1 files changed, 183 insertions, 157 deletions

diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index 5c17c370..978a1891 100644
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -15,8 +15,8 @@
 #include "nest/bird.h"
 #include "nest/iface.h"
 #include "nest/protocol.h"
-#include "nest/route.h"
-#include "nest/attrs.h"
+#include "nest/rt.h"
+#include "lib/attrs.h"
 #include "proto/mrt/mrt.h"
 #include "conf/conf.h"
 #include "lib/unaligned.h"
@@ -973,7 +973,7 @@ bgp_rx_open(struct bgp_conn *conn, byte *pkt, uint len)
       conn->hold_time, conn->keepalive_time, p->remote_as, p->remote_id, conn->as4_session);
 
   bgp_schedule_packet(conn, NULL, PKT_KEEPALIVE);
-  bgp_start_timer(conn->hold_timer, conn->hold_time);
+  bgp_start_timer(p, conn->hold_timer, conn->hold_time);
   bgp_conn_enter_openconfirm_state(conn);
 }
 
@@ -1002,7 +1002,7 @@ bgp_rx_open(struct bgp_conn *conn, byte *pkt, uint len)
 #define MISMATCHED_AF	" - mismatched address family (%I for %s)"
 
 static void
-bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll)
+bgp_apply_next_hop(struct bgp_parse_state *s, ea_list **to, ip_addr gw, ip_addr ll)
 {
   struct bgp_proto *p = s->proto;
   struct bgp_channel *c = s->channel;
@@ -1025,10 +1025,18 @@ bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll)
     if (nbr->scope == SCOPE_HOST)
       WITHDRAW(BAD_NEXT_HOP " - address %I is local", nbr->addr);
 
-    a->dest = RTD_UNICAST;
-    a->nh.gw = nbr->addr;
-    a->nh.iface = nbr->iface;
-    a->igp_metric = c->cf->cost;
+    ea_set_attr_u32(to, &ea_gen_igp_metric, 0, c->cf->cost);
+
+    struct nexthop_adata nhad = {
+      .nh = {
+	.gw = nbr->addr,
+	.iface = nbr->iface,
+      },
+      .ad = {
+	.length = sizeof nhad - sizeof nhad.ad,
+      },
+    };
+    ea_set_attr_data(to, &ea_gen_nexthop, 0, nhad.ad.data, nhad.ad.length);
   }
   else /* GW_RECURSIVE */
   {
@@ -1037,59 +1045,52 @@ bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll)
 
     rtable *tab = ipa_is_ip4(gw) ? c->igp_table_ip4 : c->igp_table_ip6;
     ip_addr lla = (c->cf->next_hop_prefer == NHP_LOCAL) ? ll : IPA_NONE;
-    s->hostentry = rt_get_hostentry(tab, gw, lla, c->c.table);
-
-    if (!s->mpls)
-      rta_apply_hostentry(a, s->hostentry, NULL);
 
-    /* With MPLS, hostentry is applied later in bgp_apply_mpls_labels() */
+    if (s->mpls)
+    {
+      u32 labels[BGP_MPLS_MAX];
+      ea_set_hostentry(to, c->c.table, tab, gw, lla, BGP_MPLS_MAX, labels);
+    }
+    else
+      ea_set_hostentry(to, c->c.table, tab, gw, lla, 0, NULL);
   }
 }
 
 static void
-bgp_apply_mpls_labels(struct bgp_parse_state *s, rta *a, u32 *labels, uint lnum)
+bgp_apply_mpls_labels(struct bgp_parse_state *s, ea_list **to, u32 lnum, u32 labels[lnum])
 {
   if (lnum > MPLS_MAX_LABEL_STACK)
   {
     REPORT("Too many MPLS labels ($u)", lnum);
 
-    a->dest = RTD_UNREACHABLE;
-    a->hostentry = NULL;
-    a->nh = (struct nexthop) { };
+    ea_set_dest(to, 0, RTD_UNREACHABLE);
     return;
   }
 
   /* Handle implicit NULL as empty MPLS stack */
   if ((lnum == 1) && (labels[0] == BGP_MPLS_NULL))
-    lnum = 0;
+    lnum = s->mpls_labels->length = 0;
 
   if (s->channel->cf->gw_mode == GW_DIRECT)
   {
-    a->nh.labels = lnum;
-    memcpy(a->nh.label, labels, 4*lnum);
+    eattr *e = ea_find(*to, &ea_gen_nexthop);
+    struct {
+      struct nexthop_adata nhad;
+      u32 labels[MPLS_MAX_LABEL_STACK];
+    } nh;
+    
+    memcpy(&nh.nhad, e->u.ptr, sizeof(struct adata) + e->u.ptr->length);
+    nh.nhad.nh.labels = lnum;
+    memcpy(nh.labels, labels, lnum * sizeof(u32));
+    nh.nhad.ad.length = sizeof nh.nhad + lnum * sizeof(u32);
   }
   else /* GW_RECURSIVE */
   {
-    mpls_label_stack ms;
-
-    ms.len = lnum;
-    memcpy(ms.stack, labels, 4*lnum);
-    rta_apply_hostentry(a, s->hostentry, &ms);
-  }
-}
-
-static void
-bgp_apply_flow_validation(struct bgp_parse_state *s, const net_addr *n, rta *a)
-{
-  struct bgp_channel *c = s->channel;
-  int valid = rt_flowspec_check(c->base_table, c->c.table, n, a, s->proto->is_interior);
-  a->dest = valid ? RTD_NONE : RTD_UNREACHABLE;
-
-  /* Invalidate cached rta if dest changes */
-  if (s->cached_rta && (s->cached_rta->dest != a->dest))
-  {
-    rta_free(s->cached_rta);
-    s->cached_rta = NULL;
+    eattr *e = ea_find(*to, &ea_gen_hostentry);
+    ASSERT_DIE(e);
+    struct hostentry_adata *head = (void *) e->u.ptr;
+    memcpy(&head->labels, labels, lnum * sizeof(u32));
+    head->ad.length = (void *)(&head->labels[lnum]) - (void *) head->ad.data;
   }
 }
 
@@ -1130,7 +1131,7 @@ bgp_use_next_hop(struct bgp_export_state *s, eattr *a)
     return 0;
 
   /* Do not pass NEXT_HOP between different VRFs */
-  if (p->p.vrf_set && s->src && s->src->p.vrf_set && (p->p.vrf != s->src->p.vrf))
+  if (p->p.vrf && s->src && s->src->p.vrf && (p->p.vrf != s->src->p.vrf))
     return 0;
 
   /* Keep it when exported to internal peers */
@@ -1143,35 +1144,45 @@ bgp_use_next_hop(struct bgp_export_state *s, eattr *a)
   return p->neigh && (p->neigh->iface == ifa);
 }
 
-static inline int
+static inline struct nexthop *
 bgp_use_gateway(struct bgp_export_state *s)
 {
   struct bgp_proto *p = s->proto;
   struct bgp_channel *c = s->channel;
-  rta *ra = s->route->attrs;
+  ea_list *ra = s->route->attrs;
 
   /* Handle next hop self option - also applies to gateway */
   if (c->cf->next_hop_self && bgp_match_src(s, c->cf->next_hop_self))
-    return 0;
+    return NULL;
+
+  eattr *nhea = ea_find(ra, &ea_gen_nexthop);
+  if (!nhea)
+    return NULL;
 
   /* We need one valid global gateway */
-  if ((ra->dest != RTD_UNICAST) || ra->nh.next || ipa_zero(ra->nh.gw) || ipa_is_link_local(ra->nh.gw))
-    return 0;
+  struct nexthop_adata *nhad = (struct nexthop_adata *) nhea->u.ptr;
+  if (!NEXTHOP_IS_REACHABLE(nhad) ||
+      !NEXTHOP_ONE(nhad) || ipa_zero(nhad->nh.gw) ||
+      ipa_is_link_local(nhad->nh.gw))
+    return NULL;
 
   /* Check for non-matching AF */
-  if ((ipa_is_ip4(ra->nh.gw) != bgp_channel_is_ipv4(c)) && !c->ext_next_hop)
-    return 0;
+  if ((ipa_is_ip4(nhad->nh.gw) != bgp_channel_is_ipv4(c)) && !c->ext_next_hop)
+    return NULL;
 
   /* Do not use gateway from different VRF */
-  if (p->p.vrf_set && ra->nh.iface && (p->p.vrf != ra->nh.iface->master))
+  if (p->p.vrf && nhad->nh.iface && (p->p.vrf != nhad->nh.iface->master))
     return 0;
 
   /* Use it when exported to internal peers */
   if (p->is_interior)
-    return 1;
+    return &nhad->nh;
 
   /* Use it when forwarded to single-hop BGP peer on on the same iface */
-  return p->neigh && (p->neigh->iface == ra->nh.iface);
+  if (p->neigh && (p->neigh->iface == nhad->nh.iface))
+    return &nhad->nh;
+
+  return NULL;
 }
 
 static void
@@ -1179,36 +1190,36 @@ bgp_update_next_hop_ip(struct bgp_export_state *s, eattr *a, ea_list **to)
 {
   if (!a || !bgp_use_next_hop(s, a))
   {
-    if (bgp_use_gateway(s))
+    struct nexthop *nhloc;
+    if (nhloc = bgp_use_gateway(s))
     {
-      rta *ra = s->route->attrs;
-      ip_addr nh[1] = { ra->nh.gw };
-      bgp_set_attr_data(to, s->pool, BA_NEXT_HOP, 0, nh, 16);
+      ip_addr nh[1] = { nhloc->gw };
+      bgp_set_attr_data(to, BA_NEXT_HOP, 0, nh, 16);
 
       if (s->mpls)
       {
 	u32 implicit_null = BGP_MPLS_NULL;
-	u32 *labels = ra->nh.labels ? ra->nh.label : &implicit_null;
-	uint lnum = ra->nh.labels ? ra->nh.labels : 1;
-	bgp_set_attr_data(to, s->pool, BA_MPLS_LABEL_STACK, 0, labels, lnum * 4);
+	u32 *labels = nhloc->labels ? nhloc->label : &implicit_null;
+	uint lnum = nhloc->labels ? nhloc->labels : 1;
+	bgp_set_attr_data(to, BA_MPLS_LABEL_STACK, 0, labels, lnum * 4);
       }
       else
-	bgp_unset_attr(to, s->pool, BA_MPLS_LABEL_STACK);
+	bgp_unset_attr(to, BA_MPLS_LABEL_STACK);
     }
     else
     {
       ip_addr nh[2] = { s->channel->next_hop_addr, s->channel->link_addr };
-      bgp_set_attr_data(to, s->pool, BA_NEXT_HOP, 0, nh, ipa_nonzero(nh[1]) ? 32 : 16);
+      bgp_set_attr_data(to, BA_NEXT_HOP, 0, nh, ipa_nonzero(nh[1]) ? 32 : 16);
       s->local_next_hop = 1;
 
       /* TODO: Use local MPLS assigned label */
       if (s->mpls)
       {
 	u32 implicit_null = BGP_MPLS_NULL;
-	bgp_set_attr_data(to, s->pool, BA_MPLS_LABEL_STACK, 0, &implicit_null, 4);
+	bgp_set_attr_data(to, BA_MPLS_LABEL_STACK, 0, &implicit_null, 4);
       }
       else
-	bgp_unset_attr(to, s->pool, BA_MPLS_LABEL_STACK);
+	bgp_unset_attr(to, BA_MPLS_LABEL_STACK);
     }
   }
 
@@ -1270,7 +1281,7 @@ bgp_encode_next_hop_ip(struct bgp_write_state *s, eattr *a, byte *buf, uint size
 }
 
 static void
-bgp_decode_next_hop_ip(struct bgp_parse_state *s, byte *data, uint len, rta *a)
+bgp_decode_next_hop_ip(struct bgp_parse_state *s, byte *data, uint len, ea_list **to)
 {
   struct bgp_channel *c = s->channel;
   struct adata *ad = lp_alloc_adata(s->pool, 32);
@@ -1311,8 +1322,8 @@ bgp_decode_next_hop_ip(struct bgp_parse_state *s, byte *data, uint len, rta *a)
 
   // XXXX validate next hop
 
-  bgp_set_attr_ptr(&(a->eattrs), s->pool, BA_NEXT_HOP, 0, ad);
-  bgp_apply_next_hop(s, a, nh[0], nh[1]);
+  bgp_set_attr_ptr(to, BA_NEXT_HOP, 0, ad);
+  bgp_apply_next_hop(s, to, nh[0], nh[1]);
 }
 
 static uint
@@ -1350,7 +1361,7 @@ bgp_encode_next_hop_vpn(struct bgp_write_state *s, eattr *a, byte *buf, uint siz
 }
 
 static void
-bgp_decode_next_hop_vpn(struct bgp_parse_state *s, byte *data, uint len, rta *a)
+bgp_decode_next_hop_vpn(struct bgp_parse_state *s, byte *data, uint len, ea_list **to)
 {
   struct bgp_channel *c = s->channel;
   struct adata *ad = lp_alloc_adata(s->pool, 32);
@@ -1392,8 +1403,8 @@ bgp_decode_next_hop_vpn(struct bgp_parse_state *s, byte *data, uint len, rta *a)
 
   // XXXX validate next hop
 
-  bgp_set_attr_ptr(&(a->eattrs), s->pool, BA_NEXT_HOP, 0, ad);
-  bgp_apply_next_hop(s, a, nh[0], nh[1]);
+  bgp_set_attr_ptr(to, BA_NEXT_HOP, 0, ad);
+  bgp_apply_next_hop(s, to, nh[0], nh[1]);
 }
 
 
@@ -1405,7 +1416,7 @@ bgp_encode_next_hop_none(struct bgp_write_state *s UNUSED, eattr *a UNUSED, byte
 }
 
 static void
-bgp_decode_next_hop_none(struct bgp_parse_state *s UNUSED, byte *data UNUSED, uint len UNUSED, rta *a UNUSED)
+bgp_decode_next_hop_none(struct bgp_parse_state *s UNUSED, byte *data UNUSED, uint len UNUSED, ea_list **to UNUSED)
 {
   /*
    * Although we expect no next hop and RFC 7606 7.11 states that attribute
@@ -1417,11 +1428,11 @@ bgp_decode_next_hop_none(struct bgp_parse_state *s UNUSED, byte *data UNUSED, ui
 }
 
 static void
-bgp_update_next_hop_none(struct bgp_export_state *s, eattr *a, ea_list **to)
+bgp_update_next_hop_none(struct bgp_export_state *s UNUSED, eattr *a, ea_list **to)
 {
   /* NEXT_HOP shall not pass */
   if (a)
-    bgp_unset_attr(to, s->pool, BA_NEXT_HOP);
+    bgp_unset_attr(to, BA_NEXT_HOP);
 }
 
 
@@ -1430,15 +1441,17 @@ bgp_update_next_hop_none(struct bgp_export_state *s, eattr *a, ea_list **to)
  */
 
 static void
-bgp_rte_update(struct bgp_parse_state *s, const net_addr *n, u32 path_id, rta *a0)
+bgp_rte_update(struct bgp_parse_state *s, const net_addr *n, u32 path_id, ea_list *a0)
 {
   if (path_id != s->last_id)
   {
+    rt_unlock_source(s->last_src);
+
     s->last_src = rt_get_source(&s->proto->p, path_id);
     s->last_id = path_id;
 
-    rta_free(s->cached_rta);
-    s->cached_rta = NULL;
+    ea_free(s->cached_ea);
+    s->cached_ea = NULL;
   }
 
   if (!a0)
@@ -1448,23 +1461,20 @@ bgp_rte_update(struct bgp_parse_state *s, const net_addr *n, u32 path_id, rta *a
       REPORT("Invalid route %N withdrawn", n);
 
     /* Route withdraw */
-    rte_update3(&s->channel->c, n, NULL, s->last_src);
+    rte_update(&s->channel->c, n, NULL, s->last_src);
     return;
   }
 
   /* Prepare cached route attributes */
-  if (s->cached_rta == NULL)
-  {
-    /* Workaround for rta_lookup() breaking eattrs */
-    ea_list *ea = a0->eattrs;
-    s->cached_rta = rta_lookup(a0);
-    a0->eattrs = ea;
-  }
+  if (s->cached_ea == NULL)
+    s->cached_ea = ea_lookup(a0, 0);
 
-  rta *a = rta_clone(s->cached_rta);
-  rte *e = rte_get_temp(a, s->last_src);
+  rte e0 = {
+    .attrs = s->cached_ea,
+    .src = s->last_src,
+  };
 
-  rte_update3(&s->channel->c, n, e, s->last_src);
+  rte_update(&s->channel->c, n, &e0, s->last_src);
 }
 
 static void
@@ -1487,9 +1497,10 @@ bgp_encode_mpls_labels(struct bgp_write_state *s UNUSED, const adata *mpls, byte
 }
 
 static void
-bgp_decode_mpls_labels(struct bgp_parse_state *s, byte **pos, uint *len, uint *pxlen, rta *a)
+bgp_decode_mpls_labels(struct bgp_parse_state *s, byte **pos, uint *len, uint *pxlen, ea_list **to)
 {
-  u32 labels[BGP_MPLS_MAX], label;
+  u32 labels[BGP_MPLS_MAX];
+  u32 label;
   uint lnum = 0;
 
   do {
@@ -1508,26 +1519,15 @@ bgp_decode_mpls_labels(struct bgp_parse_state *s, byte **pos, uint *len, uint *p
   }
   while (!(label & BGP_MPLS_BOS));
 
-  if (!a)
+  if (!*to)
     return;
 
-  /* Attach MPLS attribute unless we already have one */
-  if (!s->mpls_labels)
-  {
-    s->mpls_labels = lp_alloc_adata(s->pool, 4*BGP_MPLS_MAX);
-    bgp_set_attr_ptr(&(a->eattrs), s->pool, BA_MPLS_LABEL_STACK, 0, s->mpls_labels);
-  }
-
-  /* Overwrite data in the attribute */
-  s->mpls_labels->length = 4*lnum;
-  memcpy(s->mpls_labels->data, labels, 4*lnum);
-
   /* Update next hop entry in rta */
-  bgp_apply_mpls_labels(s, a, labels, lnum);
+  bgp_apply_mpls_labels(s, to, lnum, labels);
 
   /* Attributes were changed, invalidate cached entry */
-  rta_free(s->cached_rta);
-  s->cached_rta = NULL;
+  rta_free(s->cached_ea);
+  s->cached_ea = NULL;
 
   return;
 }
@@ -1563,14 +1563,14 @@ bgp_encode_nlri_ip4(struct bgp_write_state *s, struct bgp_bucket *buck, byte *bu
     memcpy(pos, &a, b);
     ADVANCE(pos, size, b);
 
-    bgp_free_prefix(s->channel, px);
+    bgp_done_prefix(s->channel, px, buck);
   }
 
   return pos - buf;
 }
 
 static void
-bgp_decode_nlri_ip4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
+bgp_decode_nlri_ip4(struct bgp_parse_state *s, byte *pos, uint len, ea_list *a)
 {
   while (len)
   {
@@ -1596,7 +1596,7 @@ bgp_decode_nlri_ip4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
 
     /* Decode MPLS labels */
     if (s->mpls)
-      bgp_decode_mpls_labels(s, &pos, &len, &l, a);
+      bgp_decode_mpls_labels(s, &pos, &len, &l, &a);
 
     if (l > IP4_MAX_PREFIX_LENGTH)
       bgp_parse_error(s, 10);
@@ -1648,14 +1648,14 @@ bgp_encode_nlri_ip6(struct bgp_write_state *s, struct bgp_bucket *buck, byte *bu
     memcpy(pos, &a, b);
     ADVANCE(pos, size, b);
 
-    bgp_free_prefix(s->channel, px);
+    bgp_done_prefix(s->channel, px, buck);
   }
 
   return pos - buf;
 }
 
 static void
-bgp_decode_nlri_ip6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
+bgp_decode_nlri_ip6(struct bgp_parse_state *s, byte *pos, uint len, ea_list *a)
 {
   while (len)
   {
@@ -1681,7 +1681,7 @@ bgp_decode_nlri_ip6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
 
     /* Decode MPLS labels */
     if (s->mpls)
-      bgp_decode_mpls_labels(s, &pos, &len, &l, a);
+      bgp_decode_mpls_labels(s, &pos, &len, &l, &a);
 
     if (l > IP6_MAX_PREFIX_LENGTH)
       bgp_parse_error(s, 10);
@@ -1736,14 +1736,14 @@ bgp_encode_nlri_vpn4(struct bgp_write_state *s, struct bgp_bucket *buck, byte *b
     memcpy(pos, &a, b);
     ADVANCE(pos, size, b);
 
-    bgp_free_prefix(s->channel, px);
+    bgp_done_prefix(s->channel, px, buck);
   }
 
   return pos - buf;
 }
 
 static void
-bgp_decode_nlri_vpn4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
+bgp_decode_nlri_vpn4(struct bgp_parse_state *s, byte *pos, uint len, ea_list *a)
 {
   while (len)
   {
@@ -1769,7 +1769,7 @@ bgp_decode_nlri_vpn4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
 
     /* Decode MPLS labels */
     if (s->mpls)
-      bgp_decode_mpls_labels(s, &pos, &len, &l, a);
+      bgp_decode_mpls_labels(s, &pos, &len, &l, &a);
 
     /* Decode route distinguisher */
     if (l < 64)
@@ -1833,14 +1833,14 @@ bgp_encode_nlri_vpn6(struct bgp_write_state *s, struct bgp_bucket *buck, byte *b
     memcpy(pos, &a, b);
     ADVANCE(pos, size, b);
 
-    bgp_free_prefix(s->channel, px);
+    bgp_done_prefix(s->channel, px, buck);
   }
 
   return pos - buf;
 }
 
 static void
-bgp_decode_nlri_vpn6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
+bgp_decode_nlri_vpn6(struct bgp_parse_state *s, byte *pos, uint len, ea_list *a)
 {
   while (len)
   {
@@ -1866,7 +1866,7 @@ bgp_decode_nlri_vpn6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
 
     /* Decode MPLS labels */
     if (s->mpls)
-      bgp_decode_mpls_labels(s, &pos, &len, &l, a);
+      bgp_decode_mpls_labels(s, &pos, &len, &l, &a);
 
     /* Decode route distinguisher */
     if (l < 64)
@@ -1920,14 +1920,14 @@ bgp_encode_nlri_flow4(struct bgp_write_state *s, struct bgp_bucket *buck, byte *
     memcpy(pos, net->data, flen);
     ADVANCE(pos, size, flen);
 
-    bgp_free_prefix(s->channel, px);
+    bgp_done_prefix(s->channel, px, buck);
   }
 
   return pos - buf;
 }
 
 static void
-bgp_decode_nlri_flow4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
+bgp_decode_nlri_flow4(struct bgp_parse_state *s, byte *pos, uint len, ea_list *a)
 {
   while (len)
   {
@@ -1978,10 +1978,6 @@ bgp_decode_nlri_flow4(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
     net_fill_flow4(n, px, pxlen, pos, flen);
     ADVANCE(pos, len, flen);
 
-    /* Apply validation procedure per RFC 8955 (6) */
-    if (a && s->channel->cf->validate)
-      bgp_apply_flow_validation(s, n, a);
-
     bgp_rte_update(s, n, path_id, a);
   }
 }
@@ -2012,14 +2008,14 @@ bgp_encode_nlri_flow6(struct bgp_write_state *s, struct bgp_bucket *buck, byte *
     memcpy(pos, net->data, flen);
     ADVANCE(pos, size, flen);
 
-    bgp_free_prefix(s->channel, px);
+    bgp_done_prefix(s->channel, px, buck);
   }
 
   return pos - buf;
 }
 
 static void
-bgp_decode_nlri_flow6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
+bgp_decode_nlri_flow6(struct bgp_parse_state *s, byte *pos, uint len, ea_list *a)
 {
   while (len)
   {
@@ -2070,10 +2066,6 @@ bgp_decode_nlri_flow6(struct bgp_parse_state *s, byte *pos, uint len, rta *a)
     net_fill_flow6(n, px, pxlen, pos, flen);
     ADVANCE(pos, len, flen);
 
-    /* Apply validation procedure per RFC 8955 (6) */
-    if (a && s->channel->cf->validate)
-      bgp_apply_flow_validation(s, n, a);
-
     bgp_rte_update(s, n, path_id, a);
   }
 }
@@ -2250,6 +2242,8 @@ bgp_create_ip_reach(struct bgp_write_state *s, struct bgp_bucket *buck, byte *bu
    *	var	IPv4 Network Layer Reachability Information
    */
 
+  ASSERT_DIE(s->channel->ptx->withdraw_bucket != buck);
+
   int lr, la;
 
   la = bgp_encode_attrs(s, buck->eattrs, buf+4, buf + MAX_ATTRS_LENGTH);
@@ -2271,6 +2265,8 @@ bgp_create_ip_reach(struct bgp_write_state *s, struct bgp_bucket *buck, byte *bu
 static byte *
 bgp_create_mp_reach(struct bgp_write_state *s, struct bgp_bucket *buck, byte *buf, byte *end)
 {
+  ASSERT_DIE(s->channel->ptx->withdraw_bucket != buck);
+
   /*
    *	2 B	IPv4 Withdrawn Routes Length (zero)
    *	---	IPv4 Withdrawn Routes NLRI (unused)
@@ -2409,7 +2405,7 @@ again: ;
   };
 
   /* Try unreachable bucket */
-  if ((buck = c->withdraw_bucket) && !EMPTY_LIST(buck->prefixes))
+  if ((buck = c->ptx->withdraw_bucket) && !EMPTY_LIST(buck->prefixes))
   {
     res = (c->afi == BGP_AF_IPV4) && !c->ext_next_hop ?
       bgp_create_ip_unreach(&s, buck, buf, end):
@@ -2419,14 +2415,13 @@ again: ;
   }
 
   /* Try reachable buckets */
-  if (!EMPTY_LIST(c->bucket_queue))
+  if (!EMPTY_LIST(c->ptx->bucket_queue))
   {
-    buck = HEAD(c->bucket_queue);
+    buck = HEAD(c->ptx->bucket_queue);
 
     /* Cleanup empty buckets */
-    if (EMPTY_LIST(buck->prefixes))
+    if (bgp_done_bucket(c, buck))
     {
-      bgp_free_bucket(c, buck);
       lp_restore(tmp_linpool, &tmpp);
       goto again;
     }
@@ -2435,10 +2430,7 @@ again: ;
       bgp_create_ip_reach(&s, buck, buf, end):
       bgp_create_mp_reach(&s, buck, buf, end);
 
-    if (EMPTY_LIST(buck->prefixes))
-      bgp_free_bucket(c, buck);
-    else
-      bgp_defer_bucket(c, buck);
+    bgp_done_bucket(c, buck);
 
     if (!res)
     {
@@ -2522,7 +2514,6 @@ static inline void
 bgp_decode_nlri(struct bgp_parse_state *s, u32 afi, byte *nlri, uint len, ea_list *ea, byte *nh, uint nh_len)
 {
   struct bgp_channel *c = bgp_get_channel(s->proto, afi);
-  rta *a = NULL;
 
   if (!c)
     DISCARD(BAD_AFI, BGP_AFI(afi), BGP_SAFI(afi));
@@ -2533,6 +2524,7 @@ bgp_decode_nlri(struct bgp_parse_state *s, u32 afi, byte *nlri, uint len, ea_lis
 
   s->last_id = 0;
   s->last_src = s->proto->p.main_source;
+  rt_lock_source(s->last_src);
 
   /*
    * IPv4 BGP and MP-BGP may be used together in one update, therefore we do not
@@ -2543,26 +2535,24 @@ bgp_decode_nlri(struct bgp_parse_state *s, u32 afi, byte *nlri, uint len, ea_lis
 
   if (ea)
   {
-    a = allocz(RTA_MAX_SIZE);
+    ea_set_attr_data(&ea, &ea_gen_from, 0, &s->proto->remote_ip, sizeof(ip_addr));
+    ea_set_attr_u32(&ea, &ea_gen_preference, 0, c->c.preference);
+    ea_set_attr_u32(&ea, &ea_gen_source, 0, RTS_BGP);
 
-    a->source = RTS_BGP;
-    a->scope = SCOPE_UNIVERSE;
-    a->from = s->proto->remote_ip;
-    a->eattrs = ea;
-    a->pref = c->c.preference;
-
-    c->desc->decode_next_hop(s, nh, nh_len, a);
-    bgp_finish_attrs(s, a);
+    c->desc->decode_next_hop(s, nh, nh_len, &ea);
+    bgp_finish_attrs(s, &ea);
 
     /* Handle withdraw during next hop decoding */
     if (s->err_withdraw)
-      a = NULL;
+      ea = NULL;
   }
 
-  c->desc->decode_nlri(s, nlri, len, a);
+  c->desc->decode_nlri(s, nlri, len, ea);
+
+  rta_free(s->cached_ea);
+  s->cached_ea = NULL;
 
-  rta_free(s->cached_rta);
-  s->cached_rta = NULL;
+  rt_unlock_source(s->last_src);
 }
 
 static void
@@ -2582,7 +2572,7 @@ bgp_rx_update(struct bgp_conn *conn, byte *pkt, uint len)
   if (conn->state != BS_ESTABLISHED)
   { bgp_error(conn, 5, fsm_err_subcode[conn->state], NULL, 0); return; }
 
-  bgp_start_timer(conn->hold_timer, conn->hold_time);
+  bgp_start_timer(p, conn->hold_timer, conn->hold_time);
 
   struct lp_state tmpp;
   lp_save(tmp_linpool, &tmpp);
@@ -2667,7 +2657,7 @@ bgp_rx_update(struct bgp_conn *conn, byte *pkt, uint len)
 		    ea, s.mp_next_hop_data, s.mp_next_hop_len);
 
 done:
-  rta_free(s.cached_rta);
+  rta_free(s.cached_ea);
   lp_restore(tmp_linpool, &tmpp);
   return;
 }
@@ -2923,7 +2913,7 @@ bgp_fire_tx(struct bgp_conn *conn)
   {
     conn->packets_to_send &= ~(1 << PKT_KEEPALIVE);
     BGP_TRACE(D_PACKETS, "Sending KEEPALIVE");
-    bgp_start_timer(conn->keepalive_timer, conn->keepalive_time);
+    bgp_start_timer(p, conn->keepalive_timer, conn->keepalive_time);
     return bgp_send(conn, PKT_KEEPALIVE, BGP_HEADER_LENGTH);
   }
   else while (conn->channels_to_send)
@@ -2991,7 +2981,11 @@ bgp_schedule_packet(struct bgp_conn *conn, struct bgp_channel *c, int type)
 {
   ASSERT(conn->sk);
 
-  DBG("BGP: Scheduling packet type %d\n", type);
+  struct bgp_proto *p = conn->bgp;
+  if (c)
+    BGP_TRACE(D_PACKETS, "Scheduling packet type %d for channel %s", type, c->c.name);
+  else
+    BGP_TRACE(D_PACKETS, "Scheduling packet type %d", type);
 
   if (c)
   {
@@ -3008,7 +3002,7 @@ bgp_schedule_packet(struct bgp_conn *conn, struct bgp_channel *c, int type)
     conn->packets_to_send |= 1 << type;
 
   if ((conn->sk->tpos == conn->sk->tbuf) && !ev_active(conn->tx_ev))
-    ev_schedule(conn->tx_ev);
+    proto_send_event(&p->p, conn->tx_ev);
 }
 void
 bgp_kick_tx(void *vconn)
@@ -3021,7 +3015,7 @@ bgp_kick_tx(void *vconn)
     ;
 
   if (!max && !ev_active(conn->tx_ev))
-    ev_schedule(conn->tx_ev);
+    proto_send_event(&conn->bgp->p, conn->tx_ev);
 }
 
 void
@@ -3029,13 +3023,14 @@ bgp_tx(sock *sk)
 {
   struct bgp_conn *conn = sk->data;
 
+  ASSERT_DIE(birdloop_inside(conn->bgp->p.loop));
   DBG("BGP: TX hook\n");
   uint max = 1024;
   while (--max && (bgp_fire_tx(conn) > 0))
     ;
 
   if (!max && !ev_active(conn->tx_ev))
-    ev_schedule(conn->tx_ev);
+    proto_send_event(&conn->bgp->p, conn->tx_ev);
 }
 
 
@@ -3224,7 +3219,7 @@ bgp_rx_keepalive(struct bgp_conn *conn)
   struct bgp_proto *p = conn->bgp;
 
   BGP_TRACE(D_PACKETS, "Got KEEPALIVE");
-  bgp_start_timer(conn->hold_timer, conn->hold_time);
+  bgp_start_timer(p, conn->hold_timer, conn->hold_time);
 
   if (conn->state == BS_OPENCONFIRM)
   { bgp_conn_enter_established_state(conn); return; }
@@ -3266,6 +3261,30 @@ bgp_rx_packet(struct bgp_conn *conn, byte *pkt, uint len)
   }
 }
 
+void
+bgp_uncork(void *vp)
+{
+  /* The uncork event is run from &main_birdloop and there is no useful way how
+   * to assign the target loop to it, thus we have to lock it ourselves. */
+
+  struct bgp_proto *p = vp;
+  if (!p)
+    return;
+
+  birdloop_enter(p->p.loop);
+
+  if (p && p->conn && (p->conn->state == BS_ESTABLISHED) && !p->conn->sk->rx_hook)
+  {
+    struct birdsock *sk = p->conn->sk;
+    ASSERT_DIE(sk->rpos > sk->rbuf);
+    sk_resume_rx(p->p.loop, sk, bgp_rx);
+    bgp_rx(sk, sk->rpos - sk->rbuf);
+    BGP_TRACE(D_PACKETS, "Uncorked");
+  }
+
+  birdloop_leave(p->p.loop);
+}
+
 /**
  * bgp_rx - handle received data
  * @sk: socket
@@ -3280,6 +3299,7 @@ int
 bgp_rx(sock *sk, uint size)
 {
   struct bgp_conn *conn = sk->data;
+  struct bgp_proto *p = conn->bgp;
   byte *pkt_start = sk->rbuf;
   byte *end = pkt_start + size;
   uint i, len;
@@ -3289,6 +3309,12 @@ bgp_rx(sock *sk, uint size)
     {
       if ((conn->state == BS_CLOSE) || (conn->sk != sk))
 	return 0;
+      if ((conn->state == BS_ESTABLISHED) && rt_cork_check(conn->bgp->uncork_ev))
+      {
+	sk_pause_rx(p->p.loop, sk);
+	BGP_TRACE(D_PACKETS, "Corked");
+	return 0;
+      }
       for(i=0; i<16; i++)
 	if (pkt_start[i] != 0xff)
 	  {