_ip;
-  ip_addr source_addr;			/* Source address to use */
   struct iface *iface;			/* Interface for link-local addresses */
+  u16 local_port;			/* Local listening port */
   u16 remote_port; 			/* Neighbor destination port */
   int multihop;				/* Number of hops if multihop */
+  int strict_bind;			/* Bind listening socket to local address XXXX */
   int ttl_security;			/* Enable TTL security [RFC 5082] */
-  int next_hop_self;			/* Always set next hop to local IP address */
-  int next_hop_keep;			/* Do not touch next hop attribute */
-  int missing_lladdr;			/* What we will do when we don' know link-local addr, see MLL_* */
-  int gw_mode;				/* How we compute route gateway from next_hop attr, see GW_* */
   int compare_path_lengths;		/* Use path lengths when selecting best route */
   int med_metric;			/* Compare MULTI_EXIT_DISC even between routes from differen ASes */
   int igp_metric;			/* Use IGP metrics when selecting best route */
@@ -44,11 +86,10 @@ struct bgp_config {
   u32 rr_cluster_id;			/* Route reflector cluster ID, if different from local ID */
   int rr_client;			/* Whether neighbor is RR client of me */
   int rs_client;			/* Whether neighbor is RS client of me */
-  int advertise_ipv4;			/* Whether we should add IPv4 capability advertisement to OPEN message */
+  u32 confederation;			/* Confederation ID, or zero if confeds not active */
+  int confederation_member;		/* Whether neighbor AS is member of our confederation */
   int passive;				/* Do not initiate outgoing connection */
   int interpret_communities;		/* Hardwired handling of well-known communities */
-  int secondary;			/* Accept also non-best routes (i.e. RA_ACCEPTED) */
-  int add_path;				/* Use ADD-PATH extension [draft] */
   int allow_local_as;			/* Allow that number of local ASNs in incoming AS_PATHs */
   int gr_mode;				/* Graceful restart mode (BGP_GR_*) */
   int setkey;				/* Set MD5 password to system SA/SP database */
@@ -63,11 +104,27 @@ struct bgp_config {
   unsigned disable_after_error;		/* Disable the protocol when error is detected */
 
   char *password;			/* Password used for MD5 authentication */
-  struct rtable_config *igp_table;	/* Table used for recursive next hop lookups */
   int check_link;			/* Use iface link state for liveness detection */
   int bfd;				/* Use BFD for liveness detection */
 };
 
+struct bgp_channel_config {
+  struct channel_config c;
+
+  u32 afi;
+
+  ip_addr next_hop_addr;		/* Local address for NEXT_HOP attribute */
+  u8 next_hop_self;			/* Always set next hop to local IP address */
+  u8 next_hop_keep;			/* Do not touch next hop attribute */
+  u8 missing_lladdr;			/* What we will do when we don' know link-local addr, see MLL_* */
+  u8 gw_mode;				/* How we compute route gateway from next_hop attr, see GW_* */
+  u8 secondary;				/* Accept also non-best routes (i.e. RA_ACCEPTED) */
+  u8 gr_able;				/* Allow full graceful restart for the channel */
+  u8 add_path;				/* Use ADD-PATH extension [RFC 7911] */
+
+  struct rtable_config *igp_table;	/* Table used for recursive next hop lookups */
+};
+
 #define MLL_SELF 1
 #define MLL_DROP 2
 #define MLL_IGNORE 3
@@ -75,112 +132,226 @@ struct bgp_config {
 #define GW_DIRECT 1
 #define GW_RECURSIVE 2
 
-#define ADD_PATH_RX 1
+#define BGP_ADD_PATH_RX		1
-#define ADD_PATH_TX 2
+#define BGP_ADD_PATH_TX		2
-#define ADD_PATH_FULL 3
+#define BGP_ADD_PATH_FULL	3
 
 #define BGP_GR_ABLE		1
 #define BGP_GR_AWARE		2
 
-/* For peer_gr_flags */
+/* For GR capability common flags */
 #define BGP_GRF_RESTART 0x80
 
-/* For peer_gr_aflags */
+/* For GR capability per-AF flags */
 #define BGP_GRF_FORWARDING 0x80
 
 
+struct bgp_af_caps {
+  u32 afi;
+  u8 ready;				/* Multiprotocol capability, RFC 4760 */
+  u8 gr_able;				/* Graceful restart support, RFC 4724 */
+  u8 gr_af_flags;			/* Graceful restart per-AF flags */
+  u8 add_path;				/* Multiple paths support,   RFC 7911 */
+};
+
+struct bgp_caps {
+  u32 as4_number;			/* Announced ASN */
+
+  u8 as4_support;			/* Four-octet AS capability, RFC 6793 */
+  u8 ext_messages;			/* Extended message length,  RFC draft */
+  u8 route_refresh;			/* Route refresh capability, RFC 2918 */
+  u8 enhanced_refresh;			/* Enhanced route refresh,   RFC 7313 */
+
+  u8 gr_aware;				/* Graceful restart capability, RFC 4724 */
+  u8 gr_flags;				/* Graceful restart flags */
+  u16 gr_time;				/* Graceful restart time in seconds */
+
+  u16 af_count;				/* Number of af_data items */
+
+  struct bgp_af_caps af_data[0];	/* Per-AF capability data */
+};
+
+struct bgp_socket {
+  node n;				/* Node in global bgp_sockets */
+  sock *sk;				/* Real listening socket */
+  u32 uc;				/* Use count */
+};
+
 struct bgp_conn {
   struct bgp_proto *bgp;
   struct birdsock *sk;
-  uint state;				/* State of connection state machine */
+  u8 state;				/* State of connection state machine */
-  struct timer *connect_retry_timer;
+  u8 as4_session;			/* Session uses 4B AS numbers in AS_PATH (both sides support it) */
+  u8 ext_messages;			/* Session uses extended message length */
+
+  struct bgp_caps *local_caps;
+  struct bgp_caps *remote_caps;
+  struct timer *connect_timer;
   struct timer *hold_timer;
   struct timer *keepalive_timer;
   struct event *tx_ev;
-  int packets_to_send;			/* Bitmap of packet types to be sent */
+  u32 packets_to_send;			/* Bitmap of packet types to be sent */
+  u32 channels_to_send;			/* Bitmap of channels with packets to be sent */
+  u8 last_channel;			/* Channel used last time for TX */
+  u8 last_channel_count;		/* Number of times the last channel was used in succession */
   int notify_code, notify_subcode, notify_size;
   byte *notify_data;
-  u32 advertised_as;			/* Temporary value for AS number received */
+
-  int start_state;			/* protocol start_state snapshot when connection established */
+  uint hold_time, keepalive_time;	/* Times calculated from my and neighbor's requirements */
-  u8 peer_refresh_support;		/* Peer supports route refresh [RFC2918] */
-  u8 peer_as4_support;			/* Peer supports 4B AS numbers [RFC4893] */
-  u8 peer_add_path;			/* Peer supports ADD-PATH [draft] */
-  u8 peer_enhanced_refresh_support;	/* Peer supports enhanced refresh [RFC7313] */
-  u8 peer_gr_aware;
-  u8 peer_gr_able;
-  u16 peer_gr_time;
-  u8 peer_gr_flags;
-  u8 peer_gr_aflags;
-  u8 peer_ext_messages_support;		/* Peer supports extended message length [draft] */
-  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 */
   u32 local_as, remote_as;
-  int start_state;			/* Substates that partitions BS_START */
+  u32 public_as;			/* Externally visible ASN (local_as or confederation id) */
-  u8 is_internal;			/* Internal BGP connection (local_as == remote_as) */
-  u8 as4_session;			/* Session uses 4B AS numbers in AS_PATH (both sides support it) */
-  u8 add_path_rx;			/* Session expects receive of ADD-PATH extended NLRI */
-  u8 add_path_tx;			/* Session expects transmit of ADD-PATH extended NLRI */
-  u8 ext_messages;			/* Session allows to use extended messages (both sides support it) */
   u32 local_id;				/* BGP identifier of this router */
   u32 remote_id;			/* BGP identifier of the neighbor */
   u32 rr_cluster_id;			/* Route reflector cluster ID */
-  int rr_client;			/* Whether neighbor is RR client of me */
+  int start_state;			/* Substates that partitions BS_START */
-  int rs_client;			/* Whether neighbor is RS client of me */
+  u8 is_internal;			/* Internal BGP session (local_as == remote_as) */
+  u8 is_interior;			/* Internal or intra-confederation BGP session */
+  u8 as4_session;			/* Session uses 4B AS numbers in AS_PATH (both sides support it) */
+  u8 rr_client;				/* Whether neighbor is RR client of me */
+  u8 rs_client;				/* Whether neighbor is RS client of me */
+  u8 route_refresh;			/* Route refresh allowed to send [RFC 2918] */
+  u8 enhanced_refresh;			/* Enhanced refresh is negotiated [RFC 7313] */
   u8 gr_ready;				/* Neighbor could do graceful restart */
-  u8 gr_active;				/* Neighbor is doing graceful restart */
+  u8 gr_active_num;			/* Neighbor is doing GR, number of active channels */
-  u8 feed_state;			/* Feed state (TX) for EoR, RR packets, see BFS_* */
+  u8 channel_count;			/* Number of active channels */
-  u8 load_state;			/* Load state (RX) for EoR, RR packets, see BFS_* */
+  u32 *afi_map;				/* Map channel index -> AFI */
+  struct bgp_channel **channel_map;	/* Map channel index -> channel */
   struct bgp_conn *conn;		/* Connection we have established */
   struct bgp_conn outgoing_conn;	/* Outgoing connection we're working with */
   struct bgp_conn incoming_conn;	/* Incoming connection we have neither accepted nor rejected yet */
   struct object_lock *lock;		/* Lock for neighbor connection */
   struct neighbor *neigh;		/* Neighbor entry corresponding to remote ip, NULL if multihop */
+  struct bgp_socket *sock;		/* Shared listening socket */
   struct bfd_request *bfd_req;		/* BFD request, if BFD is used */
   ip_addr source_addr;			/* Local address used as an advertised next hop */
-  rtable *igp_table;			/* Table used for recursive next hop lookups */
+  ip_addr link_addr;			/* Link-local version of source_addr */
   struct event *event;			/* Event for respawning and shutting process */
   struct timer *startup_timer;		/* Timer used to delay protocol startup due to previous errors (startup_delay) */
   struct timer *gr_timer;		/* Timer waiting for reestablishment after graceful restart */
-  struct bgp_bucket **bucket_hash;	/* Hash table of attribute buckets */
-  uint hash_size, hash_count, hash_limit;
-  HASH(struct bgp_prefix) prefix_hash;	/* Prefixes to be sent */
-  slab *prefix_slab;			/* Slab holding prefix nodes */
-  list bucket_queue;			/* Queue of buckets to send */
-  struct bgp_bucket *withdraw_bucket;	/* Withdrawn routes */
   unsigned startup_delay;		/* Time to delay protocol startup by due to errors */
   bird_clock_t last_proto_error;	/* Time of last error that leads to protocol stop */
   u8 last_error_class; 			/* Error class of last error */
   u32 last_error_code;			/* Error code of last error. BGP protocol errors
 					   are encoded as (bgp_err_code << 16 | bgp_err_subcode) */
-#ifdef IPV6
+};
-  byte *mp_reach_start, *mp_unreach_start; /* Multiprotocol BGP attribute notes */
+
-  unsigned mp_reach_len, mp_unreach_len;
+struct bgp_channel {
-  ip_addr local_link;			/* Link-level version of source_addr */
+  struct channel c;
-#endif
+
+  /* Rest are BGP specific data */
+  struct bgp_channel_config *cf;
+  pool *pool; /* XXXX */
+
+  u32 afi;
+  u32 index;
+  const struct bgp_af_desc *desc;
+
+  HASH(struct bgp_bucket) bucket_hash;	/* Hash table of route buckets */
+  struct bgp_bucket *withdraw_bucket;	/* Withdrawn routes */
+  list bucket_queue;			/* Queue of buckets to send (struct bgp_bucket) */
+
+  HASH(struct bgp_prefix) prefix_hash;	/* Prefixes to be sent */
+  slab *prefix_slab;			/* Slab holding prefix nodes */
+
+  rtable *igp_table;			/* Table used for recursive next hop lookups */
+  ip_addr next_hop_addr;		/* Local address for NEXT_HOP attribute */
+  ip_addr link_addr;			/* Link-local version of next_hop_addr */
+
+  u32 packets_to_send;			/* Bitmap of packet types to be sent */
+
+  u8 gr_ready;				/* Neighbor could do GR on this AF */
+  u8 gr_active;				/* Neighbor is doing GR and keeping fwd state */
+
+  u8 add_path_rx;			/* Session expects receive of ADD-PATH extended NLRI */
+  u8 add_path_tx;			/* Session expects transmit of ADD-PATH extended NLRI */
+
+  u8 feed_state;			/* Feed state (TX) for EoR, RR packets, see BFS_* */
+  u8 load_state;			/* Load state (RX) for EoR, RR packets, see BFS_* */
 };
 
 struct bgp_prefix {
-  struct {
+  node buck_node;			/* Node in per-bucket list */
-    ip_addr prefix;
+  struct bgp_prefix *next;		/* Node in prefix hash table */
-    int pxlen;
+  u32 hash;
-  } n;
   u32 path_id;
-  struct bgp_prefix *next;
+  net_addr net[0];
-  node bucket_node;			/* Node in per-bucket list */
 };
 
 struct bgp_bucket {
   node send_node;			/* Node in send queue */
-  struct bgp_bucket *hash_next, *hash_prev;	/* Node in bucket hash table */
+  struct bgp_bucket *next;		/* Node in bucket hash table */
-  unsigned hash;			/* Hash over extended attributes */
+  list prefixes;			/* Prefixes in this bucket (struct bgp_prefix) */
-  list prefixes;			/* Prefixes in this buckets */
+  u32 hash;				/* Hash over extended attributes */
   ea_list eattrs[0];			/* Per-bucket extended attributes */
 };
 
+struct bgp_export_state {
+  struct bgp_proto *proto;
+  struct bgp_channel *channel;
+  struct linpool *pool;
+
+  struct bgp_proto *src;
+  rte *route;
+
+  u32 attrs_seen[1];
+  uint err_withdraw;
+};
+
+struct bgp_write_state {
+  struct bgp_proto *proto;
+  struct bgp_channel *channel;
+  struct linpool *pool;
+
+  int as4_session;
+  int add_path;
+
+  eattr *mp_next_hop;
+};
+
+struct bgp_parse_state {
+  struct bgp_proto *proto;
+  struct bgp_channel *channel;
+  struct linpool *pool;
+
+  int as4_session;
+  int add_path;
+
+  u32 attrs_seen[256/32];
+
+  u32 mp_reach_af;
+  u32 mp_unreach_af;
+
+  uint attr_len;
+  uint ip_reach_len;
+  uint ip_unreach_len;
+  uint ip_next_hop_len;
+  uint mp_reach_len;
+  uint mp_unreach_len;
+  uint mp_next_hop_len;
+
+  byte *attrs;
+  byte *ip_reach_nlri;
+  byte *ip_unreach_nlri;
+  byte *ip_next_hop_data;
+  byte *mp_reach_nlri;
+  byte *mp_unreach_nlri;
+  byte *mp_next_hop_data;
+
+  uint err_withdraw;
+  uint err_subcode;
+  jmp_buf err_jmpbuf;
+
+  /* Cached state for bgp_rte_update() */
+  u32 last_id;
+  struct rte_src *last_src;
+  rta *cached_rta;
+};
+
 #define BGP_PORT		179
 #define BGP_VERSION		4
 #define BGP_HEADER_LENGTH	19
@@ -191,8 +362,21 @@ struct bgp_bucket {
 #define BGP_RX_BUFFER_EXT_SIZE	65535
 #define BGP_TX_BUFFER_EXT_SIZE	65535
 
-static inline uint bgp_max_packet_length(struct bgp_proto *p)
+static inline int bgp_channel_is_ipv4(struct bgp_channel *c)
-{ return p->ext_messages ? BGP_MAX_EXT_MSG_LENGTH : BGP_MAX_MESSAGE_LENGTH; }
+{ return BGP_AFI(c->afi) == BGP_AFI_IPV4; }
+
+static inline int bgp_channel_is_ipv6(struct bgp_channel *c)
+{ return BGP_AFI(c->afi) == BGP_AFI_IPV6; }
+
+static inline uint bgp_max_packet_length(struct bgp_conn *conn)
+{ return conn->ext_messages ? BGP_MAX_EXT_MSG_LENGTH : BGP_MAX_MESSAGE_LENGTH; }
+
+static inline void
+bgp_parse_error(struct bgp_parse_state *s, uint subcode)
+{
+  s->err_subcode = subcode;
+  longjmp(s->err_jmpbuf, 1);
+}
 
 extern struct linpool *bgp_linpool;
 
@@ -207,9 +391,9 @@ void bgp_conn_enter_established_state(struct bgp_conn *conn);
 void bgp_conn_enter_close_state(struct bgp_conn *conn);
 void bgp_conn_enter_idle_state(struct bgp_conn *conn);
 void bgp_handle_graceful_restart(struct bgp_proto *p);
-void bgp_graceful_restart_done(struct bgp_proto *p);
+void bgp_graceful_restart_done(struct bgp_channel *c);
-void bgp_refresh_begin(struct bgp_proto *p);
+void bgp_refresh_begin(struct bgp_channel *c);
-void bgp_refresh_end(struct bgp_proto *p);
+void bgp_refresh_end(struct bgp_channel *c);
 void bgp_store_error(struct bgp_proto *p, struct bgp_conn *c, u8 class, u32 code);
 void bgp_stop(struct bgp_proto *p, unsigned subcode);
 
@@ -232,46 +416,83 @@ struct rte_source *bgp_get_source(struct bgp_proto *p, u32 path_id);
 
 /* attrs.c */
 
+static inline eattr *
+bgp_find_attr(ea_list *attrs, uint code)
+{
+  return ea_find(attrs, EA_CODE(EAP_BGP, code));
+}
+
+eattr *
+bgp_set_attr(ea_list **attrs, struct linpool *pool, uint code, uint flags, uintptr_t val);
+
+static inline void
+bgp_set_attr_u32(ea_list **to, struct linpool *pool, uint code, uint flags, u32 val)
+{ bgp_set_attr(to, pool, code, flags, (uintptr_t) val); }
+
+static inline void
+bgp_set_attr_ptr(ea_list **to, struct linpool *pool, uint code, uint flags, struct adata *val)
+{ bgp_set_attr(to, pool, code, flags, (uintptr_t) val); }
+
+static inline void
+bgp_set_attr_data(ea_list **to, struct linpool *pool, uint code, uint flags, void *data, uint len)
+{
+  struct adata *a = lp_alloc_adata(pool, len);
+  memcpy(a->data, data, len);
+  bgp_set_attr(to, pool, code, flags, (uintptr_t) a);
+}
+
+static inline void
+bgp_unset_attr(ea_list **to, struct linpool *pool, uint code)
+{ eattr *e = bgp_set_attr(to, pool, code, 0, 0); e->type = EAF_TYPE_UNDEF; }
+
+
+
+
 /* Hack: although BA_NEXT_HOP attribute has type EAF_TYPE_IP_ADDRESS, in IPv6
  * we store two addesses in it - a global address and a link local address.
  */
-#ifdef IPV6
+#ifdef XXX
 #define NEXT_HOP_LENGTH (2*sizeof(ip_addr))
 static inline void set_next_hop(byte *b, ip_addr addr) { ((ip_addr *) b)[0] = addr; ((ip_addr *) b)[1] = IPA_NONE; }
-#else
 #define NEXT_HOP_LENGTH sizeof(ip_addr)
 static inline void set_next_hop(byte *b, ip_addr addr) { ((ip_addr *) b)[0] = addr; }
 #endif
 
-void bgp_attach_attr(struct ea_list **to, struct linpool *pool, unsigned attr, uintptr_t val);
+int bgp_encode_attrs(struct bgp_write_state *s, ea_list *attrs, byte *buf, byte *end);
-byte *bgp_attach_attr_wa(struct ea_list **to, struct linpool *pool, unsigned attr, unsigned len);
+ea_list * bgp_decode_attrs(struct bgp_parse_state *s, byte *data, uint len);
-struct rta *bgp_decode_attrs(struct bgp_conn *conn, byte *a, uint len, struct linpool *pool, int mandatory);
+
-int bgp_get_attr(struct eattr *e, byte *buf, int buflen);
+void bgp_init_bucket_table(struct bgp_channel *c);
+void bgp_free_bucket(struct bgp_channel *c, struct bgp_bucket *b);
+void bgp_defer_bucket(struct bgp_channel *c, struct bgp_bucket *b);
+void bgp_withdraw_bucket(struct bgp_channel *c, struct bgp_bucket *b);
+
+void bgp_init_prefix_table(struct bgp_channel *c);
+void bgp_free_prefix(struct bgp_channel *c, struct bgp_prefix *bp);
+
 int bgp_rte_better(struct rte *, struct rte *);
 int bgp_rte_mergable(rte *pri, rte *sec);
 int bgp_rte_recalculate(rtable *table, net *net, rte *new, rte *old, rte *old_best);
-void bgp_rt_notify(struct proto *P, rtable *tbl UNUSED, net *n, rte *new, rte *old UNUSED, ea_list *attrs);
+void bgp_rt_notify(struct proto *P, struct channel *C, net *n, rte *new, rte *old, ea_list *attrs);
 int bgp_import_control(struct proto *, struct rte **, struct ea_list **, struct linpool *);
-void bgp_init_bucket_table(struct bgp_proto *);
+int bgp_get_attr(struct eattr *e, byte *buf, int buflen);
-void bgp_free_bucket(struct bgp_proto *p, struct bgp_bucket *buck);
-void bgp_init_prefix_table(struct bgp_proto *p, u32 order);
-void bgp_free_prefix(struct bgp_proto *p, struct bgp_prefix *bp);
-uint bgp_encode_attrs(struct bgp_proto *p, byte *w, ea_list *attrs, int remains);
 void bgp_get_route_info(struct rte *, byte *buf, struct ea_list *attrs);
 
-inline static void bgp_attach_attr_ip(struct ea_list **to, struct linpool *pool, unsigned attr, ip_addr a)
-{ *(ip_addr *) bgp_attach_attr_wa(to, pool, attr, sizeof(ip_addr)) = a; }
 
 /* packets.c */
 
 void mrt_dump_bgp_state_change(struct bgp_conn *conn, unsigned old, unsigned new);
-void bgp_schedule_packet(struct bgp_conn *conn, int type);
+const struct bgp_af_desc *bgp_get_af_desc(u32 afi);
+const struct bgp_af_caps *bgp_find_af_caps(struct bgp_caps *caps, u32 afi);
+void bgp_schedule_packet(struct bgp_conn *conn, struct bgp_channel *c, int type);
 void bgp_kick_tx(void *vconn);
 void bgp_tx(struct birdsock *sk);
 int bgp_rx(struct birdsock *sk, uint size);
 const char * bgp_error_dsc(unsigned code, unsigned subcode);
 void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsigned subcode, byte *data, unsigned len);
 
+void bgp_update_next_hop(struct bgp_export_state *s, eattr *a, ea_list **to);
+
+
 /* Packet types */
 
 #define PKT_OPEN		0x01
@@ -305,9 +526,9 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
 #define BA_RCID_PATH		0x0d
 #define BA_MP_REACH_NLRI	0x0e	/* [RFC2283] */
 #define BA_MP_UNREACH_NLRI	0x0f
-#define BA_EXT_COMMUNITY	0x10	/* [RFC4360] */
+#define BA_EXT_COMMUNITY	0x10	/* RFC 4360 */
-#define BA_AS4_PATH             0x11    /* [RFC4893] */
+#define BA_AS4_PATH             0x11	/* RFC 6793 */
-#define BA_AS4_AGGREGATOR       0x12
+#define BA_AS4_AGGREGATOR       0x12	/* RFC 6793 */
 #define BA_LARGE_COMMUNITY	0x20	/* [draft-ietf-idr-large-community] */
 
 /* BGP connection states */
@@ -328,14 +549,12 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
  *
  * When BGP protocol is started by core, it goes to BSS_PREPARE. When BGP
  * protocol done what is neccessary to start itself (like acquiring the lock),
- * it goes to BSS_CONNECT.  When some connection attempt failed because of
+ * it goes to BSS_CONNECT.
- * option or capability error, it goes to BSS_CONNECT_NOCAP.
  */
 
 #define BSS_PREPARE		0	/* Used before ordinary BGP started, i. e. waiting for lock */
 #define BSS_DELAY		1	/* Startup delay due to previous errors */
 #define BSS_CONNECT		2	/* Ordinary BGP connecting */
-#define BSS_CONNECT_NOCAP	3	/* Legacy BGP connecting (without capabilities) */
 
 
 /* BGP feed states (TX)
@@ -344,7 +563,7 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
  *
  * RFC 7313 specifies that a route refresh should be demarcated by BoRR and EoRR packets.
  *
- * These states (stored in p->feed_state) are used to keep track of these
+ * These states (stored in c->feed_state) are used to keep track of these
  * requirements. When such feed is started, BFS_LOADING / BFS_REFRESHING is
  * set. When it ended, BFS_LOADED / BFS_REFRESHED is set to schedule End-of-RIB
  * or EoRR packet. When the packet is sent, the state returned to BFS_NONE.
@@ -400,15 +619,5 @@ void bgp_log_error(struct bgp_proto *p, u8 class, char *msg, unsigned code, unsi
 #define ORIGIN_EGP		1
 #define ORIGIN_INCOMPLETE	2
 
-/* Address families */
-
-#define BGP_AF_IPV4		1
-#define BGP_AF_IPV6		2
-
-#ifdef IPV6
-#define BGP_AF BGP_AF_IPV6
-#else
-#define BGP_AF BGP_AF_IPV4
-#endif
 
 #endif

proto/bgp/config.Y

@@ -13,28 +13,32 @@ CF_HDR
 CF_DEFINES
 
 #define BGP_CFG ((struct bgp_config *) this_proto)
+#define BGP_CC ((struct bgp_channel_config *) this_channel)
 
 CF_DECLS
 
-CF_KEYWORDS(BGP, LOCAL, NEIGHBOR, AS, HOLD, TIME, CONNECT, RETRY,
+CF_KEYWORDS(BGP, LOCAL, NEIGHBOR, AS, HOLD, TIME, CONNECT, RETRY, KEEPALIVE,
-	KEEPALIVE, MULTIHOP, STARTUP, VIA, NEXT, HOP, SELF, DEFAULT,
+	MULTIHOP, STARTUP, VIA, NEXT, HOP, SELF, DEFAULT, PATH, METRIC, ERROR,
-	PATH, METRIC, ERROR, START, DELAY, FORGET, WAIT, ENABLE,
+	START, DELAY, FORGET, WAIT, ENABLE, DISABLE, AFTER, BGP_PATH,
-	DISABLE, AFTER, BGP_PATH, BGP_LOCAL_PREF, BGP_MED, BGP_ORIGIN,
+	BGP_LOCAL_PREF, BGP_MED, BGP_ORIGIN, BGP_NEXT_HOP, BGP_ATOMIC_AGGR,
-	BGP_NEXT_HOP, BGP_ATOMIC_AGGR, BGP_AGGREGATOR, BGP_COMMUNITY,
+	BGP_AGGREGATOR, BGP_COMMUNITY, BGP_EXT_COMMUNITY, BGP_LARGE_COMMUNITY,
-	BGP_EXT_COMMUNITY, SOURCE, ADDRESS, PASSWORD, RR, RS, CLIENT,
+	SOURCE, ADDRESS, PASSWORD, RR, RS, CLIENT, CLUSTER, ID, AS4, ADVERTISE,
-	CLUSTER, ID, AS4, ADVERTISE, IPV4, CAPABILITIES, LIMIT, PASSIVE,
+	IPV4, CAPABILITIES, LIMIT, PASSIVE, PREFER, OLDER, MISSING, LLADDR,
-	PREFER, OLDER, MISSING, LLADDR, DROP, IGNORE, ROUTE, REFRESH,
+	DROP, IGNORE, ROUTE, REFRESH, INTERPRET, COMMUNITIES, BGP_ORIGINATOR_ID,
-	INTERPRET, COMMUNITIES, BGP_ORIGINATOR_ID, BGP_CLUSTER_LIST, IGP,
+	BGP_CLUSTER_LIST, IGP, TABLE, GATEWAY, DIRECT, RECURSIVE, MED, TTL,
-	TABLE, GATEWAY, DIRECT, RECURSIVE, MED, TTL, SECURITY, DETERMINISTIC,
+	SECURITY, DETERMINISTIC, SECONDARY, ALLOW, BFD, ADD, PATHS, RX, TX,
-	SECONDARY, ALLOW, BFD, ADD, PATHS, RX, TX, GRACEFUL, RESTART, AWARE,
+	GRACEFUL, RESTART, AWARE, CHECK, LINK, PORT, EXTENDED, MESSAGES, SETKEY,
-	CHECK, LINK, PORT, EXTENDED, MESSAGES, SETKEY, BGP_LARGE_COMMUNITY)
+	STRICT, BIND, CONFEDERATION, MEMBER, MULTICAST)
+
+%type <i32> bgp_afi
 
 CF_GRAMMAR
 
-CF_ADDTO(proto, bgp_proto '}' { bgp_check_config(BGP_CFG); } )
+CF_ADDTO(proto, bgp_proto '}' )
 
 bgp_proto_start: proto_start BGP {
      this_proto = proto_config_new(&proto_bgp, $1);
+     BGP_CFG->local_port = BGP_PORT;
      BGP_CFG->remote_port = BGP_PORT;
      BGP_CFG->multihop = -1;	/* undefined */
      BGP_CFG->hold_time = 240;
@@ -49,7 +53,6 @@ bgp_proto_start: proto_start BGP {
      BGP_CFG->enable_refresh = 1;
      BGP_CFG->enable_as4 = 1;
      BGP_CFG->capabilities = 2;
-     BGP_CFG->advertise_ipv4 = 1;
      BGP_CFG->interpret_communities = 1;
      BGP_CFG->default_local_pref = 100;
      BGP_CFG->gr_mode = BGP_GR_AWARE;
@@ -58,6 +61,12 @@ bgp_proto_start: proto_start BGP {
    }
  ;
 
+bgp_loc_opts:
+   /* empty */
+ | bgp_loc_opts PORT expr { BGP_CFG->local_port = $3; if (($3<1) || ($3>65535)) cf_error("Invalid port number"); }
+ | bgp_loc_opts AS expr { BGP_CFG->local_as = $3; }
+ ;
+
 bgp_nbr_opts:
    /* empty */
  | bgp_nbr_opts PORT expr { BGP_CFG->remote_port = $3; if (($3<1) || ($3>65535)) cf_error("Invalid port number"); }
@@ -67,8 +76,12 @@ bgp_nbr_opts:
 bgp_proto:
    bgp_proto_start proto_name '{'
  | bgp_proto proto_item ';'
- | bgp_proto LOCAL AS expr ';' { BGP_CFG->local_as = $4; }
+ | bgp_proto bgp_proto_channel ';'
- | bgp_proto LOCAL ipa AS expr ';' { BGP_CFG->source_addr = $3; BGP_CFG->local_as = $5; }
+ | bgp_proto LOCAL bgp_loc_opts ';'
+ | bgp_proto LOCAL ipa ipa_scope bgp_loc_opts ';' {
+     BGP_CFG->local_ip = $3;
+     if ($4) BGP_CFG->iface = $4;
+   }
  | bgp_proto NEIGHBOR bgp_nbr_opts ';'
  | bgp_proto NEIGHBOR ipa ipa_scope bgp_nbr_opts ';' {
      if (ipa_nonzero(BGP_CFG->remote_ip))
@@ -78,20 +91,16 @@ bgp_proto:
    }
  | bgp_proto INTERFACE TEXT ';' { BGP_CFG->iface = if_get_by_name($3); }
  | bgp_proto RR CLUSTER ID idval ';' { BGP_CFG->rr_cluster_id = $5; }
- | bgp_proto RR CLIENT ';' { BGP_CFG->rr_client = 1; }
+ | bgp_proto RR CLIENT bool ';' { BGP_CFG->rr_client = $4; }
- | bgp_proto RS CLIENT ';' { BGP_CFG->rs_client = 1; }
+ | bgp_proto RS CLIENT bool ';' { BGP_CFG->rs_client = $4; }
+ | bgp_proto CONFEDERATION expr ';' { BGP_CFG->confederation = $3; }
+ | bgp_proto CONFEDERATION MEMBER bool ';' { BGP_CFG->confederation_member = $4; }
  | bgp_proto HOLD TIME expr ';' { BGP_CFG->hold_time = $4; }
  | bgp_proto STARTUP HOLD TIME expr ';' { BGP_CFG->initial_hold_time = $5; }
  | bgp_proto DIRECT ';' { BGP_CFG->multihop = 0; }
  | bgp_proto MULTIHOP ';' { BGP_CFG->multihop = 64; }
  | bgp_proto MULTIHOP expr ';' { BGP_CFG->multihop = $3; if (($3<1) || ($3>255)) cf_error("Multihop must be in range 1-255"); }
- | bgp_proto NEXT HOP SELF ';' { BGP_CFG->next_hop_self = 1; BGP_CFG->next_hop_keep = 0; }
+ | bgp_proto STRICT BIND bool ';' { BGP_CFG->strict_bind = $4; }
- | bgp_proto NEXT HOP KEEP ';' { BGP_CFG->next_hop_keep = 1; BGP_CFG->next_hop_self = 0; }
- | bgp_proto MISSING LLADDR SELF ';' { BGP_CFG->missing_lladdr = MLL_SELF; }
- | bgp_proto MISSING LLADDR DROP ';' { BGP_CFG->missing_lladdr = MLL_DROP; }
- | bgp_proto MISSING LLADDR IGNORE ';' { BGP_CFG->missing_lladdr = MLL_IGNORE; }
- | bgp_proto GATEWAY DIRECT ';' { BGP_CFG->gw_mode = GW_DIRECT; }
- | bgp_proto GATEWAY RECURSIVE ';' { BGP_CFG->gw_mode = GW_RECURSIVE; }
  | bgp_proto PATH METRIC bool ';' { BGP_CFG->compare_path_lengths = $4; }
  | bgp_proto MED METRIC bool ';' { BGP_CFG->med_metric = $4; }
  | bgp_proto IGP METRIC bool ';' { BGP_CFG->igp_metric = $4; }
@@ -99,7 +108,7 @@ bgp_proto:
  | bgp_proto DETERMINISTIC MED bool ';' { BGP_CFG->deterministic_med = $4; }
  | bgp_proto DEFAULT BGP_MED expr ';' { BGP_CFG->default_med = $4; }
  | bgp_proto DEFAULT BGP_LOCAL_PREF expr ';' { BGP_CFG->default_local_pref = $4; }
- | bgp_proto SOURCE ADDRESS ipa ';' { BGP_CFG->source_addr = $4; }
+ | bgp_proto SOURCE ADDRESS ipa ';' { BGP_CFG->local_ip = $4; }
  | bgp_proto START DELAY TIME expr ';' { BGP_CFG->connect_delay_time = $5; log(L_WARN "%s: Start delay time option is deprecated, use connect delay time", this_proto->name); }
  | bgp_proto CONNECT DELAY TIME expr ';' { BGP_CFG->connect_delay_time = $5; }
  | bgp_proto CONNECT RETRY TIME expr ';' { BGP_CFG->connect_retry_time = $5; }
@@ -111,26 +120,79 @@ bgp_proto:
  | bgp_proto ENABLE AS4 bool ';' { BGP_CFG->enable_as4 = $4; }
  | bgp_proto ENABLE EXTENDED MESSAGES bool ';' { BGP_CFG->enable_extended_messages = $5; }
  | bgp_proto CAPABILITIES bool ';' { BGP_CFG->capabilities = $3; }
- | bgp_proto ADVERTISE IPV4 bool ';' { BGP_CFG->advertise_ipv4 = $4; }
  | bgp_proto PASSWORD text ';' { BGP_CFG->password = $3; }
  | bgp_proto SETKEY bool ';' { BGP_CFG->setkey = $3; }
  | bgp_proto PASSIVE bool ';' { BGP_CFG->passive = $3; }
  | bgp_proto INTERPRET COMMUNITIES bool ';' { BGP_CFG->interpret_communities = $4; }
- | bgp_proto SECONDARY bool ';' { BGP_CFG->secondary = $3; }
- | bgp_proto ADD PATHS RX ';' { BGP_CFG->add_path = ADD_PATH_RX; }
- | bgp_proto ADD PATHS TX ';' { BGP_CFG->add_path = ADD_PATH_TX; }
- | bgp_proto ADD PATHS bool ';' { BGP_CFG->add_path = $4 ? ADD_PATH_FULL : 0; }
  | bgp_proto ALLOW LOCAL AS ';' { BGP_CFG->allow_local_as = -1; }
  | bgp_proto ALLOW LOCAL AS expr ';' { BGP_CFG->allow_local_as = $5; }
  | bgp_proto GRACEFUL RESTART bool ';' { BGP_CFG->gr_mode = $4; }
  | bgp_proto GRACEFUL RESTART AWARE ';' { BGP_CFG->gr_mode = BGP_GR_AWARE; }
  | bgp_proto GRACEFUL RESTART TIME expr ';' { BGP_CFG->gr_time = $5; }
- | bgp_proto IGP TABLE rtable ';' { BGP_CFG->igp_table = $4; }
  | bgp_proto TTL SECURITY bool ';' { BGP_CFG->ttl_security = $4; }
  | bgp_proto CHECK LINK bool ';' { BGP_CFG->check_link = $4; }
  | bgp_proto BFD bool ';' { BGP_CFG->bfd = $3; cf_check_bfd($3); }
  ;
 
+bgp_afi:
+   IPV4 { $$ = BGP_AF_IPV4; }
+ | IPV6 { $$ = BGP_AF_IPV6; }
+ | IPV4 MULTICAST { $$ = BGP_AF_IPV4_MC; }
+ | IPV6 MULTICAST { $$ = BGP_AF_IPV6_MC; }
+ ;
+
+bgp_channel_start: bgp_afi
+{
+  const struct bgp_af_desc *desc = bgp_get_af_desc($1);
+
+  if (!desc)
+    cf_error("Unknown AFI/SAFI");
+
+  this_channel = channel_config_new(&channel_bgp, desc->net, this_proto);
+  BGP_CC->c.name = desc->name;
+  BGP_CC->afi = $1;
+  BGP_CC->gr_able = 0xff;	/* undefined */
+};
+
+bgp_channel_item:
+   channel_item
+ | NEXT HOP ADDRESS ipa { BGP_CC->next_hop_addr = $4; }
+ | NEXT HOP SELF { BGP_CC->next_hop_self = 1; BGP_CC->next_hop_keep = 0; }
+ | NEXT HOP KEEP { BGP_CC->next_hop_keep = 1; BGP_CC->next_hop_self = 0; }
+ | MISSING LLADDR SELF { BGP_CC->missing_lladdr = MLL_SELF; }
+ | MISSING LLADDR DROP { BGP_CC->missing_lladdr = MLL_DROP; }
+ | MISSING LLADDR IGNORE { BGP_CC->missing_lladdr = MLL_IGNORE; }
+ | GATEWAY DIRECT { BGP_CC->gw_mode = GW_DIRECT; }
+ | GATEWAY RECURSIVE { BGP_CC->gw_mode = GW_RECURSIVE; }
+ | SECONDARY bool { BGP_CC->secondary = $2; }
+ | GRACEFUL RESTART bool { BGP_CC->gr_able = $3; }
+ | ADD PATHS RX { BGP_CC->add_path = BGP_ADD_PATH_RX; }
+ | ADD PATHS TX { BGP_CC->add_path = BGP_ADD_PATH_TX; }
+ | ADD PATHS bool { BGP_CC->add_path = $3 ? BGP_ADD_PATH_FULL : 0; }
+ | IGP TABLE rtable { BGP_CC->igp_table = $3; }
+ ;
+
+bgp_channel_opts:
+   /* empty */
+ | bgp_channel_opts bgp_channel_item ';'
+ ;
+
+bgp_channel_opt_list:
+   /* empty */
+ | '{' bgp_channel_opts '}'
+ ;
+
+bgp_channel_end:
+{
+  if (!this_channel->table)
+    cf_error("Routing table not specified");
+
+  this_channel = NULL;
+};
+
+bgp_proto_channel: bgp_channel_start bgp_channel_opt_list bgp_channel_end;
+
+
 CF_ADDTO(dynamic_attr, BGP_ORIGIN
 	{ $$ = f_new_dynamic_attr(EAF_TYPE_INT, T_ENUM_BGP_ORIGIN, EA_CODE(EAP_BGP, BA_ORIGIN)); })
 CF_ADDTO(dynamic_attr, BGP_PATH

proto/static/static.c

@@ -104,7 +104,7 @@ static_install(struct proto *p, struct static_route *r, struct iface *ifa)
     }
 
   if (r->dest == RTDX_RECURSIVE)
-    rta_set_recursive_next_hop(p->main_channel->table, &a, p_igp_table(p), &r->via, &r->via);
+    rta_set_recursive_next_hop(p->main_channel->table, &a, p_igp_table(p), r->via, IPA_NONE);
 
   /* We skip rta_lookup() here */
 

sysdep/unix/io.c

@@ -1255,7 +1255,7 @@ sk_setup(sock *s)
 
   if (sk_is_ipv6(s))
   {
-    if (s->type != SK_IP)
+    if ((s->type == SK_TCP_PASSIVE) || (s->type == SK_TCP_ACTIVE) || (s->type == SK_UDP))
       if (setsockopt(fd, SOL_IPV6, IPV6_V6ONLY, &y, sizeof(y)) < 0)
 	ERR("IPV6_V6ONLY");