From 4e276a8920ed0496836f002f144943ab42f120f6 Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Fri, 6 May 2016 15:48:35 +0200
Subject: Merged multipath and single-path data structures.

Dropped struct mpnh and mpnh_*()
Now struct nexthop exists, nexthop_*(), and also included struct nexthop
into struct rta.

Also converted RTD_DEVICE and RTD_ROUTER to RTD_UNICAST. If it is needed
to distinguish between these two cases, RTD_DEVICE is equivalent to
IPA_ZERO(a->nh.gw), RTD_ROUTER is then IPA_NONZERO(a->nh.gw).

From now on, we also explicitely want C99 compatible compiler. We assume
that this 20-year norm should be known almost everywhere.
---
 nest/route.h | 45 ++++++++++++++++++++++-----------------------
 1 file changed, 22 insertions(+), 23 deletions(-)

(limited to 'nest/route.h')

diff --git a/nest/route.h b/nest/route.h
index d652ca15..37c9abfb 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -195,7 +195,7 @@ struct hostentry {
   unsigned hash_key;			/* Hash key */
   unsigned uc;				/* Use count */
   struct rta *src;			/* Source rta entry */
-  ip_addr gw;				/* Chosen next hop */
+  struct nexthop *nh;			/* Chosen next hop */
   byte dest;				/* Chosen route destination type (RTD_...) */
   u32 igp_metric;			/* Chosen route IGP metric */
 };
@@ -332,11 +332,11 @@ void rt_show(struct rt_show_data *);
  *	construction of BGP route attribute lists.
  */
 
-/* Multipath next-hop */
-struct mpnh {
+/* Nexthop structure */
+struct nexthop {
   ip_addr gw;				/* Next hop */
   struct iface *iface;			/* Outgoing interface */
-  struct mpnh *next;
+  struct nexthop *next;
   byte weight;
 };
 
@@ -353,20 +353,19 @@ typedef struct rta {
   struct rta *next, **pprev;		/* Hash chain */
   u32 uc;				/* Use count */
   u32 hash_key;				/* Hash over important fields */
-  struct mpnh *nexthops;		/* Next-hops for multipath routes */
   struct ea_list *eattrs;		/* Extended Attribute chain */
   struct rte_src *src;			/* Route source that created the route */
   struct hostentry *hostentry;		/* Hostentry for recursive next-hops */
-  struct iface *iface;			/* Outgoing interface */
-  ip_addr gw;				/* Next hop */
   ip_addr from;				/* Advertising router */
   u32 igp_metric;			/* IGP metric to next hop (for iBGP routes) */
-  byte source;				/* Route source (RTS_...) */
-  byte scope;				/* Route scope (SCOPE_... -- see ip.h) */
-  byte cast;				/* Casting type (RTC_...) */
-  byte dest;				/* Route destination type (RTD_...) */
-  byte flags;				/* Route flags (RTF_...), now unused */
-  byte aflags;				/* Attribute cache flags (RTAF_...) */
+  u32 bf[0];
+  u32 source:6;				/* Route source (RTS_...) */
+  u32 scope:6;				/* Route scope (SCOPE_... -- see ip.h) */
+  u32 cast:6;				/* Casting type (RTC_...) */
+  u32 dest:6;				/* Route destination type (RTD_...) */
+//  u32 eflags:8;				/* Flags (RTAF_...) */
+  u32 aflags:8;
+  struct nexthop nh;			/* Next hop */
 } rta;
 
 #define RTS_DUMMY 0			/* Dummy route to be removed soon */
@@ -391,12 +390,10 @@ typedef struct rta {
 #define RTC_MULTICAST 2
 #define RTC_ANYCAST 3			/* IPv6 Anycast */
 
-#define RTD_ROUTER 0			/* Next hop is neighbor router */
-#define RTD_DEVICE 1			/* Points to device */
+#define RTD_UNICAST 0			/* Next hop is neighbor router */
 #define RTD_BLACKHOLE 2			/* Silently drop packets */
 #define RTD_UNREACHABLE 3		/* Reject as unreachable */
 #define RTD_PROHIBIT 4			/* Administratively prohibited */
-#define RTD_MULTIPATH 5			/* Multipath route (nexthops != NULL) */
 #define RTD_NONE 6			/* Invalid RTD */
 
 					/* Flags for net->n.flags, used by kernel syncer */
@@ -411,7 +408,7 @@ typedef struct rta {
 
 /* Route has regular, reachable nexthop (i.e. not RTD_UNREACHABLE and like) */
 static inline int rte_is_reachable(rte *r)
-{ uint d = r->attrs->dest; return (d == RTD_ROUTER) || (d == RTD_DEVICE) || (d == RTD_MULTIPATH); }
+{ uint d = r->attrs->dest; return (d == RTD_UNICAST); }
 
 
 /*
@@ -516,12 +513,14 @@ uint ea_hash(ea_list *e);	/* Calculate 16-bit hash value */
 ea_list *ea_append(ea_list *to, ea_list *what);
 void ea_format_bitfield(struct eattr *a, byte *buf, int bufsize, const char **names, int min, int max);
 
-int mpnh__same(struct mpnh *x, struct mpnh *y); /* Compare multipath nexthops */
-static inline int mpnh_same(struct mpnh *x, struct mpnh *y)
-{ return (x == y) || mpnh__same(x, y); }
-struct mpnh *mpnh_merge(struct mpnh *x, struct mpnh *y, int rx, int ry, int max, linpool *lp);
-void mpnh_insert(struct mpnh **n, struct mpnh *y);
-int mpnh_is_sorted(struct mpnh *x);
+int nexthop__same(struct nexthop *x, struct nexthop *y); /* Compare multipath nexthops */
+static inline int nexthop_same(struct nexthop *x, struct nexthop *y)
+{ return (x == y) || nexthop__same(x, y); }
+struct nexthop *nexthop_merge(struct nexthop *x, struct nexthop *y, int rx, int ry, int max, linpool *lp);
+static inline void nexthop_link(struct rta *a, struct nexthop *from)
+{ a->nh.gw = from->gw; a->nh.iface = from->iface; a->nh.weight = from->weight; a->nh.next = from->next; }
+void nexthop_insert(struct nexthop *n, struct nexthop *y);
+int nexthop_is_sorted(struct nexthop *x);
 
 void rta_init(void);
 rta *rta_lookup(rta *);			/* Get rta equivalent to this one, uc++ */
-- 
cgit v1.2.3


From 5b208e296fed0beddce16188478c5119df610d89 Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Tue, 7 Jun 2016 11:46:07 +0200
Subject: Removing (struct rta)->cast. Never used.

---
 filter/config.Y        |  3 +--
 filter/filter.c        |  1 -
 filter/filter.h        |  7 +++----
 nest/config.Y          |  1 -
 nest/route.h           | 11 ++++-------
 nest/rt-attr.c         | 13 ++++++-------
 nest/rt-dev.c          |  1 -
 proto/bgp/packets.c    |  1 -
 proto/ospf/rt.c        |  1 -
 proto/rip/rip.c        |  1 -
 proto/rpki/rpki.c      |  1 -
 proto/static/static.c  | 11 +++++------
 sysdep/bsd/krt-sock.c  |  2 --
 sysdep/linux/netlink.c |  4 ----
 14 files changed, 19 insertions(+), 39 deletions(-)

(limited to 'nest/route.h')

diff --git a/filter/config.Y b/filter/config.Y
index 7b4178be..94a7e307 100644
--- a/filter/config.Y
+++ b/filter/config.Y
@@ -397,7 +397,7 @@ CF_KEYWORDS(FUNCTION, PRINT, PRINTN, UNSET, RETURN,
 	SET, STRING, BGPMASK, BGPPATH, CLIST, ECLIST, LCLIST,
 	IF, THEN, ELSE, CASE,
 	TRUE, FALSE, RT, RO, UNKNOWN, GENERIC,
-	FROM, GW, NET, MASK, PROTO, SOURCE, SCOPE, CAST, DEST, IFNAME, IFINDEX,
+	FROM, GW, NET, MASK, PROTO, SOURCE, SCOPE, DEST, IFNAME, IFINDEX,
 	PREFERENCE,
 	ROA_CHECK, ASN,
 	LEN, MAXLEN,
@@ -854,7 +854,6 @@ static_attr:
  | PROTO   { $$ = f_new_inst(); $$->aux = T_STRING;     $$->a2.i = SA_PROTO; }
  | SOURCE  { $$ = f_new_inst(); $$->aux = T_ENUM_RTS;   $$->a2.i = SA_SOURCE; }
  | SCOPE   { $$ = f_new_inst(); $$->aux = T_ENUM_SCOPE; $$->a2.i = SA_SCOPE;	$$->a1.i = 1; }
- | CAST    { $$ = f_new_inst(); $$->aux = T_ENUM_RTC;   $$->a2.i = SA_CAST; }
  | DEST    { $$ = f_new_inst(); $$->aux = T_ENUM_RTD;   $$->a2.i = SA_DEST;	$$->a1.i = 1; }
  | IFNAME  { $$ = f_new_inst(); $$->aux = T_STRING;     $$->a2.i = SA_IFNAME; }
  | IFINDEX { $$ = f_new_inst(); $$->aux = T_INT;        $$->a2.i = SA_IFINDEX; }
diff --git a/filter/filter.c b/filter/filter.c
index 926316ac..bc80997f 100644
--- a/filter/filter.c
+++ b/filter/filter.c
@@ -905,7 +905,6 @@ interpret(struct f_inst *what)
       case SA_PROTO:	res.val.s = rta->src->proto->name; break;
       case SA_SOURCE:	res.val.i = rta->source; break;
       case SA_SCOPE:	res.val.i = rta->scope; break;
-      case SA_CAST:	res.val.i = rta->cast; break;
       case SA_DEST:	res.val.i = rta->dest; break;
       case SA_IFNAME:	res.val.s = rta->nh.iface ? rta->nh.iface->name : ""; break;
       case SA_IFINDEX:	res.val.i = rta->nh.iface ? rta->nh.iface->index : 0; break;
diff --git a/filter/filter.h b/filter/filter.h
index a4808731..0482b83b 100644
--- a/filter/filter.h
+++ b/filter/filter.h
@@ -174,10 +174,9 @@ void val_format(struct f_val v, buffer *buf);
 #define SA_PROTO	 4
 #define SA_SOURCE	 5
 #define SA_SCOPE	 6
-#define SA_CAST		 7
-#define SA_DEST		 8
-#define SA_IFNAME	 9
-#define SA_IFINDEX	10
+#define SA_DEST    	 8
+#define SA_IFNAME  	 9
+#define SA_IFINDEX    	10
 
 
 struct f_tree {
diff --git a/nest/config.Y b/nest/config.Y
index 95ce59cd..511936ef 100644
--- a/nest/config.Y
+++ b/nest/config.Y
@@ -78,7 +78,6 @@ CF_KEYWORDS(GRACEFUL, RESTART, WAIT, MAX, FLUSH, AS)
 CF_ENUM(T_ENUM_RTS, RTS_, DUMMY, STATIC, INHERIT, DEVICE, STATIC_DEVICE, REDIRECT,
 	RIP, OSPF, OSPF_IA, OSPF_EXT1, OSPF_EXT2, BGP, PIPE, BABEL)
 CF_ENUM(T_ENUM_SCOPE, SCOPE_, HOST, LINK, SITE, ORGANIZATION, UNIVERSE, UNDEFINED)
-CF_ENUM(T_ENUM_RTC, RTC_, UNICAST, BROADCAST, MULTICAST, ANYCAST)
 CF_ENUM(T_ENUM_RTD, RTD_, UNICAST, BLACKHOLE, UNREACHABLE, PROHIBIT)
 CF_ENUM(T_ENUM_ROA, ROA_, UNKNOWN, VALID, INVALID)
 
diff --git a/nest/route.h b/nest/route.h
index 37c9abfb..154e027e 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -358,13 +358,10 @@ typedef struct rta {
   struct hostentry *hostentry;		/* Hostentry for recursive next-hops */
   ip_addr from;				/* Advertising router */
   u32 igp_metric;			/* IGP metric to next hop (for iBGP routes) */
-  u32 bf[0];
-  u32 source:6;				/* Route source (RTS_...) */
-  u32 scope:6;				/* Route scope (SCOPE_... -- see ip.h) */
-  u32 cast:6;				/* Casting type (RTC_...) */
-  u32 dest:6;				/* Route destination type (RTD_...) */
-//  u32 eflags:8;				/* Flags (RTAF_...) */
-  u32 aflags:8;
+  u8 source;				/* Route source (RTS_...) */
+  u8 scope;				/* Route scope (SCOPE_... -- see ip.h) */
+  u8 dest;				/* Route destination type (RTD_...) */
+  u8 aflags;
   struct nexthop nh;			/* Next hop */
 } rta;
 
diff --git a/nest/rt-attr.c b/nest/rt-attr.c
index 0eacfe3f..e575ba4a 100644
--- a/nest/rt-attr.c
+++ b/nest/rt-attr.c
@@ -1008,7 +1008,9 @@ rta_hash(rta *a)
   MIX(hostentry);
   MIX(from);
   MIX(igp_metric);
-  mem_hash_mix(&h, a->bf, sizeof(u32));
+  MIX(source);
+  MIX(scope);
+  MIX(dest);
 #undef MIX
 
   return mem_hash_value(&h) ^ nexthop_hash(&(a->nh)) ^ ea_hash(a->eattrs);
@@ -1020,7 +1022,6 @@ rta_same(rta *x, rta *y)
   return (x->src == y->src &&
 	  x->source == y->source &&
 	  x->scope == y->scope &&
-	  x->cast == y->cast &&
 	  x->dest == y->dest &&
 	  x->igp_metric == y->igp_metric &&
 	  ipa_equal(x->nh.gw, y->nh.gw) &&
@@ -1163,11 +1164,10 @@ rta_dump(rta *a)
 			 "RTS_STAT_DEV", "RTS_REDIR", "RTS_RIP",
 			 "RTS_OSPF", "RTS_OSPF_IA", "RTS_OSPF_EXT1",
                          "RTS_OSPF_EXT2", "RTS_BGP", "RTS_PIPE", "RTS_BABEL" };
-  static char *rtc[] = { "", " BC", " MC", " AC" };
   static char *rtd[] = { "", " DEV", " HOLE", " UNREACH", " PROHIBIT" };
 
-  debug("p=%s uc=%d %s %s%s%s h=%04x",
-	a->src->proto->name, a->uc, rts[a->source], ip_scope_text(a->scope), rtc[a->cast],
+  debug("p=%s uc=%d %s %s%s h=%04x",
+	a->src->proto->name, a->uc, rts[a->source], ip_scope_text(a->scope),
 	rtd[a->dest], a->hash_key);
   if (!(a->aflags & RTAF_CACHED))
     debug(" !CACHED");
@@ -1213,10 +1213,9 @@ rta_show(struct cli *c, rta *a, ea_list *eal)
 {
   static char *src_names[] = { "dummy", "static", "inherit", "device", "static-device", "redirect",
 			       "RIP", "OSPF", "OSPF-IA", "OSPF-E1", "OSPF-E2", "BGP", "pipe" };
-  static char *cast_names[] = { "unicast", "broadcast", "multicast", "anycast" };
   int i;
 
-  cli_printf(c, -1008, "\tType: %s %s %s", src_names[a->source], cast_names[a->cast], ip_scope_text(a->scope));
+  cli_printf(c, -1008, "\tType: %s %s", src_names[a->source], ip_scope_text(a->scope));
   if (!eal)
     eal = a->eattrs;
   for(; eal; eal=eal->next)
diff --git a/nest/rt-dev.c b/nest/rt-dev.c
index 43628af8..5edd1c5d 100644
--- a/nest/rt-dev.c
+++ b/nest/rt-dev.c
@@ -78,7 +78,6 @@ dev_ifa_notify(struct proto *P, uint flags, struct ifa *ad)
 	.src = src,
 	.source = RTS_DEVICE,
 	.scope = SCOPE_UNIVERSE,
-	.cast = RTC_UNICAST,
 	.dest = RTD_UNICAST,
 	.nh = {
 	  .iface = ad->iface
diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index 0baa84c9..9c59e6d8 100644
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -1439,7 +1439,6 @@ bgp_decode_nlri(struct bgp_parse_state *s, u32 afi, byte *nlri, uint len, ea_lis
 
     a->source = RTS_BGP;
     a->scope = SCOPE_UNIVERSE;
-    a->cast = RTC_UNICAST;
     a->dest = RTD_UNREACHABLE;
     a->from = s->proto->cf->remote_ip;
     a->eattrs = ea;
diff --git a/proto/ospf/rt.c b/proto/ospf/rt.c
index 09cc5776..d28d463b 100644
--- a/proto/ospf/rt.c
+++ b/proto/ospf/rt.c
@@ -1951,7 +1951,6 @@ again1:
 	.src = p->p.main_source,
 	.source = nf->n.type,
 	.scope = SCOPE_UNIVERSE,
-	.cast = RTC_UNICAST
       };
 
       nexthop_link(&a0, nf->n.nhs);
diff --git a/proto/rip/rip.c b/proto/rip/rip.c
index 8b09330c..9bed9249 100644
--- a/proto/rip/rip.c
+++ b/proto/rip/rip.c
@@ -147,7 +147,6 @@ rip_announce_rte(struct rip_proto *p, struct rip_entry *en)
       .src = p->p.main_source,
       .source = RTS_RIP,
       .scope = SCOPE_UNIVERSE,
-      .cast = RTC_UNICAST
     };
 
     u8 rt_metric = rt->metric;
diff --git a/proto/rpki/rpki.c b/proto/rpki/rpki.c
index 6360dbaf..81268e83 100644
--- a/proto/rpki/rpki.c
+++ b/proto/rpki/rpki.c
@@ -124,7 +124,6 @@ rpki_table_add_roa(struct rpki_cache *cache, struct channel *channel, const net_
     .src = p->p.main_source,
     .source = RTS_RPKI,
     .scope = SCOPE_UNIVERSE,
-    .cast = RTC_UNICAST,
     .dest = RTD_BLACKHOLE,
   };
 
diff --git a/proto/static/static.c b/proto/static/static.c
index f3cfec01..a63d4d29 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -71,7 +71,6 @@ static_install(struct proto *p, struct static_route *r)
   a.src = p->main_source;
   a.source = ((r->dest == RTD_UNICAST) && ipa_zero(r->via)) ? RTS_STATIC_DEVICE : RTS_STATIC;
   a.scope = SCOPE_UNIVERSE;
-  a.cast = RTC_UNICAST;
   a.dest = r->dest;
   if (r->dest == RTD_UNICAST)
     {
@@ -210,13 +209,13 @@ static_add(struct proto *p, struct static_config *cf, struct static_route *r)
 	  }
 
 	if (count)
-	  static_install(p, r, NULL);
+	  static_install(p, r);
 
 	break;
       }
 
     default:
-      static_install(p, r, NULL);
+      static_install(p, r);
     }
 }
 
@@ -300,7 +299,7 @@ static_update_rte(struct proto *p, struct static_route *r)
     return;
 
   if (static_decide((struct static_config *) p->cf, r))
-    static_install(p, r, r->neigh->iface);
+    static_install(p, r);
   else
     static_remove(p, r);
 }
@@ -367,7 +366,7 @@ static_if_notify(struct proto *p, unsigned flags, struct iface *i)
     {
       WALK_LIST(r, c->iface_routes)
 	if (!strcmp(r->if_name, i->name))
-	  static_install(p, r, i);
+	  static_install(p, r);
     }
   else if (flags & IF_CHANGE_DOWN)
     {
@@ -535,7 +534,7 @@ static_reconfigure(struct proto *p, struct proto_config *CF)
     {
       struct iface *ifa;
       if ((ifa = if_find_by_name(r->if_name)) && (ifa->flags & IF_UP))
-	static_install(p, r, ifa);
+	static_install(p, r);
     }
   WALK_LIST(r, n->other_routes)
     static_add(p, n, r);
diff --git a/sysdep/bsd/krt-sock.c b/sysdep/bsd/krt-sock.c
index fbaa8e32..2e0f194b 100644
--- a/sysdep/bsd/krt-sock.c
+++ b/sysdep/bsd/krt-sock.c
@@ -147,7 +147,6 @@ krt_capable(rte *e)
   rta *a = e->attrs;
 
   return
-    a->cast == RTC_UNICAST &&
     ((a->dest == RTD_UNICAST && !a->nh.next) /* No multipath support */
 #ifdef RTF_REJECT
      || a->dest == RTD_UNREACHABLE
@@ -470,7 +469,6 @@ krt_read_route(struct ks_msg *msg, struct krt_proto *p, int scan)
     .src = p->p.main_source,
     .source = RTS_INHERIT,
     .scope = SCOPE_UNIVERSE,
-    .cast = RTC_UNICAST
   };
 
   /* reject/blackhole routes have also set RTF_GATEWAY,
diff --git a/sysdep/linux/netlink.c b/sysdep/linux/netlink.c
index 6e75ee53..c9a30719 100644
--- a/sysdep/linux/netlink.c
+++ b/sysdep/linux/netlink.c
@@ -952,9 +952,6 @@ krt_capable(rte *e)
 {
   rta *a = e->attrs;
 
-  if (a->cast != RTC_UNICAST)
-    return 0;
-
   switch (a->dest)
     {
     case RTD_UNICAST:
@@ -1345,7 +1342,6 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
   ra->src = p->p.main_source;
   ra->source = RTS_INHERIT;
   ra->scope = SCOPE_UNIVERSE;
-  ra->cast = RTC_UNICAST;
 
   switch (i->rtm_type)
     {
-- 
cgit v1.2.3


From ec5e5d23faa482495c84163c4ae38d9a31bdc00f Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Fri, 10 Jun 2016 14:34:41 +0200
Subject: Nexthop: Support for label stack in nest

---
 nest/route.h          |  9 +++++-
 nest/rt-attr.c        | 82 ++++++++++++++++++++++++++++++++++++++-------------
 nest/rt-table.c       |  2 +-
 proto/ospf/rt.c       |  1 +
 proto/pipe/pipe.c     | 17 ++++++-----
 proto/static/static.c |  1 +
 6 files changed, 83 insertions(+), 29 deletions(-)

(limited to 'nest/route.h')

diff --git a/nest/route.h b/nest/route.h
index 154e027e..d9a1737b 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -338,6 +338,8 @@ struct nexthop {
   struct iface *iface;			/* Outgoing interface */
   struct nexthop *next;
   byte weight;
+  byte labels;				/* Number of labels appended */
+  u32 label[0];
 };
 
 struct rte_src {
@@ -510,16 +512,21 @@ uint ea_hash(ea_list *e);	/* Calculate 16-bit hash value */
 ea_list *ea_append(ea_list *to, ea_list *what);
 void ea_format_bitfield(struct eattr *a, byte *buf, int bufsize, const char **names, int min, int max);
 
+#define NEXTHOP_MAX_LABEL_STACK 8
+
+static inline size_t nexthop_size(const struct nexthop *nh)
+{ return sizeof(struct nexthop) + sizeof(u32)*nh->labels; }
 int nexthop__same(struct nexthop *x, struct nexthop *y); /* Compare multipath nexthops */
 static inline int nexthop_same(struct nexthop *x, struct nexthop *y)
 { return (x == y) || nexthop__same(x, y); }
 struct nexthop *nexthop_merge(struct nexthop *x, struct nexthop *y, int rx, int ry, int max, linpool *lp);
 static inline void nexthop_link(struct rta *a, struct nexthop *from)
-{ a->nh.gw = from->gw; a->nh.iface = from->iface; a->nh.weight = from->weight; a->nh.next = from->next; }
+{ memcpy(&a->nh, from, nexthop_size(from)); }
 void nexthop_insert(struct nexthop *n, struct nexthop *y);
 int nexthop_is_sorted(struct nexthop *x);
 
 void rta_init(void);
+static inline size_t rta_size(const rta *a) { return sizeof(rta) + sizeof(u32)*a->nh.labels; }
 rta *rta_lookup(rta *);			/* Get rta equivalent to this one, uc++ */
 static inline int rta_is_cached(rta *r) { return r->aflags & RTAF_CACHED; }
 static inline rta *rta_clone(rta *r) { r->uc++; return r; }
diff --git a/nest/rt-attr.c b/nest/rt-attr.c
index e575ba4a..d3671a53 100644
--- a/nest/rt-attr.c
+++ b/nest/rt-attr.c
@@ -60,8 +60,8 @@
 
 pool *rta_pool;
 
-static slab *rta_slab;
-static slab *nexthop_slab;
+static slab *rta_slab_[4];
+static slab *nexthop_slab_[4];
 static slab *rte_src_slab;
 
 static struct idm src_ids;
@@ -148,7 +148,11 @@ nexthop_hash(struct nexthop *x)
 {
   u32 h = 0;
   for (; x; x = x->next)
-    h ^= ipa_hash(x->gw);
+  {
+    h ^= ipa_hash(x->gw) ^ (h << 5) ^ (h >> 9);
+    for (int i=0; i<x->labels; i++)
+      h ^= x->label[i] ^ (h << 6) ^ (h >> 7);
+  }
 
   return h;
 }
@@ -157,10 +161,15 @@ int
 nexthop__same(struct nexthop *x, struct nexthop *y)
 {
   for (; x && y; x = x->next, y = y->next)
-    if (!ipa_equal(x->gw, y->gw) || (x->iface != y->iface) || (x->weight != y->weight))
+  {
+    if (!ipa_equal(x->gw, y->gw) || (x->iface != y->iface) || (x->weight != y->weight) || (x->labels != y->labels))
       return 0;
+    for (int i=0; i<x->labels; i++)
+      if (x->label[i] != y->label[i])
+	return 0;
+  }
 
-  return x == y;
+  return 1;
 }
 
 static int
@@ -182,17 +191,28 @@ nexthop_compare_node(struct nexthop *x, struct nexthop *y)
   if (r)
     return r;
 
+  r = ((int) y->labels) - ((int) x->labels);
+  if (r)
+    return r;
+
+  for (int i=0; i<y->labels; i++)
+  {
+    r = ((int) y->label[i]) - ((int) x->label[i]);
+    if (r)
+      return r;
+  }
+
   return ((int) x->iface->index) - ((int) y->iface->index);
 }
 
 static inline struct nexthop *
 nexthop_copy_node(const struct nexthop *src, linpool *lp)
 {
-  struct nexthop *n = lp_alloc(lp, sizeof(struct nexthop));
-  n->gw = src->gw;
-  n->iface = src->iface;
+  struct nexthop *n = lp_alloc(lp, nexthop_size(src));
+
+  memcpy(n, src, nexthop_size(src));
   n->next = NULL;
-  n->weight = src->weight;
+
   return n;
 }
 
@@ -291,6 +311,12 @@ nexthop_is_sorted(struct nexthop *x)
   return 1;
 }
 
+static inline slab *
+nexthop_slab(struct nexthop *nh)
+{
+  return nexthop_slab_[nh->labels > 2 ? 3 : nh->labels];
+}
+
 static struct nexthop *
 nexthop_copy(struct nexthop *o)
 {
@@ -299,7 +325,7 @@ nexthop_copy(struct nexthop *o)
 
   for (; o; o = o->next)
     {
-      struct nexthop *n = sl_alloc(nexthop_slab);
+      struct nexthop *n = sl_alloc(nexthop_slab(o));
       n->gw = o->gw;
       n->iface = o->iface;
       n->next = NULL;
@@ -320,7 +346,7 @@ nexthop_free(struct nexthop *o)
   while (o)
     {
       n = o->next;
-      sl_free(nexthop_slab, o);
+      sl_free(nexthop_slab(o), o);
       o = n;
     }
 }
@@ -1024,20 +1050,24 @@ rta_same(rta *x, rta *y)
 	  x->scope == y->scope &&
 	  x->dest == y->dest &&
 	  x->igp_metric == y->igp_metric &&
-	  ipa_equal(x->nh.gw, y->nh.gw) &&
 	  ipa_equal(x->from, y->from) &&
-	  x->nh.iface == y->nh.iface &&
 	  x->hostentry == y->hostentry &&
 	  nexthop_same(&(x->nh), &(y->nh)) &&
 	  ea_same(x->eattrs, y->eattrs));
 }
 
+static inline slab *
+rta_slab(rta *a)
+{
+  return rta_slab_[a->nh.labels > 2 ? 3 : a->nh.labels];
+}
+
 static rta *
 rta_copy(rta *o)
 {
-  rta *r = sl_alloc(rta_slab);
+  rta *r = sl_alloc(rta_slab(o));
 
-  memcpy(r, o, sizeof(rta));
+  memcpy(r, o, rta_size(o));
   r->uc = 1;
   r->nh.next = nexthop_copy(o->nh.next);
   r->eattrs = ea_list_copy(o->eattrs);
@@ -1138,14 +1168,14 @@ rta__free(rta *a)
   if (a->nh.next)
     nexthop_free(a->nh.next);
   ea_free(a->eattrs);
-  sl_free(rta_slab, a);
+  sl_free(rta_slab(a), a);
 }
 
 rta *
 rta_do_cow(rta *o, linpool *lp)
 {
-  rta *r = lp_alloc(lp, sizeof(rta));
-  memcpy(r, o, sizeof(rta));
+  rta *r = lp_alloc(lp, rta_size(o));
+  memcpy(r, o, rta_size(o));
   r->aflags = 0;
   r->uc = 0;
   return r;
@@ -1176,6 +1206,9 @@ rta_dump(rta *a)
     for (struct nexthop *nh = &(a->nh); nh; nh = nh->next)
       {
 	if (ipa_nonzero(nh->gw)) debug(" ->%I", nh->gw);
+	if (nh->labels) debug(" L %d", nh->label[0]);
+	for (int i=1; i<nh->labels; i++)
+	  debug("/%d", nh->label[i]);
 	debug(" [%s]", nh->iface ? nh->iface->name : "???");
       }
   if (a->eattrs)
@@ -1233,8 +1266,17 @@ void
 rta_init(void)
 {
   rta_pool = rp_new(&root_pool, "Attributes");
-  rta_slab = sl_new(rta_pool, sizeof(rta));
-  nexthop_slab = sl_new(rta_pool, sizeof(struct nexthop));
+
+  rta_slab_[0] = sl_new(rta_pool, sizeof(rta));
+  rta_slab_[1] = sl_new(rta_pool, sizeof(rta) + sizeof(u32));
+  rta_slab_[2] = sl_new(rta_pool, sizeof(rta) + sizeof(u32)*2);
+  rta_slab_[3] = sl_new(rta_pool, sizeof(rta) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK);
+
+  nexthop_slab_[0] = sl_new(rta_pool, sizeof(struct nexthop));
+  nexthop_slab_[1] = sl_new(rta_pool, sizeof(struct nexthop) + sizeof(u32));
+  nexthop_slab_[2] = sl_new(rta_pool, sizeof(struct nexthop) + sizeof(u32)*2);
+  nexthop_slab_[3] = sl_new(rta_pool, sizeof(struct nexthop) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK);
+
   rta_alloc_hash();
   rte_src_init();
 }
diff --git a/nest/rt-table.c b/nest/rt-table.c
index 46857d0d..a7ceddb8 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -1782,7 +1782,7 @@ static inline rte *
 rt_next_hop_update_rte(rtable *tab UNUSED, rte *old)
 {
   rta a;
-  memcpy(&a, old->attrs, sizeof(rta));
+  memcpy(&a, old->attrs, rta_size(old->attrs));
   rta_apply_hostentry(&a, old->attrs->hostentry);
   a.aflags = 0;
 
diff --git a/proto/ospf/rt.c b/proto/ospf/rt.c
index d28d463b..74f47810 100644
--- a/proto/ospf/rt.c
+++ b/proto/ospf/rt.c
@@ -37,6 +37,7 @@ static inline struct nexthop *
 new_nexthop(struct ospf_proto *p, ip_addr gw, struct iface *iface, byte weight)
 {
   struct nexthop *nh = lp_alloc(p->nhpool, sizeof(struct nexthop));
+  nh->labels = 0;
   nh->gw = gw;
   nh->iface = iface;
   nh->next = NULL;
diff --git a/proto/pipe/pipe.c b/proto/pipe/pipe.c
index 8924c200..a4d371fe 100644
--- a/proto/pipe/pipe.c
+++ b/proto/pipe/pipe.c
@@ -43,6 +43,8 @@
 
 #include "pipe.h"
 
+#include <alloca.h>
+
 static void
 pipe_rt_notify(struct proto *P, struct channel *src_ch, net *n, rte *new, rte *old, ea_list *attrs)
 {
@@ -51,7 +53,7 @@ pipe_rt_notify(struct proto *P, struct channel *src_ch, net *n, rte *new, rte *o
   struct rte_src *src;
 
   rte *e;
-  rta a;
+  rta *a;
 
   if (!new && !old)
     return;
@@ -65,12 +67,13 @@ pipe_rt_notify(struct proto *P, struct channel *src_ch, net *n, rte *new, rte *o
 
   if (new)
     {
-      memcpy(&a, new->attrs, sizeof(rta));
+      a = alloca(rta_size(new->attrs));
+      memcpy(a, new->attrs, rta_size(new->attrs));
 
-      a.aflags = 0;
-      a.eattrs = attrs;
-      a.hostentry = NULL;
-      e = rte_get_temp(&a);
+      a->aflags = 0;
+      a->eattrs = attrs;
+      a->hostentry = NULL;
+      e = rte_get_temp(a);
       e->pflags = 0;
 
       /* Copy protocol specific embedded attributes. */
@@ -78,7 +81,7 @@ pipe_rt_notify(struct proto *P, struct channel *src_ch, net *n, rte *new, rte *o
       e->pref = new->pref;
       e->pflags = new->pflags;
 
-      src = a.src;
+      src = a->src;
     }
   else
     {
diff --git a/proto/static/static.c b/proto/static/static.c
index a63d4d29..9ecf6df8 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -88,6 +88,7 @@ static_install(struct proto *p, struct static_route *r)
 	    nh->gw = r2->via;
 	    nh->iface = r2->neigh->iface;
 	    nh->weight = r2->weight;
+	    nh->labels = 0;
 	    if (a.nh.next)
 	      nexthop_insert(&(a.nh), nh);
 	    r2->state |= STS_INSTALLED;
-- 
cgit v1.2.3


From f2010f9c65ca69584c34c762fb3e5e957958478e Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Mon, 13 Jun 2016 15:49:53 +0200
Subject: Static: Protocol rework wrt. struct nexthop changes; MPLS label
 support

---
 conf/confbase.Y       |  18 +++++
 doc/bird.sgml         |  14 ++--
 nest/route.h          |   2 +
 nest/rt-attr.c        |   9 +++
 nest/rt-table.c       |  28 ++++---
 proto/static/config.Y |  95 +++++++++++-----------
 proto/static/static.c | 213 +++++++++++++++++++++++++++++++++++---------------
 proto/static/static.h |  17 ++--
 8 files changed, 259 insertions(+), 137 deletions(-)

(limited to 'nest/route.h')

diff --git a/conf/confbase.Y b/conf/confbase.Y
index aec4aeb4..11393fe2 100644
--- a/conf/confbase.Y
+++ b/conf/confbase.Y
@@ -64,6 +64,7 @@ CF_DECLS
   struct proto_spec ps;
   struct channel_limit cl;
   struct timeformat *tf;
+  u32 *lbl;
 }
 
 %token END CLI_MARKER INVALID_TOKEN ELSECOL DDOT
@@ -82,6 +83,7 @@ CF_DECLS
 %type <a> ipa
 %type <net> net_ip4_ net_ip6_ net_ip6 net_ip_ net_ip net_or_ipa
 %type <net_ptr> net_ net_any net_roa4_ net_roa6_ net_roa_
+%type <lbl> label_stack_start label_stack
 
 %type <t> text opttext
 
@@ -266,6 +268,22 @@ net_or_ipa:
    }
  ;
 
+label_stack_start: NUM
+{
+  $$ = cfg_allocz(sizeof(u32) * (NEXTHOP_MAX_LABEL_STACK+1));
+  $$[0] = 1;
+  $$[1] = $1;
+};
+
+label_stack:
+    label_stack_start
+  | label_stack '/' NUM {
+    if ($1[0] >= NEXTHOP_MAX_LABEL_STACK)
+      cf_error("Too many labels in stack.");
+    $1[++$1[0]] = $3;
+    $$ = $1;
+  }
+;
 
 datetime:
    TEXT {
diff --git a/doc/bird.sgml b/doc/bird.sgml
index 999fa294..23026eae 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -4141,8 +4141,8 @@ specific destination for them and you don't want to send them out through the
 default route to prevent routing loops).
 
 <p>There are five types of static routes: `classical' routes telling to forward
-packets to a neighboring router, multipath routes specifying several (possibly
-weighted) neighboring routers, device routes specifying forwarding to hosts on a
+packets to a neighboring router (single path or multipath, possibly weighted),
+device routes specifying forwarding to hosts on a
 directly connected network, recursive routes computing their nexthops by doing
 route table lookups for a given IP, and special routes (sink, blackhole etc.)
 which specify a special action to be done instead of forwarding the packet.
@@ -4174,14 +4174,14 @@ definition of the protocol contains mainly a list of static routes.
 <p>Route definitions (each may also contain a block of per-route options):
 
 <descrip>
-	<tag><label id="static-route-via-ip">route <m/prefix/ via <m/ip/</tag>
-	Static route through a neighboring router. For link-local next hops,
+	<tag><label id="static-route-via-ip">route <m/prefix/ via <m/ip/ [mpls <m/num/[/<m/num/[/<m/num/[...]]]]</tag>
+	Static single path route through a neighboring router. For link-local next hops,
 	interface can be specified as a part of the address (e.g.,
-	<cf/via fe80::1234%eth0/).
+	<cf/via fe80::1234%eth0/). MPLS labels should be specified in outer-first order.
 
-	<tag><label id="static-route-via-mpath">route <m/prefix/ multipath via <m/ip/ [weight <m/num/] [bfd <m/switch/] [via <m/.../]</tag>
+	<tag><label id="static-route-via-mpath">route <m/prefix/ via <m/ip/ [mpls <m/num/[/<m/num/[/<m/num/[...]]]] [weight <m/num/] [bfd <m/switch/] [via ...]</tag>
 	Static multipath route. Contains several nexthops (gateways), possibly
-	with their weights.
+	with their weights and MPLS labels.
 
 	<tag><label id="static-route-via-iface">route <m/prefix/ via <m/"interface"/</tag>
 	Static device route through an interface to hosts on a directly
diff --git a/nest/route.h b/nest/route.h
index d9a1737b..c293fbb1 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -513,6 +513,7 @@ ea_list *ea_append(ea_list *to, ea_list *what);
 void ea_format_bitfield(struct eattr *a, byte *buf, int bufsize, const char **names, int min, int max);
 
 #define NEXTHOP_MAX_LABEL_STACK 8
+#define NEXTHOP_MAX_SIZE (sizeof(struct nexthop) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK)
 
 static inline size_t nexthop_size(const struct nexthop *nh)
 { return sizeof(struct nexthop) + sizeof(u32)*nh->labels; }
@@ -527,6 +528,7 @@ int nexthop_is_sorted(struct nexthop *x);
 
 void rta_init(void);
 static inline size_t rta_size(const rta *a) { return sizeof(rta) + sizeof(u32)*a->nh.labels; }
+#define RTA_MAX_SIZE (sizeof(rta) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK)
 rta *rta_lookup(rta *);			/* Get rta equivalent to this one, uc++ */
 static inline int rta_is_cached(rta *r) { return r->aflags & RTAF_CACHED; }
 static inline rta *rta_clone(rta *r) { r->uc++; return r; }
diff --git a/nest/rt-attr.c b/nest/rt-attr.c
index d3671a53..22a5cb14 100644
--- a/nest/rt-attr.c
+++ b/nest/rt-attr.c
@@ -330,6 +330,9 @@ nexthop_copy(struct nexthop *o)
       n->iface = o->iface;
       n->next = NULL;
       n->weight = o->weight;
+      n->labels = o->labels;
+      for (int i=0; i<o->labels; i++)
+	n->label[i] = o->label[i];
 
       *last = n;
       last = &(n->next);
@@ -1176,6 +1179,12 @@ rta_do_cow(rta *o, linpool *lp)
 {
   rta *r = lp_alloc(lp, rta_size(o));
   memcpy(r, o, rta_size(o));
+  for (struct nexthop **nhn = &(r->nh.next), *nho = o->nh.next; nho; nho = nho->next)
+    {
+      *nhn = lp_alloc(lp, nexthop_size(nho));
+      memcpy(*nhn, nho, nexthop_size(nho));
+      nhn = &((*nhn)->next);
+    }
   r->aflags = 0;
   r->uc = 0;
   return r;
diff --git a/nest/rt-table.c b/nest/rt-table.c
index a7ceddb8..92542ea8 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -2443,14 +2443,7 @@ rt_format_via(rte *e)
 
   switch (a->dest)
     {
-    case RTD_UNICAST:	if (a->nh.next)
-			  bsprintf(via, "multipath");
-			else
-			  {
-			    if (ipa_nonzero(a->nh.gw)) bsprintf(via, "via %I ", a->nh.gw);
-			    bsprintf(via, "dev %s", a->nh.iface->name);
-			  }
-			break;
+    case RTD_UNICAST:	bsprintf(via, "unicast"); break;
     case RTD_BLACKHOLE:	bsprintf(via, "blackhole"); break;
     case RTD_UNREACHABLE:	bsprintf(via, "unreachable"); break;
     case RTD_PROHIBIT:	bsprintf(via, "prohibited"); break;
@@ -2492,11 +2485,24 @@ rt_show_rte(struct cli *c, byte *ia, rte *e, struct rt_show_data *d, ea_list *tm
     bsprintf(info, " (%d)", e->pref);
   cli_printf(c, -1007, "%-18s %s [%s %s%s]%s%s", ia, rt_format_via(e), a->src->proto->name,
 	     tm, from, primary ? (sync_error ? " !" : " *") : "", info);
-  if (a->nh.next)
-    for (nh = &(a->nh); nh; nh = nh->next)
-      cli_printf(c, -1007, "\tvia %I on %s weight %d", nh->gw, nh->iface->name, nh->weight + 1);
+  for (nh = &(a->nh); nh; nh = nh->next)
+    {
+      char ls[NEXTHOP_MAX_LABEL_STACK*8 + 5]; char *lsp = ls;
+      if (nh->labels)
+	{
+	  lsp += bsprintf(lsp, " mpls %d", nh->label[0]);
+	  for (int i=1;i<nh->labels; i++)
+	    lsp += bsprintf(lsp, "/%d", nh->label[i]);
+	  *lsp++ = '\0';
+	}
+      if (a->nh.next)
+	cli_printf(c, -1007, "\tvia %I%s on %s weight %d", nh->gw, (nh->labels ? ls : ""), nh->iface->name, nh->weight + 1);
+      else
+	cli_printf(c, -1007, "\tvia %I%s on %s", nh->gw, (nh->labels ? ls : ""), nh->iface->name);
+    }
   if (d->verbose)
     rta_show(c, a, tmpa);
+
 }
 
 static void
diff --git a/proto/static/config.Y b/proto/static/config.Y
index 8103166d..a461bd27 100644
--- a/proto/static/config.Y
+++ b/proto/static/config.Y
@@ -18,19 +18,12 @@ static struct f_inst **this_srt_last_cmd;
 
 static void
 static_route_finish(void)
-{
-  struct static_route *r;
-
-  /* Update undefined use_bfd entries in multipath nexthops */
-  for (r = this_srt->mp_next; r; r = r->mp_next)
-    if (r->use_bfd < 0)
-      r->use_bfd = this_srt->use_bfd;
-}
+{ }
 
 CF_DECLS
 
 CF_KEYWORDS(STATIC, ROUTE, VIA, DROP, REJECT, PROHIBIT, PREFERENCE, CHECK, LINK)
-CF_KEYWORDS(MULTIPATH, WEIGHT, RECURSIVE, IGP, TABLE, BLACKHOLE, UNREACHABLE, BFD)
+CF_KEYWORDS(WEIGHT, RECURSIVE, IGP, TABLE, BLACKHOLE, UNREACHABLE, BFD, MPLS)
 
 
 CF_GRAMMAR
@@ -52,6 +45,50 @@ static_proto:
  | static_proto stat_route stat_route_opt_list ';' { static_route_finish(); }
  ;
 
+stat_nexthop_via: VIA
+{
+  if (last_srt)
+  {
+    last_srt = (last_srt->mp_next = cfg_allocz(sizeof(struct static_route)));
+    last_srt->net = this_srt->net;
+  }
+  else
+  {
+    last_srt = this_srt;
+    rem_node(&this_srt->n);
+  }
+
+  last_srt->mp_head = this_srt;
+  last_srt->dest = RTD_UNICAST;
+};
+
+stat_nexthop_ident:
+    stat_nexthop_via ipa ipa_scope {
+      last_srt->via = $2;
+      last_srt->iface = $3;
+      add_tail(&STATIC_CFG->neigh_routes, &last_srt->n);
+    }
+  | stat_nexthop_via TEXT {
+      last_srt->via = IPA_NONE;
+      last_srt->if_name = $2;
+      add_tail(&STATIC_CFG->iface_routes, &last_srt->n);
+    }
+  | stat_nexthop_ident MPLS label_stack {
+    last_srt->label_count = $3[0];
+    last_srt->label_stack = &($3[1]);
+  }
+  | stat_nexthop_ident WEIGHT expr {
+    last_srt->weight = $3 - 1;
+    if (($3<1) || ($3>256)) cf_error("Weight must be in range 1-256");
+  }
+  | stat_nexthop_ident BFD bool { last_srt->use_bfd = $3; cf_check_bfd($3); }
+;
+
+stat_nexthop:
+    stat_nexthop_ident
+  | stat_nexthop stat_nexthop_ident
+;
+
 stat_route0: ROUTE net_any {
      this_srt = cfg_allocz(sizeof(struct static_route));
      add_tail(&STATIC_CFG->other_routes, &this_srt->n);
@@ -62,45 +99,8 @@ stat_route0: ROUTE net_any {
   }
  ;
 
-stat_multipath1:
-   VIA ipa ipa_scope {
-     last_srt = last_srt ? last_srt->mp_next = cfg_allocz(sizeof(struct static_route)) : this_srt;
-
-     last_srt->dest = RTD_UNICAST;
-     last_srt->via = $2;
-     last_srt->via_if = $3;
-     last_srt->if_name = (void *) this_srt; /* really */
-     last_srt->use_bfd = -1; /* undefined */
-     last_srt->mp_next = NULL;
-   }
- | stat_multipath1 WEIGHT expr {
-     last_srt->weight = $3 - 1;
-     if (($3<1) || ($3>256)) cf_error("Weight must be in range 1-256"); 
-   }
- | stat_multipath1 BFD bool {
-     last_srt->use_bfd = $3; cf_check_bfd($3);
-   }
- ;
-
-stat_multipath:
-   stat_multipath1
- | stat_multipath stat_multipath1
- ;
-
 stat_route:
-   stat_route0 VIA ipa ipa_scope {
-      this_srt->dest = RTD_UNICAST;
-      this_srt->via = $3;
-      this_srt->via_if = $4;
-   }
- | stat_route0 VIA TEXT {
-      this_srt->dest = RTD_UNICAST;
-      this_srt->via = IPA_NONE;
-      this_srt->if_name = $3;
-      rem_node(&this_srt->n);
-      add_tail(&STATIC_CFG->iface_routes, &this_srt->n);
-   }
- | stat_route0 MULTIPATH stat_multipath
+   stat_route0 stat_nexthop
  | stat_route0 RECURSIVE ipa {
       this_srt->dest = RTDX_RECURSIVE;
       this_srt->via = $3;
@@ -115,7 +115,6 @@ stat_route:
 
 stat_route_item:
    cmd { *this_srt_last_cmd = $1; this_srt_last_cmd = &($1->next); }
- | BFD bool ';' { this_srt->use_bfd = $2; cf_check_bfd($2); }
  ;
 
 stat_route_opts:
diff --git a/proto/static/static.c b/proto/static/static.c
index 9ecf6df8..878248c1 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -60,37 +60,51 @@ p_igp_table(struct proto *p)
 static void
 static_install(struct proto *p, struct static_route *r)
 {
-  rta a;
+  rta *ap = alloca(RTA_MAX_SIZE);
   rte *e;
 
-  if (!(r->state & STS_WANT) && r->dest != RTD_UNICAST)
-    return;
+  if (!(r->state & STS_WANT) && (r->state & (STS_INSTALLED | STS_FORCE)) && r->dest != RTD_UNICAST)
+    goto drop;
 
   DBG("Installing static route %N, rtd=%d\n", r->net, r->dest);
-  bzero(&a, sizeof(a));
-  a.src = p->main_source;
-  a.source = ((r->dest == RTD_UNICAST) && ipa_zero(r->via)) ? RTS_STATIC_DEVICE : RTS_STATIC;
-  a.scope = SCOPE_UNIVERSE;
-  a.dest = r->dest;
+  bzero(ap, RTA_MAX_SIZE);
+  ap->src = p->main_source;
+  ap->source = ((r->dest == RTD_UNICAST) && ipa_zero(r->via)) ? RTS_STATIC_DEVICE : RTS_STATIC;
+  ap->scope = SCOPE_UNIVERSE;
+  ap->dest = r->dest;
+
   if (r->dest == RTD_UNICAST)
     {
       struct static_route *r2;
-      int num = 0;
+      int num = 0, update = 0;
 
       for (r2 = r; r2; r2 = r2->mp_next)
       {
-	if ((r2->state & STS_INSTALLED) && !(r2->state & STS_FORCE))
-	  continue;
+
+	if ((r2->state & STS_FORCE) ||
+	    (!!(r2->state & STS_INSTALLED) != !!(r2->state & STS_WANT)))
+	  update++;
 
 	if (r2->state & STS_WANT)
 	  {
-	    struct nexthop *nh = (a.nh.next) ? alloca(sizeof(struct nexthop)) : &(a.nh);
-	    nh->gw = r2->via;
-	    nh->iface = r2->neigh->iface;
+	    struct nexthop *nh = (ap->nh.next) ? alloca(NEXTHOP_MAX_SIZE) : &(ap->nh);
+	    if (ipa_zero(r2->via)) // Device nexthop
+	      {
+		nh->gw = IPA_NONE;
+		nh->iface = r2->iface;
+	      }
+	    else // Router nexthop
+	      {
+		nh->gw = r2->via;
+		nh->iface = r2->neigh->iface;
+	      }
 	    nh->weight = r2->weight;
-	    nh->labels = 0;
-	    if (a.nh.next)
-	      nexthop_insert(&(a.nh), nh);
+	    nh->labels = r2->label_count;
+	    for (int i=0; i<nh->labels; i++)
+	      nh->label[i] = r2->label_stack[i];
+
+	    if (ap->nh.next)
+	      nexthop_insert(&(ap->nh), nh);
 	    r2->state |= STS_INSTALLED;
 	    num++;
 	  }
@@ -98,21 +112,27 @@ static_install(struct proto *p, struct static_route *r)
 	  r2->state = 0;
       }
 
+      if (!update) // Nothing changed
+	return;
+
+      r = r->mp_head;
+
       if (!num) // No nexthop to install
       {
-	if (r->state & STS_INSTALLED_ANY)
-	  rte_update(p, r->net, NULL);
-	
+drop:
+	rte_update(p, r->net, NULL);
 	return;
       }
     }
-
+  else
+    r->state |= STS_INSTALLED;
+  
   if (r->dest == RTDX_RECURSIVE)
-    rta_set_recursive_next_hop(p->main_channel->table, &a, p_igp_table(p), r->via, IPA_NONE);
+    rta_set_recursive_next_hop(p->main_channel->table, ap, p_igp_table(p), r->via, IPA_NONE);
 
   /* We skip rta_lookup() here */
 
-  e = rte_get_temp(&a);
+  e = rte_get_temp(ap);
   e->pflags = 0;
 
   if (r->cmds)
@@ -124,18 +144,6 @@ static_install(struct proto *p, struct static_route *r)
     lp_flush(static_lp);
 }
 
-static void
-static_remove(struct proto *p, struct static_route *r)
-{
-  if (!(r->state & STS_INSTALLED_ANY))
-    return;
-
-  DBG("Removing static route %N via %I\n", r->net, r->via);
-  rte_update(p, r->net, NULL);
-
-  r->state &= ~(STS_INSTALLED | STS_INSTALLED_ANY);
-}
-
 static void
 static_bfd_notify(struct bfd_request *req);
 
@@ -165,6 +173,8 @@ static_decide(struct static_config *cf, struct static_route *r)
   /* r->dest != RTD_MULTIPATH, but may be RTD_NONE (part of multipath route)
      the route also have to be valid (r->neigh != NULL) */
 
+  r->state &= ~STS_WANT;
+
   if (r->neigh->scope < 0)
     return 0;
 
@@ -174,6 +184,7 @@ static_decide(struct static_config *cf, struct static_route *r)
   if (r->bfd_req && r->bfd_req->state != BFD_STATE_UP)
     return 0;
 
+  r->state |= STS_WANT;
   return 1;
 }
 
@@ -181,6 +192,9 @@ static_decide(struct static_config *cf, struct static_route *r)
 static void
 static_add(struct proto *p, struct static_config *cf, struct static_route *r)
 {
+  if (r->mp_head && r != r->mp_head)
+    return;
+
   DBG("static_add(%N,%d)\n", r->net, r->dest);
   switch (r->dest)
     {
@@ -191,7 +205,10 @@ static_add(struct proto *p, struct static_config *cf, struct static_route *r)
 
 	for (r2 = r; r2; r2 = r2->mp_next)
 	  {
-	    struct neighbor *n = neigh_find2(p, &r2->via, r2->via_if, NEF_STICKY);
+	    if (ipa_zero(r2->via)) // No struct neighbor for device routes
+	      continue;
+
+	    struct neighbor *n = neigh_find2(p, &r2->via, r2->iface, NEF_STICKY);
 	    if (n)
 	      {
 		r2->chain = n->data;
@@ -199,7 +216,7 @@ static_add(struct proto *p, struct static_config *cf, struct static_route *r)
 		r2->neigh = n;
 
 		static_update_bfd(p, r2);
-		r2->state = static_decide(cf,r2) ? STS_WANT : 0;
+		static_decide(cf,r2);
 		count++;
 	      }
 	    else
@@ -223,6 +240,9 @@ static_add(struct proto *p, struct static_config *cf, struct static_route *r)
 static void
 static_rte_cleanup(struct proto *p UNUSED, struct static_route *r)
 {
+  if (r->mp_head && (r != r->mp_head))
+    return;
+
   struct static_route *r2;
   
   for (r2 = r; r2; r2 = r2->mp_next)
@@ -250,8 +270,15 @@ static_start(struct proto *p)
   /* We have to go UP before routes could be installed */
   proto_notify_state(p, PS_UP);
 
-  WALK_LIST(r, cf->other_routes)
+  WALK_LIST(r, cf->neigh_routes)
     static_add(p, cf, r);
+
+  WALK_LIST(r, cf->iface_routes)
+    static_add(p, cf, r);
+
+  WALK_LIST(r, cf->other_routes)
+    static_install(p, r);
+
   return PS_UP;
 }
 
@@ -262,9 +289,14 @@ static_shutdown(struct proto *p)
   struct static_route *r;
 
   /* Just reset the flag, the routes will be flushed by the nest */
+  WALK_LIST(r, cf->other_routes)
+  {
+    static_rte_cleanup(p, r);
+    r->state = 0;
+  }
   WALK_LIST(r, cf->iface_routes)
     r->state = 0;
-  WALK_LIST(r, cf->other_routes)
+  WALK_LIST(r, cf->neigh_routes)
   {
     static_rte_cleanup(p, r);
     r->state = 0;
@@ -275,9 +307,11 @@ static_shutdown(struct proto *p)
   cf = (void *) p->cf_new;
   if (cf)
   {
+    WALK_LIST(r, cf->other_routes)
+      r->state = 0;
     WALK_LIST(r, cf->iface_routes)
       r->state = 0;
-    WALK_LIST(r, cf->other_routes)
+    WALK_LIST(r, cf->neigh_routes)
       r->state = 0;
   }
 
@@ -299,10 +333,8 @@ static_update_rte(struct proto *p, struct static_route *r)
   if (r->dest != RTD_UNICAST)
     return;
 
-  if (static_decide((struct static_config *) p->cf, r))
-    static_install(p, r);
-  else
-    static_remove(p, r);
+  static_decide((struct static_config *) p->cf, r);
+  static_install(p, r);
 }
 
 static void
@@ -349,12 +381,15 @@ static_dump(struct proto *p)
   struct static_config *c = (void *) p->cf;
   struct static_route *r;
 
-  debug("Independent static routes:\n");
-  WALK_LIST(r, c->other_routes)
+  debug("Independent static nexthops:\n");
+  WALK_LIST(r, c->neigh_routes)
     static_dump_rt(r);
-  debug("Device static routes:\n");
+  debug("Device static nexthops:\n");
   WALK_LIST(r, c->iface_routes)
     static_dump_rt(r);
+  debug("Other static routes:\n");
+  WALK_LIST(r, c->other_routes)
+    static_dump_rt(r);
 }
 
 static void
@@ -367,13 +402,21 @@ static_if_notify(struct proto *p, unsigned flags, struct iface *i)
     {
       WALK_LIST(r, c->iface_routes)
 	if (!strcmp(r->if_name, i->name))
+	{
+	  r->state |= STS_WANT;
+	  r->iface = i;
 	  static_install(p, r);
+	}
     }
   else if (flags & IF_CHANGE_DOWN)
     {
       WALK_LIST(r, c->iface_routes)
 	if (!strcmp(r->if_name, i->name))
-	  static_remove(p, r);
+	{
+	  r->state &= ~STS_WANT;
+	  r->iface = NULL;
+	  static_install(p, r);
+	}
     }
 }
 
@@ -386,6 +429,7 @@ static_rte_mergable(rte *pri UNUSED, rte *sec UNUSED)
 void
 static_init_config(struct static_config *c)
 {
+  init_list(&c->neigh_routes);
   init_list(&c->iface_routes);
   init_list(&c->other_routes);
 }
@@ -400,8 +444,12 @@ static_postconfig(struct proto_config *CF)
     cf_error("Channel not specified");
 
 
+  WALK_LIST(r, cf->neigh_routes)
+    if (r->net && (r->net->type != CF->net_type))
+      cf_error("Route %N incompatible with channel type", r->net);
+
   WALK_LIST(r, cf->iface_routes)
-    if (r->net->type != CF->net_type)
+    if (r->net && (r->net->type != CF->net_type))
       cf_error("Route %N incompatible with channel type", r->net);
 
   WALK_LIST(r, cf->other_routes)
@@ -442,16 +490,22 @@ static_same_dest(struct static_route *x, struct static_route *y)
 	  if (ipa_nonzero(xc->via) && ipa_nonzero(yc->via))
 	  {
 	    if (!ipa_equal(x->via, y->via) ||
-		(x->via_if != y->via_if) ||
+		(x->iface != y->iface) ||
 		(x->use_bfd != y->use_bfd) ||
-		(x->weight != y->weight))
+		(x->weight != y->weight) ||
+		(x->label_count != y->label_count))
 	      return 0;
+	    for (int i=0; i<x->label_count; i++)
+	      if (x->label_stack[i] != y->label_stack[i])
+		return 0;
 	  }
 	  else
-	    if (strcmp(x->if_name, y->if_name) ||
+	    if ((!x->if_name) || (!y->if_name) ||
+		strcmp(x->if_name, y->if_name) ||
 		(x->use_bfd != y->use_bfd) ||
 		(x->weight != y->weight))
 	      return 0;
+
 	}
 	return 1;
       }
@@ -476,6 +530,9 @@ static_match(struct proto *p, struct static_route *r, struct static_config *n)
 {
   struct static_route *t;
 
+  if (r->mp_head && (r->mp_head != r))
+    return;
+
   /*
    * For given old route *r we find whether a route to the same
    * network is also in the new route list. In that case, we keep the
@@ -486,23 +543,28 @@ static_match(struct proto *p, struct static_route *r, struct static_config *n)
   if (r->neigh)
     r->neigh->data = NULL;
 
+  WALK_LIST(t, n->neigh_routes)
+    if ((!t->mp_head || (t->mp_head == t)) && net_equal(r->net, t->net))
+      goto found;
+
   WALK_LIST(t, n->iface_routes)
-    if (net_equal(r->net, t->net))
+    if ((!t->mp_head || (t->mp_head == t)) && net_equal(r->net, t->net))
       goto found;
 
   WALK_LIST(t, n->other_routes)
     if (net_equal(r->net, t->net))
       goto found;
 
-  static_remove(p, r);
+  r->state &= ~STS_WANT;
+  static_install(p, r);
   return;
 
  found:
+  t->state = r->state;
+
   /* If destination is different, force reinstall */
-  if (r->state && !static_same_rte(r, t))
-    t->state = r->state | STS_WANT | STS_FORCE;
-  else
-    t->state = r->state;
+  if (!static_same_rte(r, t))
+    t->state |= STS_FORCE;
 }
 
 static inline rtable *
@@ -525,20 +587,34 @@ static_reconfigure(struct proto *p, struct proto_config *CF)
     return 0;
 
   /* Delete all obsolete routes and reset neighbor entries */
+  WALK_LIST(r, o->other_routes)
+    static_match(p, r, n);
   WALK_LIST(r, o->iface_routes)
     static_match(p, r, n);
-  WALK_LIST(r, o->other_routes)
+  WALK_LIST(r, o->neigh_routes)
     static_match(p, r, n);
 
   /* Now add all new routes, those not changed will be ignored by static_install() */
+  WALK_LIST(r, n->neigh_routes)
+    static_add(p, n, r);
+  WALK_LIST(r, o->neigh_routes)
+    static_rte_cleanup(p, r);
+
   WALK_LIST(r, n->iface_routes)
     {
       struct iface *ifa;
       if ((ifa = if_find_by_name(r->if_name)) && (ifa->flags & IF_UP))
-	static_install(p, r);
+	{
+	  r->iface = ifa;
+	  static_install(p, r);
+	}
     }
+
   WALK_LIST(r, n->other_routes)
-    static_add(p, n, r);
+  {
+    r->state |= STS_WANT;
+    static_install(p, r);
+  }
 
   WALK_LIST(r, o->other_routes)
     static_rte_cleanup(p, r);
@@ -555,7 +631,7 @@ static_copy_routes(list *dlst, list *slst)
   WALK_LIST(sr, *slst)
     {
       struct static_route *srr, *drr = NULL;
-      for (srr = sr; srr; srr = srr->mp_next)
+      for (srr = sr->mp_head; srr; srr = srr->mp_next)
       {
 	/* copy one route */
 	struct static_route *dr = cfg_alloc(sizeof(struct static_route));
@@ -577,6 +653,7 @@ static_copy_config(struct proto_config *dest, struct proto_config *src)
   struct static_config *s = (struct static_config *) src;
 
   /* Copy route lists */
+  static_copy_routes(&d->neigh_routes, &s->neigh_routes);
   static_copy_routes(&d->iface_routes, &s->iface_routes);
   static_copy_routes(&d->other_routes, &s->other_routes);
 }
@@ -606,7 +683,7 @@ static_format_via(struct static_route *r)
   switch (r->dest)
     {
     case RTD_UNICAST:	if (ipa_zero(r->via)) bsprintf(via, "dev %s", r->if_name);
-			else bsprintf(via, "via %I%J", r->via, r->via_if);
+			else bsprintf(via, "via %I%J", r->via, r->iface);
 			break;
     case RTD_BLACKHOLE:	bsprintf(via, "blackhole"); break;
     case RTD_UNREACHABLE: bsprintf(via, "unreachable"); break;
@@ -620,13 +697,19 @@ static_format_via(struct static_route *r)
 static void
 static_show_rt(struct static_route *r)
 {
+  if (r->mp_head && (r != r->mp_head))
+    return;
   if (r->mp_next)
   {
     cli_msg(-1009, "%N", r->net);
     struct static_route *r2;
     for (r2 = r; r2; r2 = r2->mp_next)
+    {
       cli_msg(-1009, "\t%s weight %d%s%s", static_format_via(r2), r2->weight + 1,
 	      r2->bfd_req ? " (bfd)" : "", (r2->state & STS_INSTALLED) ? "" : " (dormant)");
+      if (r2->mp_next == r)
+	break;
+    }
   }
   else
     cli_msg(-1009, "%N %s%s%s", r->net, static_format_via(r),
@@ -639,9 +722,11 @@ static_show(struct proto *P)
   struct static_config *c = (void *) P->cf;
   struct static_route *r;
 
-  WALK_LIST(r, c->other_routes)
+  WALK_LIST(r, c->neigh_routes)
     static_show_rt(r);
   WALK_LIST(r, c->iface_routes)
     static_show_rt(r);
+  WALK_LIST(r, c->other_routes)
+    static_show_rt(r);
   cli_msg(0, "");
 }
diff --git a/proto/static/static.h b/proto/static/static.h
index 418b3076..aeb9660a 100644
--- a/proto/static/static.h
+++ b/proto/static/static.h
@@ -15,7 +15,8 @@
 struct static_config {
   struct proto_config c;
   list iface_routes;		/* Routes to search on interface events */
-  list other_routes;		/* Routes hooked to neighbor cache and reject routes */
+  list neigh_routes;		/* Routes to search on neighbor events */
+  list other_routes;		/* Non-nexthop routes */
   int check_link;			/* Whether iface link state is used */
   struct rtable_config *igp_table;	/* Table used for recursive next hop lookups */
 };
@@ -29,21 +30,23 @@ struct static_route {
   net_addr *net;			/* Network we route */
   int dest;				/* Destination type (RTD_*) */
   ip_addr via;				/* Destination router */
-  struct iface *via_if;			/* Destination iface, for link-local vias */
+  struct iface *iface;			/* Destination iface, for link-local vias or device routes */
   struct neighbor *neigh;
-  byte *if_name;			/* Name for RTD_DEVICE routes */
+  byte *if_name;			/* Name for device routes */
   struct static_route *mp_next;		/* Nexthops for multipath routes */
+  struct static_route *mp_head;		/* First nexthop of this route */
   struct f_inst *cmds;			/* List of commands for setting attributes */
   u32 state;				/* Current state: STS_* */
-  int use_bfd;				/* Configured to use BFD */
   int weight;				/* Multipath next hop weight */
+  byte use_bfd;				/* Configured to use BFD */
+  byte label_count;			/* Number of labels in stack */
   struct bfd_request *bfd_req;		/* BFD request, if BFD is used */
+  u32 *label_stack;			/* Label stack if label_count > 0 */
 };
 
 #define STS_INSTALLED		0x1
-#define STS_INSTALLED_ANY	0x2
-#define STS_WANT		0x4
-#define STS_FORCE		0x8
+#define STS_WANT		0x2
+#define STS_FORCE		0x4
 
 /* Dummy nodes (parts of multipath route) abuses masklen field for weight
    and if_name field for a ptr to the master (RTD_MULTIPATH) node. */
-- 
cgit v1.2.3


From d14f8c3c45f7e33a7e54ebc0d45bdb0295d70301 Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Fri, 4 Mar 2016 12:55:50 +0100
Subject: Netlink: MPLS routes in kernel

Anyway, Bird is now capable to insert both MPLS routes and MPLS encap
routes into kernel.

It was (among others) needed to define platform-specific AF_MPLS to 28
as this constant has been assigned in the linux kernel.

No support for BSD now, it may be added in the future.
---
 conf/confbase.Y        |   4 +-
 lib/ip.h               |  22 +++++
 lib/net.c              |   3 +-
 lib/net.h              |   8 ++
 nest/route.h           |   5 +-
 nest/rt-attr.c         |   4 +-
 nest/rt-table.c        |   2 +-
 sysdep/cf/linux.h      |   4 +
 sysdep/linux/netlink.c | 247 +++++++++++++++++++++++++++++++++++++++++++++----
 sysdep/unix/krt.c      |   5 +-
 10 files changed, 274 insertions(+), 30 deletions(-)

(limited to 'nest/route.h')

diff --git a/conf/confbase.Y b/conf/confbase.Y
index 11393fe2..00e78232 100644
--- a/conf/confbase.Y
+++ b/conf/confbase.Y
@@ -270,7 +270,7 @@ net_or_ipa:
 
 label_stack_start: NUM
 {
-  $$ = cfg_allocz(sizeof(u32) * (NEXTHOP_MAX_LABEL_STACK+1));
+  $$ = cfg_allocz(sizeof(u32) * (MPLS_MAX_LABEL_STACK+1));
   $$[0] = 1;
   $$[1] = $1;
 };
@@ -278,7 +278,7 @@ label_stack_start: NUM
 label_stack:
     label_stack_start
   | label_stack '/' NUM {
-    if ($1[0] >= NEXTHOP_MAX_LABEL_STACK)
+    if ($1[0] >= MPLS_MAX_LABEL_STACK)
       cf_error("Too many labels in stack.");
     $1[++$1[0]] = $3;
     $$ = $1;
diff --git a/lib/ip.h b/lib/ip.h
index 86750675..ab90bee7 100644
--- a/lib/ip.h
+++ b/lib/ip.h
@@ -325,6 +325,28 @@ static inline ip6_addr ip6_hton(ip6_addr a)
 static inline ip6_addr ip6_ntoh(ip6_addr a)
 { return _MI6(ntohl(_I0(a)), ntohl(_I1(a)), ntohl(_I2(a)), ntohl(_I3(a))); }
 
+#define MPLS_MAX_LABEL_STACK 8
+static inline int
+mpls_get(const char *buf, int buflen, u32 *stack)
+{
+  for (int i=0; (i<MPLS_MAX_LABEL_STACK) && (i*4+3 < buflen); i++)
+  {
+    u32 s = get_u32(buf + i*4);
+    stack[i] = s >> 12;
+    if (s & 0x100)
+      return i+1;
+  }
+  return -1;
+}
+
+static inline int
+mpls_put(char *buf, int len, u32 *stack)
+{
+  for (int i=0; i<len; i++)
+    put_u32(buf + i*4, stack[i] << 12 | (i+1 == len ? 0x100 : 0));
+
+  return len*4;
+}
 
 /*
  *	Unaligned data access (in network order)
diff --git a/lib/net.c b/lib/net.c
index 91a4474a..5a7e6fec 100644
--- a/lib/net.c
+++ b/lib/net.c
@@ -225,7 +225,8 @@ net_classify(const net_addr *N)
   case NET_FLOW6:
     return ip6_zero(n->ip6.prefix) ? (IADDR_HOST | SCOPE_UNIVERSE) : ip6_classify(&n->ip6.prefix);
 
-  /* classify probably not needed for NET_MPLS */
+  case NET_MPLS:
+    return IADDR_HOST | SCOPE_UNIVERSE;
   }
 
   return IADDR_INVALID;
diff --git a/lib/net.h b/lib/net.h
index 5fc4dab7..7c124fc0 100644
--- a/lib/net.h
+++ b/lib/net.h
@@ -259,6 +259,14 @@ static inline ip_addr net_prefix(const net_addr *a)
   }
 }
 
+static inline u32 net_mpls(const net_addr *a)
+{
+  if (a->type == NET_MPLS)
+    return ((net_addr_mpls *) a)->label;
+
+  bug("Can't call net_mpls on non-mpls net_addr");
+}
+
 static inline uint net4_pxlen(const net_addr *a)
 { return a->pxlen; }
 
diff --git a/nest/route.h b/nest/route.h
index c293fbb1..a2fadf62 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -512,8 +512,7 @@ uint ea_hash(ea_list *e);	/* Calculate 16-bit hash value */
 ea_list *ea_append(ea_list *to, ea_list *what);
 void ea_format_bitfield(struct eattr *a, byte *buf, int bufsize, const char **names, int min, int max);
 
-#define NEXTHOP_MAX_LABEL_STACK 8
-#define NEXTHOP_MAX_SIZE (sizeof(struct nexthop) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK)
+#define NEXTHOP_MAX_SIZE (sizeof(struct nexthop) + sizeof(u32)*MPLS_MAX_LABEL_STACK)
 
 static inline size_t nexthop_size(const struct nexthop *nh)
 { return sizeof(struct nexthop) + sizeof(u32)*nh->labels; }
@@ -528,7 +527,7 @@ int nexthop_is_sorted(struct nexthop *x);
 
 void rta_init(void);
 static inline size_t rta_size(const rta *a) { return sizeof(rta) + sizeof(u32)*a->nh.labels; }
-#define RTA_MAX_SIZE (sizeof(rta) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK)
+#define RTA_MAX_SIZE (sizeof(rta) + sizeof(u32)*MPLS_MAX_LABEL_STACK)
 rta *rta_lookup(rta *);			/* Get rta equivalent to this one, uc++ */
 static inline int rta_is_cached(rta *r) { return r->aflags & RTAF_CACHED; }
 static inline rta *rta_clone(rta *r) { r->uc++; return r; }
diff --git a/nest/rt-attr.c b/nest/rt-attr.c
index 22a5cb14..120a8e24 100644
--- a/nest/rt-attr.c
+++ b/nest/rt-attr.c
@@ -1279,12 +1279,12 @@ rta_init(void)
   rta_slab_[0] = sl_new(rta_pool, sizeof(rta));
   rta_slab_[1] = sl_new(rta_pool, sizeof(rta) + sizeof(u32));
   rta_slab_[2] = sl_new(rta_pool, sizeof(rta) + sizeof(u32)*2);
-  rta_slab_[3] = sl_new(rta_pool, sizeof(rta) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK);
+  rta_slab_[3] = sl_new(rta_pool, sizeof(rta) + sizeof(u32)*MPLS_MAX_LABEL_STACK);
 
   nexthop_slab_[0] = sl_new(rta_pool, sizeof(struct nexthop));
   nexthop_slab_[1] = sl_new(rta_pool, sizeof(struct nexthop) + sizeof(u32));
   nexthop_slab_[2] = sl_new(rta_pool, sizeof(struct nexthop) + sizeof(u32)*2);
-  nexthop_slab_[3] = sl_new(rta_pool, sizeof(struct nexthop) + sizeof(u32)*NEXTHOP_MAX_LABEL_STACK);
+  nexthop_slab_[3] = sl_new(rta_pool, sizeof(struct nexthop) + sizeof(u32)*MPLS_MAX_LABEL_STACK);
 
   rta_alloc_hash();
   rte_src_init();
diff --git a/nest/rt-table.c b/nest/rt-table.c
index 92542ea8..73b838be 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -2487,7 +2487,7 @@ rt_show_rte(struct cli *c, byte *ia, rte *e, struct rt_show_data *d, ea_list *tm
 	     tm, from, primary ? (sync_error ? " !" : " *") : "", info);
   for (nh = &(a->nh); nh; nh = nh->next)
     {
-      char ls[NEXTHOP_MAX_LABEL_STACK*8 + 5]; char *lsp = ls;
+      char ls[MPLS_MAX_LABEL_STACK*8 + 5]; char *lsp = ls;
       if (nh->labels)
 	{
 	  lsp += bsprintf(lsp, " mpls %d", nh->label[0]);
diff --git a/sysdep/cf/linux.h b/sysdep/cf/linux.h
index cec9499c..3a3a15da 100644
--- a/sysdep/cf/linux.h
+++ b/sysdep/cf/linux.h
@@ -21,6 +21,10 @@
 #define CONFIG_INCLUDE_SYSPRIV_H "sysdep/linux/syspriv.h"
 
 
+#ifndef AF_MPLS
+#define AF_MPLS 28
+#endif
+
 /*
 Link: sysdep/linux
 Link: sysdep/unix
diff --git a/sysdep/linux/netlink.c b/sysdep/linux/netlink.c
index c9a30719..23431172 100644
--- a/sysdep/linux/netlink.c
+++ b/sysdep/linux/netlink.c
@@ -31,6 +31,7 @@
 
 #include <asm/types.h>
 #include <linux/if.h>
+#include <linux/lwtunnel.h>
 #include <linux/netlink.h>
 #include <linux/rtnetlink.h>
 
@@ -51,6 +52,21 @@
 #define RTA_TABLE  15
 #endif
 
+#ifndef RTA_VIA
+#define RTA_VIA	 18
+#endif
+
+#ifndef RTA_NEWDST
+#define RTA_NEWDST  19
+#endif
+
+#ifndef RTA_ENCAP_TYPE
+#define RTA_ENCAP_TYPE	21
+#endif
+
+#ifndef RTA_ENCAP
+#define RTA_ENCAP  22
+#endif
 
 #define krt_ecmp6(p) ((p)->af == AF_INET6)
 
@@ -313,10 +329,16 @@ static struct nl_want_attrs ifa_attr_want6[BIRD_IFA_MAX] = {
 };
 
 
-#define BIRD_RTA_MAX  (RTA_TABLE+1)
+#define BIRD_RTA_MAX  (RTA_ENCAP+1)
 
 static struct nl_want_attrs nexthop_attr_want4[BIRD_RTA_MAX] = {
   [RTA_GATEWAY]	  = { 1, 1, sizeof(ip4_addr) },
+  [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) },
+  [RTA_ENCAP]	  = { 1, 0, 0 },
+};
+
+static struct nl_want_attrs encap_mpls_want[BIRD_RTA_MAX] = {
+  [RTA_DST]       = { 1, 0, 0 },
 };
 
 static struct nl_want_attrs rtm_attr_want4[BIRD_RTA_MAX] = {
@@ -329,6 +351,8 @@ static struct nl_want_attrs rtm_attr_want4[BIRD_RTA_MAX] = {
   [RTA_MULTIPATH] = { 1, 0, 0 },
   [RTA_FLOW]	  = { 1, 1, sizeof(u32) },
   [RTA_TABLE]	  = { 1, 1, sizeof(u32) },
+  [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) },
+  [RTA_ENCAP]	  = { 1, 0, 0 },
 };
 
 static struct nl_want_attrs rtm_attr_want6[BIRD_RTA_MAX] = {
@@ -341,6 +365,20 @@ static struct nl_want_attrs rtm_attr_want6[BIRD_RTA_MAX] = {
   [RTA_METRICS]	  = { 1, 0, 0 },
   [RTA_FLOW]	  = { 1, 1, sizeof(u32) },
   [RTA_TABLE]	  = { 1, 1, sizeof(u32) },
+  [RTA_ENCAP_TYPE]= { 1, 1, sizeof(u16) },
+  [RTA_ENCAP]	  = { 1, 0, 0 },
+};
+
+static struct nl_want_attrs rtm_attr_want_mpls[BIRD_RTA_MAX] = {
+  [RTA_DST]	  = { 1, 1, sizeof(u32) },
+  [RTA_IIF]	  = { 1, 1, sizeof(u32) },
+  [RTA_OIF]	  = { 1, 1, sizeof(u32) },
+  [RTA_PRIORITY]  = { 1, 1, sizeof(u32) },
+  [RTA_METRICS]	  = { 1, 0, 0 },
+  [RTA_FLOW]	  = { 1, 1, sizeof(u32) },
+  [RTA_TABLE]	  = { 1, 1, sizeof(u32) },
+  [RTA_VIA]	  = { 1, 0, 0 },
+  [RTA_NEWDST]	  = { 1, 0, 0 },
 };
 
 
@@ -373,6 +411,9 @@ nl_parse_attrs(struct rtattr *a, struct nl_want_attrs *want, struct rtattr **k,
   return 1;
 }
 
+static inline u16 rta_get_u16(struct rtattr *a)
+{ return *(u16 *) RTA_DATA(a); }
+
 static inline u32 rta_get_u32(struct rtattr *a)
 { return *(u32 *) RTA_DATA(a); }
 
@@ -390,6 +431,25 @@ static inline ip_addr rta_get_ipa(struct rtattr *a)
     return ipa_from_ip6(rta_get_ip6(a));
 }
 
+static inline ip_addr rta_get_via(struct rtattr *a)
+{
+  struct rtvia *v = RTA_DATA(a);
+  switch(v->rtvia_family) {
+    case AF_INET:  return ipa_from_ip4(ip4_ntoh(*(ip4_addr *) v->rtvia_addr));
+    case AF_INET6: return ipa_from_ip6(ip6_ntoh(*(ip6_addr *) v->rtvia_addr));
+  }
+  return IPA_NONE;
+}
+
+static u32 rta_mpls_stack[MPLS_MAX_LABEL_STACK];
+static inline int rta_get_mpls(struct rtattr *a, u32 *stack)
+{
+  if (RTA_PAYLOAD(a) % 4)
+    log(L_WARN "KRT: Strange length of received MPLS stack: %u", RTA_PAYLOAD(a));
+
+  return mpls_get(RTA_DATA(a), RTA_PAYLOAD(a) & ~0x3, stack);
+}
+
 struct rtattr *
 nl_add_attr(struct nlmsghdr *h, uint bufsize, uint code, const void *data, uint dlen)
 {
@@ -410,6 +470,24 @@ nl_add_attr(struct nlmsghdr *h, uint bufsize, uint code, const void *data, uint
   return a;
 }
 
+static inline struct rtattr *
+nl_open_attr(struct nlmsghdr *h, uint bufsize, uint code)
+{
+  return nl_add_attr(h, bufsize, code, NULL, 0);
+}
+
+static inline void
+nl_close_attr(struct nlmsghdr *h, struct rtattr *a)
+{
+  a->rta_len = (void *)h + NLMSG_ALIGN(h->nlmsg_len) - (void *)a;
+}
+
+static inline void
+nl_add_attr_u16(struct nlmsghdr *h, uint bufsize, int code, u16 data)
+{
+  nl_add_attr(h, bufsize, code, &data, 2);
+}
+
 static inline void
 nl_add_attr_u32(struct nlmsghdr *h, uint bufsize, int code, u32 data)
 {
@@ -439,16 +517,47 @@ nl_add_attr_ipa(struct nlmsghdr *h, uint bufsize, int code, ip_addr ipa)
     nl_add_attr_ip6(h, bufsize, code, ipa_to_ip6(ipa));
 }
 
-static inline struct rtattr *
-nl_open_attr(struct nlmsghdr *h, uint bufsize, uint code)
+static inline void
+nl_add_attr_mpls(struct nlmsghdr *h, uint bufsize, int code, int len, u32 *stack)
 {
-  return nl_add_attr(h, bufsize, code, NULL, 0);
+  char buf[len*4];
+  mpls_put(buf, len, stack);
+  nl_add_attr(h, bufsize, code, buf, len*4);
 }
 
 static inline void
-nl_close_attr(struct nlmsghdr *h, struct rtattr *a)
+nl_add_attr_mpls_encap(struct nlmsghdr *h, uint bufsize, int len, u32 *stack)
 {
-  a->rta_len = (void *)h + NLMSG_ALIGN(h->nlmsg_len) - (void *)a;
+  nl_add_attr_u16(h, bufsize, RTA_ENCAP_TYPE, LWTUNNEL_ENCAP_MPLS);
+
+  struct rtattr *nest = nl_open_attr(h, bufsize, RTA_ENCAP);
+  nl_add_attr_mpls(h, bufsize, RTA_DST, len, stack);
+  nl_close_attr(h, nest);
+}
+
+static inline void
+nl_add_attr_via(struct nlmsghdr *h, uint bufsize, ip_addr ipa)
+{
+  struct rtattr *nest = nl_open_attr(h, bufsize, RTA_VIA);
+  struct rtvia *via = RTA_DATA(nest);
+
+  h->nlmsg_len += sizeof(*via);
+
+  if (ipa_is_ip4(ipa)) {
+    ip4_addr ip4 = ipa_to_ip4(ipa);
+    ip4 = ip4_hton(ip4);
+    via->rtvia_family = AF_INET;
+    memcpy(via->rtvia_addr, &ip4, sizeof(ip4));
+    h->nlmsg_len += sizeof(ip4);
+  } else {
+    ip6_addr ip6 = ipa_to_ip6(ipa);
+    ip6 = ip6_hton(ip6);
+    via->rtvia_family = AF_INET6;
+    memcpy(via->rtvia_addr, &ip6, sizeof(ip6));
+    h->nlmsg_len += sizeof(ip6);
+  }
+
+  nl_close_attr(h, nest);
 }
 
 static inline struct rtnexthop *
@@ -471,8 +580,24 @@ nl_close_nexthop(struct nlmsghdr *h, struct rtnexthop *nh)
   nh->rtnh_len = (void *)h + NLMSG_ALIGN(h->nlmsg_len) - (void *)nh;
 }
 
+static inline void
+nl_add_nexthop(struct nlmsghdr *h, uint bufsize, struct nexthop *nh, int af)
+{
+  if (nh->labels > 0)
+    if (af == AF_MPLS)
+      nl_add_attr_mpls(h, bufsize, RTA_NEWDST, nh->labels, nh->label);
+    else
+      nl_add_attr_mpls_encap(h, bufsize, nh->labels, nh->label);
+
+  if (ipa_nonzero(nh->gw))
+    if (af == AF_MPLS)
+      nl_add_attr_via(h, bufsize, nh->gw);
+    else
+      nl_add_attr_ipa(h, bufsize, RTA_GATEWAY, nh->gw);
+}
+
 static void
-nl_add_multipath(struct nlmsghdr *h, uint bufsize, struct nexthop *nh)
+nl_add_multipath(struct nlmsghdr *h, uint bufsize, struct nexthop *nh, int af)
 {
   struct rtattr *a = nl_open_attr(h, bufsize, RTA_MULTIPATH);
 
@@ -484,7 +609,7 @@ nl_add_multipath(struct nlmsghdr *h, uint bufsize, struct nexthop *nh)
     rtnh->rtnh_hops = nh->weight;
     rtnh->rtnh_ifindex = nh->iface->index;
 
-    nl_add_attr_ipa(h, bufsize, RTA_GATEWAY, nh->gw);
+    nl_add_nexthop(h, bufsize, nh, af);
 
     nl_close_nexthop(h, rtnh);
   }
@@ -518,7 +643,7 @@ nl_parse_multipath(struct krt_proto *p, struct rtattr *ra)
       if (nh_buf_used == nh_buf_size)
       {
 	nh_buf_size = nh_buf_size ? (nh_buf_size * 2) : 4;
-	nh_buffer = xrealloc(nh_buffer, nh_buf_size * sizeof(struct nexthop));
+	nh_buffer = xrealloc(nh_buffer, nh_buf_size * NEXTHOP_MAX_SIZE);
       }
       *last = rv = nh_buffer + nh_buf_used++;
       rv->next = NULL;
@@ -543,7 +668,21 @@ nl_parse_multipath(struct krt_proto *p, struct rtattr *ra)
 	    return NULL;
 	}
       else
-	return NULL;
+	rv->gw = IPA_NONE;
+      if (a[RTA_ENCAP_TYPE])
+	{
+	  if (rta_get_u16(a[RTA_ENCAP_TYPE]) != LWTUNNEL_ENCAP_MPLS) {
+	    log(L_WARN "KRT: Unknown encapsulation method %d in multipath", rta_get_u16(a[RTA_ENCAP_TYPE]));
+	    return NULL;
+	  }
+
+	  struct rtattr *enca[BIRD_RTA_MAX];
+	  nl_attr_len = RTA_PAYLOAD(a[RTA_ENCAP]);
+	  nl_parse_attrs(RTA_DATA(a[RTA_ENCAP]), encap_mpls_want, enca, sizeof(enca));
+	  rv->labels = rta_get_mpls(enca[RTA_DST], rv->label);
+	  break;
+	}
+
 
       len -= NLMSG_ALIGN(nh->rtnh_len);
       nh = RTNH_NEXT(nh);
@@ -1008,7 +1147,13 @@ nl_send_route(struct krt_proto *p, rte *e, struct ea_list *eattrs, int op, int d
   r->r.rtm_dst_len = net_pxlen(net->n.addr);
   r->r.rtm_protocol = RTPROT_BIRD;
   r->r.rtm_scope = RT_SCOPE_UNIVERSE;
-  nl_add_attr_ipa(&r->h, rsize, RTA_DST, net_prefix(net->n.addr));
+  if (p->af == AF_MPLS)
+  {
+    u32 label = net_mpls(net->n.addr);
+    nl_add_attr_mpls(&r->h, rsize, RTA_DST, 1, &label);
+  }
+  else
+    nl_add_attr_ipa(&r->h, rsize, RTA_DST, net_prefix(net->n.addr));
 
   /*
    * Strange behavior for RTM_DELROUTE:
@@ -1071,13 +1216,11 @@ dest:
     case RTD_UNICAST:
       r->r.rtm_type = RTN_UNICAST;
       if (nh->next && !krt_ecmp6(p))
-	nl_add_multipath(&r->h, rsize, nh);
+	nl_add_multipath(&r->h, rsize, nh, p->af);
       else
       {
 	nl_add_attr_u32(&r->h, rsize, RTA_OIF, nh->iface->index);
-
-	if (ipa_nonzero(nh->gw))
-	  nl_add_attr_ipa(&r->h, rsize, RTA_GATEWAY, nh->gw);
+	nl_add_nexthop(&r->h, rsize, nh, p->af);
       }
       break;
     case RTD_BLACKHOLE:
@@ -1276,6 +1419,19 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
 	net_fill_ip6(&dst, IP6_NONE, 0);
       break;
 
+    case AF_MPLS:
+      if (!nl_parse_attrs(RTM_RTA(i), rtm_attr_want_mpls, a, sizeof(a)))
+	return;
+
+      if (a[RTA_DST])
+	if (rta_get_mpls(a[RTA_DST], rta_mpls_stack) == 1)
+	  net_fill_mpls(&dst, rta_mpls_stack[0]);
+	else
+	  log(L_WARN "KRT: Got multi-label MPLS RTA_DST");
+      else
+	return; /* No support for MPLS routes without RTA_DST */
+      break;
+
     default:
       return;
     }
@@ -1338,7 +1494,7 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
   if (s->net && !nl_mergable_route(s, net, p, priority, i->rtm_type))
     nl_announce_route(s);
 
-  rta *ra = lp_allocz(s->pool, sizeof(rta)); // TODO: fix alloc
+  rta *ra = lp_allocz(s->pool, RTA_MAX_SIZE);
   ra->src = p->p.main_source;
   ra->source = RTS_INHERIT;
   ra->scope = SCOPE_UNIVERSE;
@@ -1367,9 +1523,12 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
 	  return;
 	}
 
-      if (a[RTA_GATEWAY])
+      if ((i->rtm_family != AF_MPLS) && a[RTA_GATEWAY] || (i->rtm_family == AF_MPLS) && a[RTA_VIA])
 	{
-	  ra->nh.gw = rta_get_ipa(a[RTA_GATEWAY]);
+	  if (i->rtm_family == AF_MPLS)
+	    ra->nh.gw = rta_get_via(a[RTA_VIA]);
+	  else
+	    ra->nh.gw = rta_get_ipa(a[RTA_GATEWAY]);
 
 	  /* Silently skip strange 6to4 routes */
 	  const net_addr_ip6 sit = NET_ADDR_IP6(IP6_NONE, 96);
@@ -1377,7 +1536,7 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
 	    return;
 
 	  neighbor *nbr;
-	  nbr = neigh_find2(&p->p, &ra->nh.gw, ra->nh.iface,
+	  nbr = neigh_find2(&p->p, &(ra->nh.gw), ra->nh.iface,
 			    (i->rtm_flags & RTNH_F_ONLINK) ? NEF_ONLINK : 0);
 	  if (!nbr || (nbr->scope == SCOPE_HOST))
 	    {
@@ -1403,6 +1562,44 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
       return;
     }
 
+  int labels = 0;
+  if ((i->rtm_family == AF_MPLS) && a[RTA_NEWDST] && !ra->nh.next)
+    labels = rta_get_mpls(a[RTA_NEWDST], ra->nh.label);
+
+  if (a[RTA_ENCAP] && a[RTA_ENCAP_TYPE] && !ra->nh.next)
+    {
+      switch (rta_get_u16(a[RTA_ENCAP_TYPE]))
+	{
+	  case LWTUNNEL_ENCAP_MPLS:
+	    {
+	      struct rtattr *enca[BIRD_RTA_MAX];
+	      nl_attr_len = RTA_PAYLOAD(a[RTA_ENCAP]);
+	      nl_parse_attrs(RTA_DATA(a[RTA_ENCAP]), encap_mpls_want, enca, sizeof(enca));
+	      labels = rta_get_mpls(enca[RTA_DST], ra->nh.label);
+	      break;
+	    }
+	  default:
+	    SKIP("unknown encapsulation method %d\n", rta_get_u16(a[RTA_ENCAP_TYPE]));
+	    break;
+	}
+    }
+
+  if (labels < 0)
+  {
+    log(L_WARN "KRT: Too long MPLS stack received, ignoring.");
+    ra->nh.labels = 0;
+  }
+  else
+    ra->nh.labels = labels;
+
+  rte *e = rte_get_temp(ra);
+  e->net = net;
+  e->u.krt.src = src;
+  e->u.krt.proto = i->rtm_protocol;
+  e->u.krt.seen = 0;
+  e->u.krt.best = 0;
+  e->u.krt.metric = 0;
+
   if (i->rtm_scope != def_scope)
     {
       ea_list *ea = lp_alloc(s->pool, sizeof(ea_list) + sizeof(eattr));
@@ -1416,6 +1613,9 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
       ea->attrs[0].u.data = i->rtm_scope;
     }
 
+  if (a[RTA_PRIORITY])
+    e->u.krt.metric = rta_get_u32(a[RTA_PRIORITY]);
+
   if (a[RTA_PREFSRC])
     {
       ip_addr ps = rta_get_ipa(a[RTA_PREFSRC]);
@@ -1527,6 +1727,15 @@ krt_do_scan(struct krt_proto *p UNUSED)	/* CONFIG_ALL_TABLES_AT_ONCE => p is NUL
     else
       log(L_DEBUG "nl_scan_fire: Unknown packet received (type=%d)", h->nlmsg_type);
   nl_parse_end(&s);
+
+  nl_parse_begin(&s, 1, 1);
+  nl_request_dump(AF_MPLS, RTM_GETROUTE);
+  while (h = nl_get_scan())
+    if (h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE)
+      nl_parse_route(&s, h);
+    else
+      log(L_DEBUG "nl_scan_fire: Unknown packet received (type=%d)", h->nlmsg_type);
+  nl_parse_end(&s);
 }
 
 /*
diff --git a/sysdep/unix/krt.c b/sysdep/unix/krt.c
index 9f66d2f4..c273cb10 100644
--- a/sysdep/unix/krt.c
+++ b/sysdep/unix/krt.c
@@ -1153,7 +1153,8 @@ krt_start(struct proto *P)
   {
   case NET_IP4:	p->af = AF_INET; break;
   case NET_IP6:	p->af = AF_INET6; break;
-  default:	ASSERT(0);
+  case NET_MPLS: p->af = AF_MPLS; break;
+  default: log(L_ERR "KRT: Tried to start with strange net type: %d", p->p.net_type); return PS_START; break;
   }
 
   add_tail(&krt_proto_list, &p->krt_node);
@@ -1264,7 +1265,7 @@ struct protocol proto_unix_kernel = {
   .template =		"kernel%d",
   .attr_class =		EAP_KRT,
   .preference =		DEF_PREF_INHERITED,
-  .channel_mask =	NB_IP,
+  .channel_mask =	NB_IP | NB_MPLS,
   .proto_size =		sizeof(struct krt_proto),
   .config_size =	sizeof(struct krt_config),
   .preconfig =		krt_preconfig,
-- 
cgit v1.2.3


From d47c3d64b2733baea756f1bb37ef09f10d7f9644 Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Tue, 9 Aug 2016 14:47:51 +0200
Subject: MPLS: Label stack concatenation for recursive routes

---
 nest/route.h          |  3 ++-
 nest/rt-table.c       | 49 +++++++++++++++++++++++++++++++++++++++----------
 proto/static/config.Y |  7 ++++++-
 proto/static/static.c |  6 +++++-
 4 files changed, 52 insertions(+), 13 deletions(-)

(limited to 'nest/route.h')

diff --git a/nest/route.h b/nest/route.h
index a2fadf62..eb98b609 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -338,7 +338,8 @@ struct nexthop {
   struct iface *iface;			/* Outgoing interface */
   struct nexthop *next;
   byte weight;
-  byte labels;				/* Number of labels appended */
+  byte labels_append;			/* Number of labels before hostentry was applied */
+  byte labels;				/* Number of labels prepended */
   u32 label[0];
 };
 
diff --git a/nest/rt-table.c b/nest/rt-table.c
index 73b838be..ef402f28 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -1768,27 +1768,56 @@ static inline void
 rta_apply_hostentry(rta *a, struct hostentry *he)
 {
   a->hostentry = he;
-
-  a->nh.gw = ipa_nonzero(he->nh->gw) ? he->nh->gw : he->link;
-  a->nh.iface = he->nh->iface;
-  a->nh.weight = he->nh->weight;
-  a->nh.next = he->nh->next;
   
   a->dest = he->dest;
   a->igp_metric = he->igp_metric;
+
+  if (a->nh.labels_append == 0)
+  {
+    a->nh = *(he->nh);
+    a->nh.labels_append = 0;
+    return;
+  }
+
+  int labels_append = a->nh.labels_append;
+  u32 label_stack[MPLS_MAX_LABEL_STACK];
+  memcpy(label_stack, a->nh.label, labels_append * sizeof(u32));
+
+  struct nexthop *nhp = NULL;
+  for (struct nexthop *nh = he->nh; nh; nh = nh->next)
+  {
+    nhp = nhp ? (nhp->next = lp_alloc(rte_update_pool, NEXTHOP_MAX_SIZE)) : &(a->nh);
+    nhp->gw = ipa_nonzero(nh->gw) ? nh->gw : he->link;
+    nhp->iface = nh->iface; /* FIXME: This is at least strange, if not utter nonsense. */
+    nhp->weight = nh->weight;
+    nhp->labels = nh->labels + labels_append;
+    nhp->labels_append = labels_append;
+    if (nhp->labels <= MPLS_MAX_LABEL_STACK)
+    {
+      memcpy(nhp->label, nh->label, nh->labels * sizeof(u32));
+      memcpy(&(nhp->label[nh->labels]), label_stack, labels_append * sizeof(u32));
+    }
+    else
+    {
+      log(L_WARN "Sum of label stack sizes %d + %d = %d exceedes allowed maximum (%d)",
+	  nh->labels, labels_append, nhp->labels, MPLS_MAX_LABEL_STACK);
+      a->dest = RTD_UNREACHABLE;
+      break;
+    }
+  }
 }
 
 static inline rte *
 rt_next_hop_update_rte(rtable *tab UNUSED, rte *old)
 {
-  rta a;
-  memcpy(&a, old->attrs, rta_size(old->attrs));
-  rta_apply_hostentry(&a, old->attrs->hostentry);
-  a.aflags = 0;
+  rta *ap = alloca(RTA_MAX_SIZE);
+  memcpy(ap, old->attrs, rta_size(old->attrs));
+  rta_apply_hostentry(ap, old->attrs->hostentry);
+  ap->aflags = 0;
 
   rte *e = sl_alloc(rte_slab);
   memcpy(e, old, sizeof(rte));
-  e->attrs = rta_lookup(&a);
+  e->attrs = rta_lookup(ap);
 
   return e;
 }
diff --git a/proto/static/config.Y b/proto/static/config.Y
index a461bd27..2fb54448 100644
--- a/proto/static/config.Y
+++ b/proto/static/config.Y
@@ -105,7 +105,12 @@ stat_route:
       this_srt->dest = RTDX_RECURSIVE;
       this_srt->via = $3;
    }
-
+ | stat_route0 RECURSIVE ipa MPLS label_stack {
+      this_srt->dest = RTDX_RECURSIVE;
+      this_srt->via = $3;
+      this_srt->label_count = $5[0];
+      this_srt->label_stack = &($5[1]);
+   }
  | stat_route0 DROP		{ this_srt->dest = RTD_BLACKHOLE; }
  | stat_route0 REJECT		{ this_srt->dest = RTD_UNREACHABLE; }
  | stat_route0 BLACKHOLE	{ this_srt->dest = RTD_BLACKHOLE; }
diff --git a/proto/static/static.c b/proto/static/static.c
index 878248c1..3e03708c 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -128,7 +128,11 @@ drop:
     r->state |= STS_INSTALLED;
   
   if (r->dest == RTDX_RECURSIVE)
-    rta_set_recursive_next_hop(p->main_channel->table, ap, p_igp_table(p), r->via, IPA_NONE);
+    {
+      ap->nh.labels_append = ap->nh.labels = r->label_count;
+      memcpy(ap->nh.label, r->label_stack, r->label_count * sizeof(u32));
+      rta_set_recursive_next_hop(p->main_channel->table, ap, p_igp_table(p), r->via, IPA_NONE);
+    }
 
   /* We skip rta_lookup() here */
 
-- 
cgit v1.2.3


From 62e64905b76b88da72c522eac9276a74f60c9592 Mon Sep 17 00:00:00 2001
From: "Ondrej Zajicek (work)" <santiago@crfreenet.org>
Date: Mon, 20 Feb 2017 02:26:45 +0100
Subject: Several minor fixes

---
 conf/cf-lex.l          | 39 +++++++++++++++-------
 conf/confbase.Y        | 11 ++++---
 doc/bird.sgml          | 10 +++---
 filter/filter.h        |  6 ++--
 lib/alloca.h           |  2 ++
 lib/net.c              | 19 ++++++-----
 lib/net.h              | 16 ++++++----
 nest/route.h           |  8 ++---
 nest/rt-attr.c         | 42 ++++++++++--------------
 nest/rt-dev.c          |  4 +--
 nest/rt-table.c        | 16 +++++-----
 proto/bgp/attrs.c      |  3 +-
 proto/bgp/packets.c    | 12 +++----
 proto/ospf/rt.c        | 10 ++----
 proto/pipe/pipe.c      |  2 --
 proto/rip/rip.c        | 21 +++++-------
 proto/rpki/rpki.c      |  2 +-
 proto/static/config.Y  | 87 ++++++++++++++++++++++++++++----------------------
 proto/static/static.c  | 53 +++++++++++++-----------------
 sysdep/bsd/krt-sock.c  | 86 ++++++++++++++++++++++++-------------------------
 sysdep/linux/netlink.c | 52 +++++++++++++++++-------------
 sysdep/unix/krt.c      | 13 +-------
 22 files changed, 253 insertions(+), 261 deletions(-)

(limited to 'nest/route.h')

diff --git a/conf/cf-lex.l b/conf/cf-lex.l
index 8e1f60a6..fb3d59e4 100644
--- a/conf/cf-lex.l
+++ b/conf/cf-lex.l
@@ -124,39 +124,56 @@ include   ^{WHITE}*include{WHITE}*\".*\"{WHITE}*;
 }
 
 [02]:{DIGIT}+:{DIGIT}+ {
+  unsigned long int l, len1, len2;
   char *e;
-  unsigned long int l;
 
   if (yytext[0] == '0')
+  {
     cf_lval.i64 = 0;
+    len1 = 16;
+    len2 = 32;
+  }
   else
-    cf_lval.i64 = 0x2000000000000ULL;
+  {
+    cf_lval.i64 = 2ULL << 48;
+    len1 = 32;
+    len2 = 16;
+  }
 
   errno = 0;
   l = strtoul(yytext+2, &e, 10);
-  if (e && (*e != ':') || errno == ERANGE || (yytext[0] == '0') && (l >= (1<<16)))
+  if (e && (*e != ':') || (errno == ERANGE) || (l >> len1))
     cf_error("ASN out of range");
-  cf_lval.i64 |= (((u64) l) << 32);
+  cf_lval.i64 |= ((u64) l) << len2;
+
   errno = 0;
   l = strtoul(e+1, &e, 10);
-  if (e && *e || errno == ERANGE || (yytext[0] == '2') && (l >= (1<<16)))
-    cf_error("Assigned number out of range");
+  if (e && *e || (errno == ERANGE) || (l >> len2))
+    cf_error("Number out of range");
   cf_lval.i64 |= l;
+
   return VPN_RD;
 }
 
 1:{DIGIT}+\.{DIGIT}+\.{DIGIT}+\.{DIGIT}+:{DIGIT}+ {
   unsigned long int l;
-  char *e = strchr(yytext+2, ':');
-  *e++ = '\0';
   ip4_addr ip4;
+  char *e;
+
+  cf_lval.i64 = 1ULL << 48;
+
+  e = strchr(yytext+2, ':');
+  *e++ = '\0';
   if (!ip4_pton(yytext+2, &ip4))
     cf_error("Invalid IPv4 address %s in Route Distinguisher", yytext+2);
+  cf_lval.i64 |= ((u64) ip4_to_u32(ip4)) << 16;
+
   errno = 0;
   l = strtoul(e, &e, 10);
-  if (e && *e || errno == ERANGE || (l >= (1<<16)))
-    cf_error("Assigned number out of range");
-  cf_lval.i64 = (1ULL<<48) | (((u64)ip4_to_u32(ip4)) << 16) | ((u64)l);
+  if (e && *e || (errno == ERANGE) || (l >> 16))
+    cf_error("Number out of range");
+  cf_lval.i64 |= l;
+
   return VPN_RD;
 }
 
diff --git a/conf/confbase.Y b/conf/confbase.Y
index a5b8b692..291dc6a0 100644
--- a/conf/confbase.Y
+++ b/conf/confbase.Y
@@ -203,13 +203,13 @@ net_ip6_: IP6 '/' NUM
 net_vpn4_: VPN_RD net_ip4_
 {
   $$ = cfg_alloc(sizeof(net_addr_vpn4));
-  net_fill_vpn4($$, ((net_addr_ip4 *)&$2)->prefix, $2.pxlen, $1);
+  net_fill_vpn4($$, net4_prefix(&$2), net4_pxlen(&$2), $1);
 }
 
 net_vpn6_: VPN_RD net_ip6_
 {
   $$ = cfg_alloc(sizeof(net_addr_vpn6));
-  net_fill_vpn6($$, ((net_addr_ip6 *)&$2)->prefix, $2.pxlen, $1);
+  net_fill_vpn6($$, net6_prefix(&$2), net6_pxlen(&$2), $1);
 }
 
 net_roa4_: net_ip4_ MAX NUM AS NUM
@@ -229,8 +229,8 @@ net_roa6_: net_ip6_ MAX NUM AS NUM
 };
 
 net_ip_: net_ip4_ | net_ip6_ ;
-net_roa_: net_roa4_ | net_roa6_ ;
 net_vpn_: net_vpn4_ | net_vpn6_ ;
+net_roa_: net_roa4_ | net_roa6_ ;
 
 net_:
    net_ip_ { $$ = cfg_alloc($1.length); net_copy($$, &($1)); }
@@ -297,8 +297,9 @@ label_stack:
     label_stack_start
   | label_stack '/' NUM {
     if ($1[0] >= MPLS_MAX_LABEL_STACK)
-      cf_error("Too many labels in stack.");
-    $1[++$1[0]] = $3;
+      cf_error("Too many labels in stack");
+    $1[0]++;
+    $1[*$1] = $3;
     $$ = $1;
   }
 ;
diff --git a/doc/bird.sgml b/doc/bird.sgml
index 23026eae..c49e7e8e 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -4140,12 +4140,12 @@ return packets as undeliverable if they are in your IP block, you don't have any
 specific destination for them and you don't want to send them out through the
 default route to prevent routing loops).
 
-<p>There are five types of static routes: `classical' routes telling to forward
+<p>There are four types of static routes: `classical' routes telling to forward
 packets to a neighboring router (single path or multipath, possibly weighted),
-device routes specifying forwarding to hosts on a
-directly connected network, recursive routes computing their nexthops by doing
-route table lookups for a given IP, and special routes (sink, blackhole etc.)
-which specify a special action to be done instead of forwarding the packet.
+device routes specifying forwarding to hosts on a directly connected network,
+recursive routes computing their nexthops by doing route table lookups for a
+given IP, and special routes (sink, blackhole etc.)  which specify a special
+action to be done instead of forwarding the packet.
 
 <p>When the particular destination is not available (the interface is down or
 the next hop of the route is not a neighbor at the moment), Static just
diff --git a/filter/filter.h b/filter/filter.h
index 0482b83b..855219ec 100644
--- a/filter/filter.h
+++ b/filter/filter.h
@@ -174,9 +174,9 @@ void val_format(struct f_val v, buffer *buf);
 #define SA_PROTO	 4
 #define SA_SOURCE	 5
 #define SA_SCOPE	 6
-#define SA_DEST    	 8
-#define SA_IFNAME  	 9
-#define SA_IFINDEX    	10
+#define SA_DEST    	 7
+#define SA_IFNAME  	 8
+#define SA_IFINDEX    	 9
 
 
 struct f_tree {
diff --git a/lib/alloca.h b/lib/alloca.h
index f0d61bb4..e5557cdb 100644
--- a/lib/alloca.h
+++ b/lib/alloca.h
@@ -15,4 +15,6 @@
 #include <stdlib.h>
 #endif
 
+#define allocz(len) ({ void *_x = alloca(len); memset(_x, 0, len); _x; })
+
 #endif
diff --git a/lib/net.c b/lib/net.c
index e46be8b2..74cea277 100644
--- a/lib/net.c
+++ b/lib/net.c
@@ -69,21 +69,20 @@ net_format(const net_addr *N, char *buf, int buflen)
   case NET_VPN4:
     switch (n->vpn4.rd >> 48)
     {
-      case 0: return bsnprintf(buf, buflen, "0:%u:%u %I4/%d", (u32) (n->vpn4.rd >> 32), (u32) n->vpn4.rd, n->vpn4.prefix, n->vpn4.pxlen);
-      case 1: return bsnprintf(buf, buflen, "1:%I4:%u %I4/%d", ip4_from_u32(n->vpn4.rd >> 16), (u32) (n->vpn4.rd & 0xffff), n->vpn4.prefix, n->vpn4.pxlen);
-      case 2: return bsnprintf(buf, buflen, "2:%u:%u %I4/%d", (u32) (n->vpn4.rd >> 16), (u32) (n->vpn4.rd & 0xffff), n->vpn4.prefix, n->vpn4.pxlen);
+    case 0: return bsnprintf(buf, buflen, "0:%u:%u %I4/%d", (u32) (n->vpn4.rd >> 32), (u32) n->vpn4.rd, n->vpn4.prefix, n->vpn4.pxlen);
+    case 1: return bsnprintf(buf, buflen, "1:%I4:%u %I4/%d", ip4_from_u32(n->vpn4.rd >> 16), (u32) (n->vpn4.rd & 0xffff), n->vpn4.prefix, n->vpn4.pxlen);
+    case 2: return bsnprintf(buf, buflen, "2:%u:%u %I4/%d", (u32) (n->vpn4.rd >> 16), (u32) (n->vpn4.rd & 0xffff), n->vpn4.prefix, n->vpn4.pxlen);
+    default: return bsnprintf(buf, buflen, "X:%08x:%08x %I4/%d", (u32) (n->vpn4.rd >> 32), (u32) n->vpn4.rd, n->vpn4.prefix, n->vpn4.pxlen);
     }
-    return bsnprintf(buf, buflen, "X: %016x %I4/%d", (n->vpn4.rd), n->vpn4.prefix, n->vpn4.pxlen);
-
-    /* XXX: RD format is specified for VPN4; not found any for VPN6, reusing the same as for VPN4. */
   case NET_VPN6:
+    /* XXX: RD format is specified for VPN4; not found any for VPN6, reusing the same as for VPN4 */
     switch (n->vpn6.rd >> 48)
     {
-      case 0: return bsnprintf(buf, buflen, "0:%u:%u %I6/%d", (u32) (n->vpn6.rd >> 32), (u32) n->vpn6.rd, n->vpn6.prefix, n->vpn6.pxlen);
-      case 1: return bsnprintf(buf, buflen, "1:%I4:%u %I6/%d", ip4_from_u32(n->vpn6.rd >> 16), (u32) (n->vpn6.rd & 0xffff), n->vpn6.prefix, n->vpn6.pxlen);
-      case 2: return bsnprintf(buf, buflen, "2:%u:%u %I6/%d", (u32) (n->vpn6.rd >> 16), (u32) (n->vpn6.rd & 0xffff), n->vpn6.prefix, n->vpn6.pxlen);
+    case 0: return bsnprintf(buf, buflen, "0:%u:%u %I6/%d", (u32) (n->vpn6.rd >> 32), (u32) n->vpn6.rd, n->vpn6.prefix, n->vpn6.pxlen);
+    case 1: return bsnprintf(buf, buflen, "1:%I4:%u %I6/%d", ip4_from_u32(n->vpn6.rd >> 16), (u32) (n->vpn6.rd & 0xffff), n->vpn6.prefix, n->vpn6.pxlen);
+    case 2: return bsnprintf(buf, buflen, "2:%u:%u %I6/%d", (u32) (n->vpn6.rd >> 16), (u32) (n->vpn6.rd & 0xffff), n->vpn6.prefix, n->vpn6.pxlen);
+    default: return bsnprintf(buf, buflen, "X:%08x:%08x %I6/%d", (u32) (n->vpn6.rd >> 32), (u32) n->vpn6.rd, n->vpn6.prefix, n->vpn6.pxlen);
     }
-    return bsnprintf(buf, buflen, "X: %016x %I6/%d", (n->vpn6.rd), n->vpn6.prefix, n->vpn6.pxlen);
   case NET_ROA4:
     return bsnprintf(buf, buflen, "%I4/%u-%u AS%u",  n->roa4.prefix, n->roa4.pxlen, n->roa4.max_pxlen, n->roa4.asn);
   case NET_ROA6:
diff --git a/lib/net.h b/lib/net.h
index 7c124fc0..7144bcb9 100644
--- a/lib/net.h
+++ b/lib/net.h
@@ -306,6 +306,9 @@ static inline int net_equal_flow4(const net_addr_flow4 *a, const net_addr_flow4
 static inline int net_equal_flow6(const net_addr_flow6 *a, const net_addr_flow6 *b)
 { return net_equal((const net_addr *) a, (const net_addr *) b); }
 
+static inline int net_equal_mpls(const net_addr_mpls *a, const net_addr_mpls *b)
+{ return !memcmp(a, b, sizeof(net_addr_mpls)); }
+
 
 static inline int net_equal_prefix_roa4(const net_addr_roa4 *a, const net_addr_roa4 *b)
 { return ip4_equal(a->prefix, b->prefix) && (a->pxlen == b->pxlen); }
@@ -313,8 +316,6 @@ static inline int net_equal_prefix_roa4(const net_addr_roa4 *a, const net_addr_r
 static inline int net_equal_prefix_roa6(const net_addr_roa6 *a, const net_addr_roa6 *b)
 { return ip6_equal(a->prefix, b->prefix) && (a->pxlen == b->pxlen); }
 
-static inline int net_equal_mpls(const net_addr_mpls *a, const net_addr_mpls *b)
-{ return !memcmp(a, b, sizeof(net_addr_mpls)); }
 
 static inline int net_zero_ip4(const net_addr_ip4 *a)
 { return !a->pxlen && ip4_zero(a->prefix); }
@@ -404,16 +405,17 @@ static inline void net_copy_flow6(net_addr_flow6 *dst, const net_addr_flow6 *src
 static inline void net_copy_mpls(net_addr_mpls *dst, const net_addr_mpls *src)
 { memcpy(dst, src, sizeof(net_addr_mpls)); }
 
+
+/* XXXX */
+static inline u32 u64_hash(u64 a)
+{ return u32_hash(a); }
+
 static inline u32 net_hash_ip4(const net_addr_ip4 *n)
 { return ip4_hash(n->prefix) ^ ((u32) n->pxlen << 26); }
 
 static inline u32 net_hash_ip6(const net_addr_ip6 *n)
 { return ip6_hash(n->prefix) ^ ((u32) n->pxlen << 26); }
 
-/* XXXX */
-static inline u32 u64_hash(u64 a)
-{ return u32_hash(a); }
-
 static inline u32 net_hash_vpn4(const net_addr_vpn4 *n)
 { return ip4_hash(n->prefix) ^ ((u32) n->pxlen << 26) ^ u64_hash(n->rd); }
 
@@ -452,7 +454,7 @@ static inline int net_validate_ip6(const net_addr_ip6 *n)
 
 static inline int net_validate_mpls(const net_addr_mpls *n)
 {
-  return n->label < (1<<20);
+  return n->label < (1 << 20);
 }
 
 int net_validate(const net_addr *N);
diff --git a/nest/route.h b/nest/route.h
index eb98b609..1e0a14bc 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -390,11 +390,11 @@ typedef struct rta {
 #define RTC_MULTICAST 2
 #define RTC_ANYCAST 3			/* IPv6 Anycast */
 
-#define RTD_UNICAST 0			/* Next hop is neighbor router */
+#define RTD_NONE 0			/* Undefined next hop */
+#define RTD_UNICAST 1			/* Next hop is neighbor router */
 #define RTD_BLACKHOLE 2			/* Silently drop packets */
 #define RTD_UNREACHABLE 3		/* Reject as unreachable */
 #define RTD_PROHIBIT 4			/* Administratively prohibited */
-#define RTD_NONE 6			/* Invalid RTD */
 
 					/* Flags for net->n.flags, used by kernel syncer */
 #define KRF_INSTALLED 0x80		/* This route should be installed in the kernel */
@@ -408,7 +408,7 @@ typedef struct rta {
 
 /* Route has regular, reachable nexthop (i.e. not RTD_UNREACHABLE and like) */
 static inline int rte_is_reachable(rte *r)
-{ uint d = r->attrs->dest; return (d == RTD_UNICAST); }
+{ return r->attrs->dest == RTD_UNICAST; }
 
 
 /*
@@ -523,7 +523,7 @@ static inline int nexthop_same(struct nexthop *x, struct nexthop *y)
 struct nexthop *nexthop_merge(struct nexthop *x, struct nexthop *y, int rx, int ry, int max, linpool *lp);
 static inline void nexthop_link(struct rta *a, struct nexthop *from)
 { memcpy(&a->nh, from, nexthop_size(from)); }
-void nexthop_insert(struct nexthop *n, struct nexthop *y);
+void nexthop_insert(struct nexthop **n, struct nexthop *y);
 int nexthop_is_sorted(struct nexthop *x);
 
 void rta_init(void);
diff --git a/nest/rt-attr.c b/nest/rt-attr.c
index 120a8e24..afc97e22 100644
--- a/nest/rt-attr.c
+++ b/nest/rt-attr.c
@@ -150,7 +150,8 @@ nexthop_hash(struct nexthop *x)
   for (; x; x = x->next)
   {
     h ^= ipa_hash(x->gw) ^ (h << 5) ^ (h >> 9);
-    for (int i=0; i<x->labels; i++)
+
+    for (int i = 0; i < x->labels; i++)
       h ^= x->label[i] ^ (h << 6) ^ (h >> 7);
   }
 
@@ -164,12 +165,13 @@ nexthop__same(struct nexthop *x, struct nexthop *y)
   {
     if (!ipa_equal(x->gw, y->gw) || (x->iface != y->iface) || (x->weight != y->weight) || (x->labels != y->labels))
       return 0;
-    for (int i=0; i<x->labels; i++)
+
+    for (int i = 0; i < x->labels; i++)
       if (x->label[i] != y->label[i])
 	return 0;
   }
 
-  return 1;
+  return x == y;
 }
 
 static int
@@ -195,7 +197,7 @@ nexthop_compare_node(struct nexthop *x, struct nexthop *y)
   if (r)
     return r;
 
-  for (int i=0; i<y->labels; i++)
+  for (int i = 0; i < y->labels; i++)
   {
     r = ((int) y->label[i]) - ((int) x->label[i]);
     if (r)
@@ -271,34 +273,22 @@ nexthop_merge(struct nexthop *x, struct nexthop *y, int rx, int ry, int max, lin
 }
 
 void
-nexthop_insert(struct nexthop *n, struct nexthop *x)
+nexthop_insert(struct nexthop **n, struct nexthop *x)
 {
-  struct nexthop tmp;
-  memcpy(&tmp, n, sizeof(struct nexthop));
-  if (nexthop_compare_node(n, x) > 0) /* Insert to the included nexthop */
-  {
-    memcpy(n, x, sizeof(struct nexthop));
-    memcpy(x, &tmp, sizeof(struct nexthop));
-    n->next = x;
-    return;
-  }
-
-  for (struct nexthop **nn = &(n->next); *nn; nn = &((*nn)->next))
+  for (; *n; n = &((*n)->next))
   {
-    int cmp = nexthop_compare_node(*nn, x);
+    int cmp = nexthop_compare_node(*n, x);
 
     if (cmp < 0)
       continue;
-    
-    if (cmp > 0)
-    {
-      x->next = *nn;
-      *nn = x;
-    }
-    
-    return;
+    else if (cmp > 0)
+      break;
+    else
+      return;
   }
 
+  x->next = *n;
+  *n = x;
 }
 
 int
@@ -314,7 +304,7 @@ nexthop_is_sorted(struct nexthop *x)
 static inline slab *
 nexthop_slab(struct nexthop *nh)
 {
-  return nexthop_slab_[nh->labels > 2 ? 3 : nh->labels];
+  return nexthop_slab_[MIN(nh->labels, 3)];
 }
 
 static struct nexthop *
diff --git a/nest/rt-dev.c b/nest/rt-dev.c
index 5edd1c5d..9993da24 100644
--- a/nest/rt-dev.c
+++ b/nest/rt-dev.c
@@ -79,9 +79,7 @@ dev_ifa_notify(struct proto *P, uint flags, struct ifa *ad)
 	.source = RTS_DEVICE,
 	.scope = SCOPE_UNIVERSE,
 	.dest = RTD_UNICAST,
-	.nh = {
-	  .iface = ad->iface
-	}
+	.nh.iface = ad->iface,
       };
 
       a = rta_lookup(&a0);
diff --git a/nest/rt-table.c b/nest/rt-table.c
index ef402f28..a33b7909 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -1768,7 +1768,6 @@ static inline void
 rta_apply_hostentry(rta *a, struct hostentry *he)
 {
   a->hostentry = he;
-  
   a->dest = he->dest;
   a->igp_metric = he->igp_metric;
 
@@ -1810,14 +1809,14 @@ rta_apply_hostentry(rta *a, struct hostentry *he)
 static inline rte *
 rt_next_hop_update_rte(rtable *tab UNUSED, rte *old)
 {
-  rta *ap = alloca(RTA_MAX_SIZE);
-  memcpy(ap, old->attrs, rta_size(old->attrs));
-  rta_apply_hostentry(ap, old->attrs->hostentry);
-  ap->aflags = 0;
+  rta *a = alloca(RTA_MAX_SIZE);
+  memcpy(a, old->attrs, rta_size(old->attrs));
+  rta_apply_hostentry(a, old->attrs->hostentry);
+  a->aflags = 0;
 
   rte *e = sl_alloc(rte_slab);
   memcpy(e, old, sizeof(rte));
-  e->attrs = rta_lookup(ap);
+  e->attrs = rta_lookup(a);
 
   return e;
 }
@@ -2373,7 +2372,8 @@ rt_update_hostentry(rtable *tab, struct hostentry *he)
 	}
 
       if ((a->dest == RTD_UNICAST) && ipa_zero(a->nh.gw) && !a->next)
-	{ /* We have singlepath device route */
+	{
+	  /* We have singlepath device route */
 	  if (if_local_addr(he->addr, a->nh.iface))
 	    {
 	      /* The host address is a local address, this is not valid */
@@ -2389,7 +2389,7 @@ rt_update_hostentry(rtable *tab, struct hostentry *he)
       else
 	{
 	  /* The host is reachable through some route entry */
-	  he->nh = (&a->nh);
+	  he->nh = &(a->nh);
 	  he->dest = a->dest;
 	}
 
diff --git a/proto/bgp/attrs.c b/proto/bgp/attrs.c
index 11666f30..1b124a17 100644
--- a/proto/bgp/attrs.c
+++ b/proto/bgp/attrs.c
@@ -1461,8 +1461,7 @@ bgp_get_neighbor(rte *r)
 static inline int
 rte_resolvable(rte *rt)
 {
-  int rd = rt->attrs->dest;
-  return (rd == RTD_UNICAST);
+  return rt->attrs->dest == RTD_UNICAST;
 }
 
 int
diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index 9c59e6d8..9380f999 100644
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -700,9 +700,7 @@ bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll)
       WITHDRAW(BAD_NEXT_HOP);
 
     a->dest = RTD_UNICAST;
-    a->nh.gw = nbr->addr;
-    a->nh.iface = nbr->iface;
-    a->nh.next = NULL;
+    a->nh = (struct nexthop){ .gw = nbr->addr, .iface = nbr->iface };
     a->hostentry = NULL;
     a->igp_metric = 0;
   }
@@ -749,8 +747,8 @@ bgp_use_gateway(struct bgp_export_state *s)
   if (s->channel->cf->next_hop_self)
     return 0;
 
-  /* We need valid global gateway */
-  if ((ra->dest != RTD_UNICAST) || (ra->nh.next) || ipa_zero(ra->nh.gw) || ipa_is_link_local(ra->nh.gw))
+  /* 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;
 
   /* Use it when exported to internal peers */
@@ -1434,12 +1432,10 @@ bgp_decode_nlri(struct bgp_parse_state *s, u32 afi, byte *nlri, uint len, ea_lis
 
   if (ea)
   {
-    a = alloca(sizeof(struct rta));
-    memset(a, 0, sizeof(struct rta));
+    a = allocz(sizeof(struct rta));
 
     a->source = RTS_BGP;
     a->scope = SCOPE_UNIVERSE;
-    a->dest = RTD_UNREACHABLE;
     a->from = s->proto->cf->remote_ip;
     a->eattrs = ea;
 
diff --git a/proto/ospf/rt.c b/proto/ospf/rt.c
index 74f47810..1b0ac5e9 100644
--- a/proto/ospf/rt.c
+++ b/proto/ospf/rt.c
@@ -36,11 +36,9 @@ unresolved_vlink(ort *ort)
 static inline struct nexthop *
 new_nexthop(struct ospf_proto *p, ip_addr gw, struct iface *iface, byte weight)
 {
-  struct nexthop *nh = lp_alloc(p->nhpool, sizeof(struct nexthop));
-  nh->labels = 0;
+  struct nexthop *nh = lp_allocz(p->nhpool, sizeof(struct nexthop));
   nh->gw = gw;
   nh->iface = iface;
-  nh->next = NULL;
   nh->weight = weight;
   return nh;
 }
@@ -1907,7 +1905,6 @@ ort_changed(ort *nf, rta *nr)
     (nf->n.metric1 != nf->old_metric1) || (nf->n.metric2 != nf->old_metric2) ||
     (nf->n.tag != nf->old_tag) || (nf->n.rid != nf->old_rid) ||
     (nr->source != or->source) || (nr->dest != or->dest) ||
-    (nr->nh.iface != or->nh.iface) || !ipa_equal(nr->nh.gw, or->nh.gw) ||
     !nexthop_same(&(nr->nh), &(or->nh));
 }
 
@@ -1952,11 +1949,10 @@ again1:
 	.src = p->p.main_source,
 	.source = nf->n.type,
 	.scope = SCOPE_UNIVERSE,
+	.dest = RTD_UNICAST,
+	.nh = *(nf->n.nhs),
       };
 
-      nexthop_link(&a0, nf->n.nhs);
-      a0.dest = RTD_UNICAST;
-
       if (reload || ort_changed(nf, &a0))
       {
 	rta *a = rta_lookup(&a0);
diff --git a/proto/pipe/pipe.c b/proto/pipe/pipe.c
index a4d371fe..310f3c01 100644
--- a/proto/pipe/pipe.c
+++ b/proto/pipe/pipe.c
@@ -43,8 +43,6 @@
 
 #include "pipe.h"
 
-#include <alloca.h>
-
 static void
 pipe_rt_notify(struct proto *P, struct channel *src_ch, net *n, rte *new, rte *old, ea_list *attrs)
 {
diff --git a/proto/rip/rip.c b/proto/rip/rip.c
index 9bed9249..157093aa 100644
--- a/proto/rip/rip.c
+++ b/proto/rip/rip.c
@@ -147,20 +147,16 @@ rip_announce_rte(struct rip_proto *p, struct rip_entry *en)
       .src = p->p.main_source,
       .source = RTS_RIP,
       .scope = SCOPE_UNIVERSE,
+      .dest = RTD_UNICAST,
     };
 
     u8 rt_metric = rt->metric;
     u16 rt_tag = rt->tag;
-    struct rip_rte *rt2 = rt->next;
 
-    /* Find second valid rte */
-    while (rt2 && !rip_valid_rte(rt2))
-      rt2 = rt2->next;
-
-    a0.dest = RTD_UNICAST;
-    if (p->ecmp && rt2)
+    if (p->ecmp)
     {
       /* ECMP route */
+      struct nexthop *nhs = NULL;
       int num = 0;
 
       for (rt = en->routes; rt && (num < p->ecmp); rt = rt->next)
@@ -168,28 +164,27 @@ rip_announce_rte(struct rip_proto *p, struct rip_entry *en)
 	if (!rip_valid_rte(rt))
 	    continue;
 
-	struct nexthop *nh = (a0.nh.next ? &(a0.nh) : alloca(sizeof(struct nexthop)));
+	struct nexthop *nh = allocz(sizeof(struct nexthop));
 
 	nh->gw = rt->next_hop;
 	nh->iface = rt->from->nbr->iface;
 	nh->weight = rt->from->ifa->cf->ecmp_weight;
 
-	if (a0.nh.next)
-	  nexthop_insert(&(a0.nh), nh);
-
+	nexthop_insert(&nhs, nh);
 	num++;
 
 	if (rt->tag != rt_tag)
 	  rt_tag = 0;
       }
+
+      a0.nh = *nhs;
     }
     else
     {
       /* Unipath route */
-      a0.nh.next = NULL;
+      a0.from = rt->from->nbr->addr;
       a0.nh.gw = rt->next_hop;
       a0.nh.iface = rt->from->nbr->iface;
-      a0.from = rt->from->nbr->addr;
     }
 
     rta *a = rta_lookup(&a0);
diff --git a/proto/rpki/rpki.c b/proto/rpki/rpki.c
index 81268e83..497edd3c 100644
--- a/proto/rpki/rpki.c
+++ b/proto/rpki/rpki.c
@@ -124,7 +124,7 @@ rpki_table_add_roa(struct rpki_cache *cache, struct channel *channel, const net_
     .src = p->p.main_source,
     .source = RTS_RPKI,
     .scope = SCOPE_UNIVERSE,
-    .dest = RTD_BLACKHOLE,
+    .dest = RTD_NONE,
   };
 
   rta *a = rta_lookup(&a0);
diff --git a/proto/static/config.Y b/proto/static/config.Y
index 2fb54448..16c276ce 100644
--- a/proto/static/config.Y
+++ b/proto/static/config.Y
@@ -13,9 +13,33 @@ CF_HDR
 CF_DEFINES
 
 #define STATIC_CFG ((struct static_config *) this_proto)
-static struct static_route *this_srt, *last_srt;
+static struct static_route *this_srt, *this_snh;
 static struct f_inst **this_srt_last_cmd;
 
+static struct static_route *
+static_nexthop_new(void)
+{
+  struct static_route *nh;
+
+  if (!this_snh)
+  {
+    /* First next hop */
+    nh = this_srt;
+    rem_node(&this_srt->n);
+  }
+  else
+  {
+    /* Additional next hop */
+    nh = cfg_allocz(sizeof(struct static_route));
+    nh->net = this_srt->net;
+    this_snh->mp_next = nh;
+  }
+
+  nh->dest = RTD_UNICAST;
+  nh->mp_head = this_srt;
+  return nh;
+};
+
 static void
 static_route_finish(void)
 { }
@@ -45,48 +69,35 @@ static_proto:
  | static_proto stat_route stat_route_opt_list ';' { static_route_finish(); }
  ;
 
-stat_nexthop_via: VIA
-{
-  if (last_srt)
-  {
-    last_srt = (last_srt->mp_next = cfg_allocz(sizeof(struct static_route)));
-    last_srt->net = this_srt->net;
-  }
-  else
-  {
-    last_srt = this_srt;
-    rem_node(&this_srt->n);
-  }
-
-  last_srt->mp_head = this_srt;
-  last_srt->dest = RTD_UNICAST;
-};
-
-stat_nexthop_ident:
-    stat_nexthop_via ipa ipa_scope {
-      last_srt->via = $2;
-      last_srt->iface = $3;
-      add_tail(&STATIC_CFG->neigh_routes, &last_srt->n);
+stat_nexthop:
+    VIA ipa ipa_scope {
+      this_snh = static_nexthop_new();
+      this_snh->via = $2;
+      this_snh->iface = $3;
+      add_tail(&STATIC_CFG->neigh_routes, &this_snh->n);
     }
-  | stat_nexthop_via TEXT {
-      last_srt->via = IPA_NONE;
-      last_srt->if_name = $2;
-      add_tail(&STATIC_CFG->iface_routes, &last_srt->n);
+  | VIA TEXT {
+      this_snh = static_nexthop_new();
+      this_snh->via = IPA_NONE;
+      this_snh->if_name = $2;
+      add_tail(&STATIC_CFG->iface_routes, &this_snh->n);
     }
-  | stat_nexthop_ident MPLS label_stack {
-    last_srt->label_count = $3[0];
-    last_srt->label_stack = &($3[1]);
+  | stat_nexthop MPLS label_stack {
+    this_snh->label_count = $3[0];
+    this_snh->label_stack = &($3[1]);
   }
-  | stat_nexthop_ident WEIGHT expr {
-    last_srt->weight = $3 - 1;
+  | stat_nexthop WEIGHT expr {
+    this_snh->weight = $3 - 1;
     if (($3<1) || ($3>256)) cf_error("Weight must be in range 1-256");
   }
-  | stat_nexthop_ident BFD bool { last_srt->use_bfd = $3; cf_check_bfd($3); }
+  | stat_nexthop BFD bool {
+    this_snh->use_bfd = $3; cf_check_bfd($3);
+  }
 ;
 
-stat_nexthop:
-    stat_nexthop_ident
-  | stat_nexthop stat_nexthop_ident
+stat_nexthops:
+    stat_nexthop
+  | stat_nexthops stat_nexthop
 ;
 
 stat_route0: ROUTE net_any {
@@ -95,12 +106,12 @@ stat_route0: ROUTE net_any {
      this_srt->net = $2;
      this_srt_last_cmd = &(this_srt->cmds);
      this_srt->mp_next = NULL;
-     last_srt = NULL;
+     this_snh = NULL;
   }
  ;
 
 stat_route:
-   stat_route0 stat_nexthop
+   stat_route0 stat_nexthops
  | stat_route0 RECURSIVE ipa {
       this_srt->dest = RTDX_RECURSIVE;
       this_srt->via = $3;
diff --git a/proto/static/static.c b/proto/static/static.c
index 3e03708c..e5251bf6 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -60,54 +60,44 @@ p_igp_table(struct proto *p)
 static void
 static_install(struct proto *p, struct static_route *r)
 {
-  rta *ap = alloca(RTA_MAX_SIZE);
+  rta *ap = allocz(RTA_MAX_SIZE);
   rte *e;
 
   if (!(r->state & STS_WANT) && (r->state & (STS_INSTALLED | STS_FORCE)) && r->dest != RTD_UNICAST)
     goto drop;
 
   DBG("Installing static route %N, rtd=%d\n", r->net, r->dest);
-  bzero(ap, RTA_MAX_SIZE);
   ap->src = p->main_source;
-  ap->source = ((r->dest == RTD_UNICAST) && ipa_zero(r->via)) ? RTS_STATIC_DEVICE : RTS_STATIC;
+  ap->source = RTS_STATIC;
   ap->scope = SCOPE_UNIVERSE;
   ap->dest = r->dest;
 
   if (r->dest == RTD_UNICAST)
     {
+      struct nexthop *nhs = NULL;
       struct static_route *r2;
-      int num = 0, update = 0;
+      int update = 0;
 
+      r = r->mp_head;
       for (r2 = r; r2; r2 = r2->mp_next)
       {
-
 	if ((r2->state & STS_FORCE) ||
 	    (!!(r2->state & STS_INSTALLED) != !!(r2->state & STS_WANT)))
 	  update++;
 
 	if (r2->state & STS_WANT)
-	  {
-	    struct nexthop *nh = (ap->nh.next) ? alloca(NEXTHOP_MAX_SIZE) : &(ap->nh);
-	    if (ipa_zero(r2->via)) // Device nexthop
-	      {
-		nh->gw = IPA_NONE;
-		nh->iface = r2->iface;
-	      }
-	    else // Router nexthop
-	      {
-		nh->gw = r2->via;
-		nh->iface = r2->neigh->iface;
-	      }
-	    nh->weight = r2->weight;
-	    nh->labels = r2->label_count;
-	    for (int i=0; i<nh->labels; i++)
-	      nh->label[i] = r2->label_stack[i];
-
-	    if (ap->nh.next)
-	      nexthop_insert(&(ap->nh), nh);
-	    r2->state |= STS_INSTALLED;
-	    num++;
-	  }
+	{
+	  struct nexthop *nh = allocz(NEXTHOP_MAX_SIZE);
+
+	  nh->gw = r2->via;
+	  nh->iface = r2->neigh ? r2->neigh->iface : r2->iface;
+	  nh->weight = r2->weight;
+	  nh->labels = r2->label_count;
+	  memcpy(nh->label, r2->label_stack, r2->label_count * sizeof(u32));
+
+	  r2->state |= STS_INSTALLED;
+	  nexthop_insert(&nhs, nh);
+	}
 	else
 	  r2->state = 0;
       }
@@ -115,18 +105,19 @@ static_install(struct proto *p, struct static_route *r)
       if (!update) // Nothing changed
 	return;
 
-      r = r->mp_head;
-
-      if (!num) // No nexthop to install
+      if (!nhs) // No nexthop to install
       {
 drop:
 	rte_update(p, r->net, NULL);
 	return;
       }
+
+      ap->dest = RTD_UNICAST;
+      nexthop_link(ap, nhs);
     }
   else
     r->state |= STS_INSTALLED;
-  
+
   if (r->dest == RTDX_RECURSIVE)
     {
       ap->nh.labels_append = ap->nh.labels = r->label_count;
diff --git a/sysdep/bsd/krt-sock.c b/sysdep/bsd/krt-sock.c
index 2e0f194b..c65cba65 100644
--- a/sysdep/bsd/krt-sock.c
+++ b/sysdep/bsd/krt-sock.c
@@ -193,7 +193,6 @@ krt_send_route(struct krt_proto *p, int cmd, rte *e)
   struct ks_msg msg;
   char *body = (char *)msg.buf;
   sockaddr gate, mask, dst;
-  ip_addr gw;
 
   DBG("krt-sock: send %I/%d via %I\n", net->n.prefix, net->n.pxlen, a->gw);
 
@@ -223,14 +222,12 @@ krt_send_route(struct krt_proto *p, int cmd, rte *e)
     msg.rtm.rtm_flags |= RTF_BLACKHOLE;
 #endif
 
-  /* This is really very nasty, but I'm not able
-   * to add "(reject|blackhole)" route without
-   * gateway set
+  /*
+   * This is really very nasty, but I'm not able to add reject/blackhole route
+   * without gateway address.
    */
-  if(!i)
+  if (!i)
   {
-    i = HEAD(iface_list);
-
     WALK_LIST(j, iface_list)
     {
       if (j->flags & IF_LOOPBACK)
@@ -239,13 +236,13 @@ krt_send_route(struct krt_proto *p, int cmd, rte *e)
         break;
       }
     }
-  }
 
-  gw = a->nh.gw;
-
-  /* Embed interface ID to link-local address */
-  if (ipa_is_link_local(gw))
-    _I0(gw) = 0xfe800000 | (i->index & 0x0000ffff);
+    if (!i)
+    {
+      log(L_ERR "KRT: Cannot find loopback iface");
+      return -1;
+    }
+  }
 
   int af = AF_UNSPEC;
 
@@ -261,45 +258,51 @@ krt_send_route(struct krt_proto *p, int cmd, rte *e)
       return -1;
   }
 
-
   sockaddr_fill(&dst,  af, net_prefix(net->n.addr), NULL, 0);
   sockaddr_fill(&mask, af, net_pxmask(net->n.addr), NULL, 0);
-  sockaddr_fill(&gate, af, gw, NULL, 0);
 
   switch (a->dest)
   {
-    case RTD_UNICAST:
-      if (ipa_zero(gw))
-      {
-	if(i)
-	{
-#ifdef RTF_CLONING
-	  if (cmd == RTM_ADD && (i->flags & IF_MULTIACCESS) != IF_MULTIACCESS)	/* PTP */
-	    msg.rtm.rtm_flags |= RTF_CLONING;
-#endif
+  case RTD_UNICAST:
+    if (ipa_nonzero(a->nh.gw))
+    {
+      ip_addr gw = a->nh.gw;
 
-	  if(!i->addr) {
-	    log(L_ERR "KRT: interface %s has no IP addess", i->name);
-	    return -1;
-	  }
-
-	  sockaddr_fill(&gate, ipa_is_ip4(i->addr->ip) ? AF_INET : AF_INET6, i->addr->ip, NULL, 0);
-	  msg.rtm.rtm_addrs |= RTA_GATEWAY;
-	}
-      } else {
-	msg.rtm.rtm_flags |= RTF_GATEWAY;
-	msg.rtm.rtm_addrs |= RTA_GATEWAY;
-      }
+      /* Embed interface ID to link-local address */
+      if (ipa_is_link_local(gw))
+	_I0(gw) = 0xfe800000 | (i->index & 0x0000ffff);
+
+      sockaddr_fill(&gate, af, gw, NULL, 0);
+      msg.rtm.rtm_flags |= RTF_GATEWAY;
+      msg.rtm.rtm_addrs |= RTA_GATEWAY;
       break;
+    }
 
 #ifdef RTF_REJECT
-    case RTD_UNREACHABLE:
+  case RTD_UNREACHABLE:
 #endif
 #ifdef RTF_BLACKHOLE
-    case RTD_BLACKHOLE:
+  case RTD_BLACKHOLE:
 #endif
-    default:
-      bug("krt-sock: unknown flags, but not filtered");
+  {
+    /* Fallback for all other valid cases */
+    if (!i->addr)
+    {
+      log(L_ERR "KRT: interface %s has no IP addess", i->name);
+      return -1;
+    }
+
+#ifdef RTF_CLONING
+    if (cmd == RTM_ADD && (i->flags & IF_MULTIACCESS) != IF_MULTIACCESS)	/* PTP */
+      msg.rtm.rtm_flags |= RTF_CLONING;
+#endif
+
+    sockaddr_fill(&gate, ipa_is_ip4(i->addr->ip) ? AF_INET : AF_INET6, i->addr->ip, NULL, 0);
+    msg.rtm.rtm_addrs |= RTA_GATEWAY;
+  }
+
+  default:
+    bug("krt-sock: unknown flags, but not filtered");
   }
 
   msg.rtm.rtm_index = i->index;
@@ -497,7 +500,6 @@ krt_read_route(struct ks_msg *msg, struct krt_proto *p, int scan)
     }
 
   a.dest = RTD_UNICAST;
-  a.nh.next = NULL;
   if (flags & RTF_GATEWAY)
   {
     neighbor *ng;
@@ -520,8 +522,6 @@ krt_read_route(struct ks_msg *msg, struct krt_proto *p, int scan)
 	return;
       }
   }
-  else
-    a.nh.gw = IPA_NONE;
 
  done:
   e = rte_get_temp(&a);
diff --git a/sysdep/linux/netlink.c b/sysdep/linux/netlink.c
index 23431172..80439c47 100644
--- a/sysdep/linux/netlink.c
+++ b/sysdep/linux/netlink.c
@@ -320,6 +320,7 @@ static struct nl_want_attrs ifa_attr_want4[BIRD_IFA_MAX] = {
   [IFA_ADDRESS]	  = { 1, 1, sizeof(ip4_addr) },
   [IFA_LOCAL]	  = { 1, 1, sizeof(ip4_addr) },
   [IFA_BROADCAST] = { 1, 1, sizeof(ip4_addr) },
+  [IFA_FLAGS]     = { 1, 1, sizeof(u32) },
 };
 
 static struct nl_want_attrs ifa_attr_want6[BIRD_IFA_MAX] = {
@@ -543,18 +544,17 @@ nl_add_attr_via(struct nlmsghdr *h, uint bufsize, ip_addr ipa)
 
   h->nlmsg_len += sizeof(*via);
 
-  if (ipa_is_ip4(ipa)) {
-    ip4_addr ip4 = ipa_to_ip4(ipa);
-    ip4 = ip4_hton(ip4);
+  if (ipa_is_ip4(ipa))
+  {
     via->rtvia_family = AF_INET;
-    memcpy(via->rtvia_addr, &ip4, sizeof(ip4));
-    h->nlmsg_len += sizeof(ip4);
-  } else {
-    ip6_addr ip6 = ipa_to_ip6(ipa);
-    ip6 = ip6_hton(ip6);
+    put_ip4(via->rtvia_addr, ipa_to_ip4(ipa));
+    h->nlmsg_len += sizeof(ip4_addr);
+  }
+  else
+  {
     via->rtvia_family = AF_INET6;
-    memcpy(via->rtvia_addr, &ip6, sizeof(ip6));
-    h->nlmsg_len += sizeof(ip6);
+    put_ip6(via->rtvia_addr, ipa_to_ip6(ipa));
+    h->nlmsg_len += sizeof(ip6_addr);
   }
 
   nl_close_attr(h, nest);
@@ -669,6 +669,7 @@ nl_parse_multipath(struct krt_proto *p, struct rtattr *ra)
 	}
       else
 	rv->gw = IPA_NONE;
+
       if (a[RTA_ENCAP_TYPE])
 	{
 	  if (rta_get_u16(a[RTA_ENCAP_TYPE]) != LWTUNNEL_ENCAP_MPLS) {
@@ -1092,20 +1093,16 @@ krt_capable(rte *e)
   rta *a = e->attrs;
 
   switch (a->dest)
-    {
+  {
     case RTD_UNICAST:
-      for (struct nexthop *nh = &(a->nh); nh; nh = nh->next)
-	if (nh->iface)
-	  return 1;
-      return 0;
     case RTD_BLACKHOLE:
     case RTD_UNREACHABLE:
     case RTD_PROHIBIT:
-      break;
+      return 1;
+
     default:
       return 0;
-    }
-  return 1;
+  }
 }
 
 static inline int
@@ -1210,7 +1207,6 @@ nl_send_route(struct krt_proto *p, rte *e, struct ea_list *eattrs, int op, int d
 
 
 dest:
-  /* a->iface != NULL checked in krt_capable() for router and device routes */
   switch (dest)
     {
     case RTD_UNICAST:
@@ -1502,6 +1498,7 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
   switch (i->rtm_type)
     {
     case RTN_UNICAST:
+      ra->dest = RTD_UNICAST;
 
       if (a[RTA_MULTIPATH] && (i->rtm_family == AF_INET))
 	{
@@ -1512,7 +1509,7 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
 	      return;
 	    }
 
-	  nexthop_link(ra, nh);
+	  ra->nh = *nh;
 	  break;
 	}
 
@@ -1698,9 +1695,20 @@ nl_parse_route(struct nl_parse_state *s, struct nlmsghdr *h)
   else
   {
     /* Merge next hops with the stored route */
-    rta *a = s->attrs;
+    rta *oa = s->attrs;
 
-    nexthop_insert(&a->nh, &ra->nh);
+    struct nexthop *nhs = &oa->nh;
+    nexthop_insert(&nhs, &ra->nh);
+
+    /* Perhaps new nexthop is inserted at the first position */
+    if (nhs == &ra->nh)
+    {
+      /* Swap rtas */
+      s->attrs = ra;
+
+      /* Keep old eattrs */
+      ra->eattrs = oa->eattrs;
+    }
   }
 }
 
diff --git a/sysdep/unix/krt.c b/sysdep/unix/krt.c
index c273cb10..c6ff6275 100644
--- a/sysdep/unix/krt.c
+++ b/sysdep/unix/krt.c
@@ -984,7 +984,7 @@ krt_store_tmp_attrs(rte *rt, struct ea_list *attrs)
 static int
 krt_import_control(struct proto *P, rte **new, ea_list **attrs UNUSED, struct linpool *pool UNUSED)
 {
-  struct krt_proto *p = (struct krt_proto *) P;
+  // struct krt_proto *p = (struct krt_proto *) P;
   rte *e = *new;
 
   if (e->attrs->src->proto == P)
@@ -1005,17 +1005,6 @@ krt_import_control(struct proto *P, rte **new, ea_list **attrs UNUSED, struct li
     return -1;
   }
 
-  if (!KRT_CF->devroutes && (e->attrs->source != RTS_STATIC_DEVICE))
-  {
-    struct nexthop *nh = &(e->attrs->nh);
-    for (; nh; nh = nh->next)
-      if (ipa_nonzero(nh->gw))
-	break;
-
-    if (!nh) /* Gone through all the nexthops and no explicit GW found */
-      return -1;
-  }
-
   if (!krt_capable(e))
     return -1;
 
-- 
cgit v1.2.3


From 93f50ca31757fc8e416093e0c73681e070294a3d Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Wed, 22 Feb 2017 14:02:03 +0100
Subject: Nest: names for nhu_state values

It took too much time to analyze what's the meaning of nhu_state values
so I spent less than the same amount of time documenting it.
---
 nest/route.h    |  5 +++++
 nest/rt-table.c | 22 ++++++++++++++--------
 2 files changed, 19 insertions(+), 8 deletions(-)

(limited to 'nest/route.h')

diff --git a/nest/route.h b/nest/route.h
index c16b2643..928a022d 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -168,6 +168,11 @@ typedef struct rtable {
   struct fib_iterator nhu_fit;		/* Next Hop Update FIB iterator */
 } rtable;
 
+#define NHU_CLEAN	0
+#define NHU_SCHEDULED	1
+#define NHU_RUNNING	2
+#define NHU_DIRTY	3
+
 typedef struct network {
   struct rte *routes;			/* Available routes for this network */
   struct fib_node n;			/* FIB flags reserved for kernel syncer */
diff --git a/nest/rt-table.c b/nest/rt-table.c
index a33b7909..1e1dde25 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -1561,11 +1561,14 @@ rt_schedule_hcu(rtable *tab)
 static inline void
 rt_schedule_nhu(rtable *tab)
 {
-  if (tab->nhu_state == 0)
+  if (tab->nhu_state == NHU_CLEAN)
     ev_schedule(tab->rt_event);
 
-  /* state change 0->1, 2->3 */
-  tab->nhu_state |= 1;
+  /* state change:
+   *   NHU_CLEAN   -> NHU_SCHEDULED
+   *   NHU_RUNNING -> NHU_DIRTY
+   */
+  tab->nhu_state |= NHU_SCHEDULED;
 }
 
 void
@@ -1897,13 +1900,13 @@ rt_next_hop_update(rtable *tab)
   struct fib_iterator *fit = &tab->nhu_fit;
   int max_feed = 32;
 
-  if (tab->nhu_state == 0)
+  if (tab->nhu_state == NHU_CLEAN)
     return;
 
-  if (tab->nhu_state == 1)
+  if (tab->nhu_state == NHU_SCHEDULED)
     {
       FIB_ITERATE_INIT(fit, &tab->fib);
-      tab->nhu_state = 2;
+      tab->nhu_state = NHU_RUNNING;
     }
 
   FIB_ITERATE_START(&tab->fib, fit, net, n)
@@ -1918,10 +1921,13 @@ rt_next_hop_update(rtable *tab)
     }
   FIB_ITERATE_END;
 
-  /* state change 2->0, 3->1 */
+  /* State change:
+   *   NHU_DIRTY   -> NHU_SCHEDULED
+   *   NHU_RUNNING -> NHU_CLEAN
+   */
   tab->nhu_state &= 1;
 
-  if (tab->nhu_state > 0)
+  if (tab->nhu_state != NHU_CLEAN)
     ev_schedule(tab->rt_event);
 }
 
-- 
cgit v1.2.3


From 039a65d0e4f33f8432caae78cd919d2fd2052eea Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Fri, 24 Feb 2017 14:05:11 +0100
Subject: Nexthop: Fixed hostentry

---
 nest/route.h          |   6 +--
 nest/rt-attr.c        |   2 +-
 nest/rt-table.c       | 105 ++++++++++++++++++++++++--------------------------
 proto/bgp/packets.c   |   2 +-
 proto/static/static.c |   2 +-
 5 files changed, 57 insertions(+), 60 deletions(-)

(limited to 'nest/route.h')

diff --git a/nest/route.h b/nest/route.h
index 928a022d..98bef1fd 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -200,8 +200,8 @@ struct hostentry {
   unsigned hash_key;			/* Hash key */
   unsigned uc;				/* Use count */
   struct rta *src;			/* Source rta entry */
-  struct nexthop *nh;			/* Chosen next hop */
   byte dest;				/* Chosen route destination type (RTD_...) */
+  byte nexthop_linkable;		/* Nexthop list is completely non-device */
   u32 igp_metric;			/* Chosen route IGP metric */
 };
 
@@ -344,8 +344,8 @@ struct nexthop {
   struct iface *iface;			/* Outgoing interface */
   struct nexthop *next;
   byte weight;
-  byte labels_append;			/* Number of labels before hostentry was applied */
-  byte labels;				/* Number of labels prepended */
+  byte labels_orig;			/* Number of labels before hostentry was applied */
+  byte labels;				/* Number of all labels */
   u32 label[0];
 };
 
diff --git a/nest/rt-attr.c b/nest/rt-attr.c
index afc97e22..2c8ee7db 100644
--- a/nest/rt-attr.c
+++ b/nest/rt-attr.c
@@ -1155,12 +1155,12 @@ rta__free(rta *a)
   *a->pprev = a->next;
   if (a->next)
     a->next->pprev = a->pprev;
-  a->aflags = 0;		/* Poison the entry */
   rt_unlock_hostentry(a->hostentry);
   rt_unlock_source(a->src);
   if (a->nh.next)
     nexthop_free(a->nh.next);
   ea_free(a->eattrs);
+  a->aflags = 0;		/* Poison the entry */
   sl_free(rta_slab(a), a);
 }
 
diff --git a/nest/rt-table.c b/nest/rt-table.c
index 1e1dde25..8765d293 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -1764,7 +1764,7 @@ rta_next_hop_outdated(rta *a)
     return a->dest != RTD_UNREACHABLE;
 
   return (a->dest != he->dest) || (a->igp_metric != he->igp_metric) ||
-    !nexthop_same(&(a->nh), he->nh);
+    (!he->nexthop_linkable) || !nexthop_same(&(a->nh), &(he->src->nh));
 }
 
 static inline void
@@ -1774,39 +1774,49 @@ rta_apply_hostentry(rta *a, struct hostentry *he)
   a->dest = he->dest;
   a->igp_metric = he->igp_metric;
 
-  if (a->nh.labels_append == 0)
+  if ((a->nh.labels_orig == 0) && (!a->nh.next) && he->nexthop_linkable)
   {
-    a->nh = *(he->nh);
-    a->nh.labels_append = 0;
+    a->nh = he->src->nh;
     return;
   }
 
-  int labels_append = a->nh.labels_append;
-  u32 label_stack[MPLS_MAX_LABEL_STACK];
-  memcpy(label_stack, a->nh.label, labels_append * sizeof(u32));
-
-  struct nexthop *nhp = NULL;
-  for (struct nexthop *nh = he->nh; nh; nh = nh->next)
+  struct nexthop *nhp = alloca(NEXTHOP_MAX_SIZE);
+  
+  for (struct nexthop *nhe = &(a->nh); nhe; nhe = nhe->next)
   {
-    nhp = nhp ? (nhp->next = lp_alloc(rte_update_pool, NEXTHOP_MAX_SIZE)) : &(a->nh);
-    nhp->gw = ipa_nonzero(nh->gw) ? nh->gw : he->link;
-    nhp->iface = nh->iface; /* FIXME: This is at least strange, if not utter nonsense. */
-    nhp->weight = nh->weight;
-    nhp->labels = nh->labels + labels_append;
-    nhp->labels_append = labels_append;
-    if (nhp->labels <= MPLS_MAX_LABEL_STACK)
-    {
-      memcpy(nhp->label, nh->label, nh->labels * sizeof(u32));
-      memcpy(&(nhp->label[nh->labels]), label_stack, labels_append * sizeof(u32));
-    }
-    else
+    int labels_orig = nhe->labels_orig;	    /* Number of labels (at the bottom of stack) */
+    u32 label_stack[MPLS_MAX_LABEL_STACK];
+    memcpy(label_stack, nhe->label, labels_orig * sizeof(u32));
+
+    for (struct nexthop *nh = &(he->src->nh); nh; nh = nh->next)
     {
-      log(L_WARN "Sum of label stack sizes %d + %d = %d exceedes allowed maximum (%d)",
-	  nh->labels, labels_append, nhp->labels, MPLS_MAX_LABEL_STACK);
-      a->dest = RTD_UNREACHABLE;
-      break;
+      nhp->iface = nh->iface;
+      nhp->weight = nh->weight; /* FIXME: Ignoring the recursive nexthop's weight */
+      nhp->labels = nh->labels + labels_orig;
+      nhp->labels_orig = labels_orig;
+      if (nhp->labels <= MPLS_MAX_LABEL_STACK)
+      {
+	memcpy(nhp->label, nh->label, nh->labels * sizeof(u32)); /* First the hostentry labels */
+	memcpy(&(nhp->label[nh->labels]), label_stack, labels_orig * sizeof(u32)); /* Then the bottom labels */
+      }
+      else
+      {
+	log(L_WARN "Sum of label stack sizes %d + %d = %d exceedes allowed maximum (%d)",
+	    nh->labels, labels_orig, nhp->labels, MPLS_MAX_LABEL_STACK);
+	continue;
+      }
+      if (ipa_nonzero(nh->gw))
+	nhp->gw = nh->gw;		/* Router nexthop */
+      else if (ipa_nonzero(he->link))
+	nhp->gw = he->link;		/* Device nexthop with link-local address known */
+      else
+	nhp->gw = he->addr;		/* Device nexthop with link-local address unknown */
+
+      nhp = (nhp->next = lp_alloc(rte_update_pool, NEXTHOP_MAX_SIZE));
     }
   }
+
+  memcpy(&(a->nh), nhp, nexthop_size(nhp));
 }
 
 static inline rte *
@@ -2231,12 +2241,12 @@ hc_new_hostentry(struct hostcache *hc, ip_addr a, ip_addr ll, rtable *dep, unsig
 {
   struct hostentry *he = sl_alloc(hc->slab);
 
-  he->addr = a;
-  he->link = ll;
-  he->tab = dep;
-  he->hash_key = k;
-  he->uc = 0;
-  he->src = NULL;
+  *he = (struct hostentry) {
+    .addr = a,
+    .link = ll,
+    .tab = dep,
+    .hash_key = k,
+  };
 
   add_tail(&hc->hostentries, &he->ln);
   hc_insert(hc, he);
@@ -2357,6 +2367,7 @@ rt_update_hostentry(rtable *tab, struct hostentry *he)
 
   /* Reset the hostentry */
   he->src = NULL;
+  he->nexthop_linkable = 0;
   he->dest = RTD_UNREACHABLE;
   he->igp_metric = 0;
 
@@ -2377,28 +2388,14 @@ rt_update_hostentry(rtable *tab, struct hostentry *he)
 	  goto done;
 	}
 
-      if ((a->dest == RTD_UNICAST) && ipa_zero(a->nh.gw) && !a->next)
-	{
-	  /* We have singlepath device route */
-	  if (if_local_addr(he->addr, a->nh.iface))
-	    {
-	      /* The host address is a local address, this is not valid */
-	      log(L_WARN "Next hop address %I is a local address of iface %s",
-		  he->addr, a->nh.iface->name);
-	      goto done;
-      	    }
-
-	  /* The host is directly reachable, use link as a gateway */
-	  he->nh = NULL;
-	  he->dest = RTD_UNICAST;
-	}
-      else
-	{
-	  /* The host is reachable through some route entry */
-	  he->nh = &(a->nh);
-	  he->dest = a->dest;
-	}
-
+      he->nexthop_linkable = 1;
+      for (struct nexthop *nh = &(a->nh); nh; nh = nh->next)
+	if (ipa_zero(nh->gw))
+	  {
+	    he->nexthop_linkable = 0;
+	    break;
+	  }
+  
       he->src = rta_clone(a);
       he->igp_metric = rt_get_igp_metric(e);
     }
diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index 2106e0d1..bee9248a 100644
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -1705,7 +1705,7 @@ bgp_decode_nlri(struct bgp_parse_state *s, u32 afi, byte *nlri, uint len, ea_lis
 
   if (ea)
   {
-    a = allocz(sizeof(struct rta));
+    a = allocz(RTA_MAX_SIZE);
 
     a->source = RTS_BGP;
     a->scope = SCOPE_UNIVERSE;
diff --git a/proto/static/static.c b/proto/static/static.c
index e5251bf6..63ee2518 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -120,7 +120,7 @@ drop:
 
   if (r->dest == RTDX_RECURSIVE)
     {
-      ap->nh.labels_append = ap->nh.labels = r->label_count;
+      ap->nh.labels_orig = ap->nh.labels = r->label_count;
       memcpy(ap->nh.label, r->label_stack, r->label_count * sizeof(u32));
       rta_set_recursive_next_hop(p->main_channel->table, ap, p_igp_table(p), r->via, IPA_NONE);
     }
-- 
cgit v1.2.3


From 665be7f6bdbf1fd8dbac45cef533bd4b1df35d4d Mon Sep 17 00:00:00 2001
From: "Ondrej Zajicek (work)" <santiago@crfreenet.org>
Date: Wed, 8 Mar 2017 16:27:18 +0100
Subject: Nest: Minor fixes in show route

---
 nest/route.h          |  6 ++++++
 nest/rt-attr.c        |  8 ++++++++
 nest/rt-table.c       | 49 +++++++++++++++++--------------------------------
 proto/static/static.c |  9 ---------
 4 files changed, 31 insertions(+), 41 deletions(-)

(limited to 'nest/route.h')

diff --git a/nest/route.h b/nest/route.h
index 98bef1fd..65711138 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -401,6 +401,7 @@ typedef struct rta {
 #define RTD_BLACKHOLE 2			/* Silently drop packets */
 #define RTD_UNREACHABLE 3		/* Reject as unreachable */
 #define RTD_PROHIBIT 4			/* Administratively prohibited */
+#define RTD_MAX 5
 
 					/* Flags for net->n.flags, used by kernel syncer */
 #define KRF_INSTALLED 0x80		/* This route should be installed in the kernel */
@@ -412,6 +413,11 @@ typedef struct rta {
 					   protocol-specific metric is availabe */
 
 
+const char * rta_dest_names[RTD_MAX];
+
+static inline const char *rta_dest_name(uint n)
+{ return (n < RTD_MAX) ? rta_dest_names[n] : "???"; }
+
 /* Route has regular, reachable nexthop (i.e. not RTD_UNREACHABLE and like) */
 static inline int rte_is_reachable(rte *r)
 { return r->attrs->dest == RTD_UNICAST; }
diff --git a/nest/rt-attr.c b/nest/rt-attr.c
index 2c8ee7db..1b7f5836 100644
--- a/nest/rt-attr.c
+++ b/nest/rt-attr.c
@@ -58,6 +58,14 @@
 
 #include <stddef.h>
 
+const char * rta_dest_names[RTD_MAX] = {
+  [RTD_NONE]		= "",
+  [RTD_UNICAST]		= "unicast",
+  [RTD_BLACKHOLE]	= "blackhole",
+  [RTD_UNREACHABLE]	= "unreachable",
+  [RTD_PROHIBIT]	= "prohibited",
+};
+
 pool *rta_pool;
 
 static slab *rta_slab_[4];
diff --git a/nest/rt-table.c b/nest/rt-table.c
index 8765d293..92db6cc8 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -50,7 +50,6 @@ static linpool *rte_update_pool;
 
 static list routing_tables;
 
-static byte *rt_format_via(rte *e);
 static void rt_free_hostcache(rtable *tab);
 static void rt_notify_hostcache(rtable *tab, net *net);
 static void rt_update_hostcache(rtable *tab);
@@ -346,7 +345,7 @@ rte_mergable(rte *pri, rte *sec)
 static void
 rte_trace(struct proto *p, rte *e, int dir, char *msg)
 {
-  log(L_TRACE "%s %c %s %N %s", p->name, dir, msg, e->net->n.addr, rt_format_via(e));
+  log(L_TRACE "%s %c %s %N %s", p->name, dir, msg, e->net->n.addr, rta_dest_name(e->attrs->dest));
 }
 
 static inline void
@@ -2395,12 +2394,12 @@ rt_update_hostentry(rtable *tab, struct hostentry *he)
 	    he->nexthop_linkable = 0;
 	    break;
 	  }
-  
+
       he->src = rta_clone(a);
       he->igp_metric = rt_get_igp_metric(e);
     }
 
- done:
+done:
   /* Add a prefix range to the trie */
   trie_add_prefix(tab->hostcache->trie, &he_addr, pxlen, he_addr.pxlen);
 
@@ -2465,25 +2464,6 @@ rta_set_recursive_next_hop(rtable *dep, rta *a, rtable *tab, ip_addr gw, ip_addr
  *  CLI commands
  */
 
-static byte *
-rt_format_via(rte *e)
-{
-  rta *a = e->attrs;
-
-  /* Max text length w/o IP addr and interface name is 16 */
-  static byte via[IPA_MAX_TEXT_LENGTH+sizeof(a->nh.iface->name)+16];
-
-  switch (a->dest)
-    {
-    case RTD_UNICAST:	bsprintf(via, "unicast"); break;
-    case RTD_BLACKHOLE:	bsprintf(via, "blackhole"); break;
-    case RTD_UNREACHABLE:	bsprintf(via, "unreachable"); break;
-    case RTD_PROHIBIT:	bsprintf(via, "prohibited"); break;
-    default:		bsprintf(via, "???");
-    }
-  return via;
-}
-
 static void
 rt_show_rte(struct cli *c, byte *ia, rte *e, struct rt_show_data *d, ea_list *tmpa)
 {
@@ -2515,26 +2495,31 @@ rt_show_rte(struct cli *c, byte *ia, rte *e, struct rt_show_data *d, ea_list *tm
     get_route_info(e, info, tmpa);
   else
     bsprintf(info, " (%d)", e->pref);
-  cli_printf(c, -1007, "%-18s %s [%s %s%s]%s%s", ia, rt_format_via(e), a->src->proto->name,
-	     tm, from, primary ? (sync_error ? " !" : " *") : "", info);
-  for (nh = &(a->nh); nh; nh = nh->next)
+
+  cli_printf(c, -1007, "%-18s %s [%s %s%s]%s%s", ia, rta_dest_name(a->dest),
+	     a->src->proto->name, tm, from, primary ? (sync_error ? " !" : " *") : "", info);
+
+  if (a->dest == RTD_UNICAST)
+    for (nh = &(a->nh); nh; nh = nh->next)
     {
-      char ls[MPLS_MAX_LABEL_STACK*8 + 5]; char *lsp = ls;
+      char mpls[MPLS_MAX_LABEL_STACK*12 + 5], *lsp = mpls;
+
       if (nh->labels)
-	{
+        {
 	  lsp += bsprintf(lsp, " mpls %d", nh->label[0]);
 	  for (int i=1;i<nh->labels; i++)
 	    lsp += bsprintf(lsp, "/%d", nh->label[i]);
-	  *lsp++ = '\0';
 	}
+      *lsp = '\0';
+
       if (a->nh.next)
-	cli_printf(c, -1007, "\tvia %I%s on %s weight %d", nh->gw, (nh->labels ? ls : ""), nh->iface->name, nh->weight + 1);
+	cli_printf(c, -1007, "\tvia %I%s on %s weight %d", nh->gw, mpls, nh->iface->name, nh->weight + 1);
       else
-	cli_printf(c, -1007, "\tvia %I%s on %s", nh->gw, (nh->labels ? ls : ""), nh->iface->name);
+	cli_printf(c, -1007, "\tvia %I%s on %s", nh->gw, mpls, nh->iface->name);
     }
+
   if (d->verbose)
     rta_show(c, a, tmpa);
-
 }
 
 static void
diff --git a/proto/static/static.c b/proto/static/static.c
index dbe490f9..55fd957c 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -570,15 +570,6 @@ static_copy_config(struct proto_config *dest, struct proto_config *src)
   }
 }
 
-
-static const char * rta_dest_names[] = {
-  [RTD_NONE]		= "",
-  [RTD_UNICAST]		= "unicast",
-  [RTD_BLACKHOLE]	= "blackhole",
-  [RTD_UNREACHABLE]	= "unreachable",
-  [RTD_PROHIBIT]	= "prohibited",
-};
-
 static void
 static_show_rt(struct static_route *r)
 {
-- 
cgit v1.2.3


From 3c74416465d77c0e79eeaaeb988e471663484b5d Mon Sep 17 00:00:00 2001
From: Jan Moskyto Matejka <mq@ucw.cz>
Date: Fri, 17 Mar 2017 15:48:09 +0100
Subject: Nexthop: Fixed recursive route mpls label merging

---
 conf/confbase.Y       |  15 ++++---
 lib/ip.h              |   5 +++
 nest/route.h          |   2 +-
 nest/rt-table.c       | 107 ++++++++++++++++++++++++++++++++++----------------
 proto/bgp/bgp.h       |   2 +
 proto/bgp/packets.c   |   2 +-
 proto/static/config.Y |   6 +--
 proto/static/static.c |  37 +++++++++++------
 proto/static/static.h |   3 +-
 9 files changed, 119 insertions(+), 60 deletions(-)

(limited to 'nest/route.h')

diff --git a/conf/confbase.Y b/conf/confbase.Y
index 291dc6a0..d6a6951f 100644
--- a/conf/confbase.Y
+++ b/conf/confbase.Y
@@ -65,7 +65,7 @@ CF_DECLS
   struct proto_spec ps;
   struct channel_limit cl;
   struct timeformat *tf;
-  u32 *lbl;
+  mpls_label_stack *mls;
 }
 
 %token END CLI_MARKER INVALID_TOKEN ELSECOL DDOT
@@ -85,7 +85,7 @@ CF_DECLS
 %type <a> ipa
 %type <net> net_ip4_ net_ip6_ net_ip6 net_ip_ net_ip net_or_ipa
 %type <net_ptr> net_ net_any net_vpn4_ net_vpn6_ net_vpn_ net_roa4_ net_roa6_ net_roa_
-%type <lbl> label_stack_start label_stack
+%type <mls> label_stack_start label_stack
 
 %type <t> text opttext
 
@@ -288,18 +288,17 @@ net_or_ipa:
 
 label_stack_start: NUM
 {
-  $$ = cfg_allocz(sizeof(u32) * (MPLS_MAX_LABEL_STACK+1));
-  $$[0] = 1;
-  $$[1] = $1;
+  $$ = cfg_allocz(sizeof(mpls_label_stack));
+  $$->len = 1;
+  $$->stack[0] = $1;
 };
 
 label_stack:
     label_stack_start
   | label_stack '/' NUM {
-    if ($1[0] >= MPLS_MAX_LABEL_STACK)
+    if ($1->len >= MPLS_MAX_LABEL_STACK)
       cf_error("Too many labels in stack");
-    $1[0]++;
-    $1[*$1] = $3;
+    $1->stack[$1->len++] = $3;
     $$ = $1;
   }
 ;
diff --git a/lib/ip.h b/lib/ip.h
index ab90bee7..5cfce1f1 100644
--- a/lib/ip.h
+++ b/lib/ip.h
@@ -326,6 +326,11 @@ static inline ip6_addr ip6_ntoh(ip6_addr a)
 { return _MI6(ntohl(_I0(a)), ntohl(_I1(a)), ntohl(_I2(a)), ntohl(_I3(a))); }
 
 #define MPLS_MAX_LABEL_STACK 8
+typedef struct mpls_label_stack {
+  uint len;
+  u32 stack[MPLS_MAX_LABEL_STACK];
+} mpls_label_stack;
+
 static inline int
 mpls_get(const char *buf, int buflen, u32 *stack)
 {
diff --git a/nest/route.h b/nest/route.h
index 65711138..546b04c4 100644
--- a/nest/route.h
+++ b/nest/route.h
@@ -551,7 +551,7 @@ 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, mpls_label_stack *mls);
 
 /*
  * rta_set_recursive_next_hop() acquires hostentry from hostcache and fills
diff --git a/nest/rt-table.c b/nest/rt-table.c
index 92db6cc8..f8baf572 100644
--- a/nest/rt-table.c
+++ b/nest/rt-table.c
@@ -1767,55 +1767,80 @@ rta_next_hop_outdated(rta *a)
 }
 
 static inline void
-rta_apply_hostentry(rta *a, struct hostentry *he)
+rta_apply_hostentry(rta *a, struct hostentry *he, mpls_label_stack *mls)
 {
   a->hostentry = he;
   a->dest = he->dest;
   a->igp_metric = he->igp_metric;
 
-  if ((a->nh.labels_orig == 0) && (!a->nh.next) && he->nexthop_linkable)
+  if (a->dest != RTD_UNICAST)
   {
-    a->nh = he->src->nh;
+    /* No nexthop */
+no_nexthop:
+    a->nh = (struct nexthop) {};
+    if (mls)
+    { /* Store the label stack for later changes */
+      a->nh.labels_orig = a->nh.labels = mls->len;
+      memcpy(a->nh.label, mls->stack, mls->len * sizeof(u32));
+    }
     return;
   }
 
-  struct nexthop *nhp = alloca(NEXTHOP_MAX_SIZE);
+  if (((!mls) || (!mls->len)) && he->nexthop_linkable)
+  { /* Just link the nexthop chain, no label append happens. */
+    memcpy(&(a->nh), &(he->src->nh), nexthop_size(&(he->src->nh)));
+    return;
+  }
+
+  struct nexthop *nhp = NULL, *nhr = NULL;
+  int skip_nexthop = 0;
   
-  for (struct nexthop *nhe = &(a->nh); nhe; nhe = nhe->next)
+  for (struct nexthop *nh = &(he->src->nh); nh; nh = nh->next)
   {
-    int labels_orig = nhe->labels_orig;	    /* Number of labels (at the bottom of stack) */
-    u32 label_stack[MPLS_MAX_LABEL_STACK];
-    memcpy(label_stack, nhe->label, labels_orig * sizeof(u32));
+    if (skip_nexthop)
+      skip_nexthop--;
+    else
+    {
+      nhr = nhp;
+      nhp = (nhp ? (nhp->next = lp_allocz(rte_update_pool, NEXTHOP_MAX_SIZE)) : &(a->nh));
+    }
 
-    for (struct nexthop *nh = &(he->src->nh); nh; nh = nh->next)
+    nhp->iface = nh->iface;
+    nhp->weight = nh->weight;
+    if (mls)
     {
-      nhp->iface = nh->iface;
-      nhp->weight = nh->weight; /* FIXME: Ignoring the recursive nexthop's weight */
-      nhp->labels = nh->labels + labels_orig;
-      nhp->labels_orig = labels_orig;
+      nhp->labels = nh->labels + mls->len;
+      nhp->labels_orig = mls->len;
       if (nhp->labels <= MPLS_MAX_LABEL_STACK)
       {
 	memcpy(nhp->label, nh->label, nh->labels * sizeof(u32)); /* First the hostentry labels */
-	memcpy(&(nhp->label[nh->labels]), label_stack, labels_orig * sizeof(u32)); /* Then the bottom labels */
+	memcpy(&(nhp->label[nh->labels]), mls->stack, mls->len * sizeof(u32)); /* Then the bottom labels */
       }
       else
       {
 	log(L_WARN "Sum of label stack sizes %d + %d = %d exceedes allowed maximum (%d)",
-	    nh->labels, labels_orig, nhp->labels, MPLS_MAX_LABEL_STACK);
+	    nh->labels, mls->len, nhp->labels, MPLS_MAX_LABEL_STACK);
+	skip_nexthop++;
 	continue;
       }
-      if (ipa_nonzero(nh->gw))
-	nhp->gw = nh->gw;		/* Router nexthop */
-      else if (ipa_nonzero(he->link))
-	nhp->gw = he->link;		/* Device nexthop with link-local address known */
-      else
-	nhp->gw = he->addr;		/* Device nexthop with link-local address unknown */
-
-      nhp = (nhp->next = lp_alloc(rte_update_pool, NEXTHOP_MAX_SIZE));
     }
+    if (ipa_nonzero(nh->gw))
+      nhp->gw = nh->gw;		/* Router nexthop */
+    else if (ipa_nonzero(he->link))
+      nhp->gw = he->link;		/* Device nexthop with link-local address known */
+    else
+      nhp->gw = he->addr;		/* Device nexthop with link-local address unknown */
   }
 
-  memcpy(&(a->nh), nhp, nexthop_size(nhp));
+  if (skip_nexthop)
+    if (nhr)
+      nhr->next = NULL;
+    else
+    {
+      a->dest = RTD_UNREACHABLE;
+      log(L_WARN "No valid nexthop remaining, setting route unreachable");
+      goto no_nexthop;
+    }
 }
 
 static inline rte *
@@ -1823,7 +1848,11 @@ rt_next_hop_update_rte(rtable *tab UNUSED, rte *old)
 {
   rta *a = alloca(RTA_MAX_SIZE);
   memcpy(a, old->attrs, rta_size(old->attrs));
-  rta_apply_hostentry(a, old->attrs->hostentry);
+
+  mpls_label_stack mls = { .len = a->nh.labels_orig };
+  memcpy(mls.stack, &a->nh.label[a->nh.labels - mls.len], mls.len * sizeof(u32));
+
+  rta_apply_hostentry(a, old->attrs->hostentry, &mls);
   a->aflags = 0;
 
   rte *e = sl_alloc(rte_slab);
@@ -2387,13 +2416,25 @@ rt_update_hostentry(rtable *tab, struct hostentry *he)
 	  goto done;
 	}
 
+      he->dest = a->dest;
       he->nexthop_linkable = 1;
-      for (struct nexthop *nh = &(a->nh); nh; nh = nh->next)
-	if (ipa_zero(nh->gw))
-	  {
-	    he->nexthop_linkable = 0;
-	    break;
-	  }
+      if (he->dest == RTD_UNICAST)
+	{
+	  for (struct nexthop *nh = &(a->nh); nh; nh = nh->next)
+	    if (ipa_zero(nh->gw))
+	      {
+		if (if_local_addr(he->addr, nh->iface))
+		  {
+		    /* The host address is a local address, this is not valid */
+		    log(L_WARN "Next hop address %I is a local address of iface %s",
+			he->addr, nh->iface->name);
+		    goto done;
+		  }
+
+		he->nexthop_linkable = 0;
+		break;
+	      }
+	}
 
       he->src = rta_clone(a);
       he->igp_metric = rt_get_igp_metric(e);
@@ -2454,9 +2495,9 @@ 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, mpls_label_stack *mls)
 {
-  rta_apply_hostentry(a, rt_get_hostentry(tab, gw, ipa_zero(ll) ? gw : ll, dep));
+  rta_apply_hostentry(a, rt_get_hostentry(tab, gw, ipa_zero(ll) ? gw : ll, dep), mls);
 }
 
 
diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h
index 5d2539d5..e7647625 100644
--- a/proto/bgp/bgp.h
+++ b/proto/bgp/bgp.h
@@ -337,6 +337,8 @@ struct bgp_parse_state {
   u32 mp_reach_af;
   u32 mp_unreach_af;
 
+  mpls_label_stack mls;
+
   uint attr_len;
   uint ip_reach_len;
   uint ip_unreach_len;
diff --git a/proto/bgp/packets.c b/proto/bgp/packets.c
index bee9248a..f7366804 100644
--- a/proto/bgp/packets.c
+++ b/proto/bgp/packets.c
@@ -753,7 +753,7 @@ bgp_apply_next_hop(struct bgp_parse_state *s, rta *a, ip_addr gw, ip_addr ll)
     if (ipa_zero(gw))
       WITHDRAW(BAD_NEXT_HOP);
 
-    rta_set_recursive_next_hop(c->c.table, a, c->igp_table, gw, ll);
+    rta_set_recursive_next_hop(c->c.table, a, c->igp_table, gw, ll, &(s->mls));
   }
 }
 
diff --git a/proto/static/config.Y b/proto/static/config.Y
index 86fcedec..cd8bfcec 100644
--- a/proto/static/config.Y
+++ b/proto/static/config.Y
@@ -75,8 +75,7 @@ stat_nexthop:
       this_snh->iface = if_get_by_name($2);
     }
   | stat_nexthop MPLS label_stack {
-    this_snh->label_count = $3[0];
-    this_snh->label_stack = &($3[1]);
+    this_snh->mls = $3;
   }
   | stat_nexthop WEIGHT expr {
     this_snh->weight = $3 - 1;
@@ -111,8 +110,7 @@ stat_route:
  | stat_route0 RECURSIVE ipa MPLS label_stack {
       this_srt->dest = RTDX_RECURSIVE;
       this_srt->via = $3;
-      this_srt->label_count = $5[0];
-      this_srt->label_stack = &($5[1]);
+      this_srt->mls = $5;
    }
  | stat_route0 DROP		{ this_srt->dest = RTD_BLACKHOLE; }
  | stat_route0 REJECT		{ this_srt->dest = RTD_UNREACHABLE; }
diff --git a/proto/static/static.c b/proto/static/static.c
index 55fd957c..adefa0b2 100644
--- a/proto/static/static.c
+++ b/proto/static/static.c
@@ -79,8 +79,11 @@ static_announce_rte(struct static_proto *p, struct static_route *r)
       nh->gw = r2->via;
       nh->iface = r2->neigh->iface;
       nh->weight = r2->weight;
-      nh->labels = r2->label_count;
-      memcpy(nh->label, r2->label_stack, r2->label_count * sizeof(u32));
+      if (r2->mls)
+      {
+	nh->labels = r2->mls->len;
+	memcpy(nh->label, r2->mls->stack, r2->mls->len * sizeof(u32));
+      }
 
       nexthop_insert(&nhs, nh);
     }
@@ -92,11 +95,7 @@ static_announce_rte(struct static_proto *p, struct static_route *r)
   }
 
   if (r->dest == RTDX_RECURSIVE)
-  {
-    a->nh.labels_orig = a->nh.labels = r->label_count;
-    memcpy(a->nh.label, r->label_stack, r->label_count * sizeof(u32));
-    rta_set_recursive_next_hop(p->p.main_channel->table, a, p_igp_table(p), r->via, IPA_NONE);
-  }
+    rta_set_recursive_next_hop(p->p.main_channel->table, a, p_igp_table(p), r->via, IPA_NONE, r->mls);
 
   /* Already announced */
   if (r->state == SRS_CLEAN)
@@ -274,17 +273,33 @@ static_same_dest(struct static_route *x, struct static_route *y)
 	  (x->iface != y->iface) ||
 	  (x->use_bfd != y->use_bfd) ||
 	  (x->weight != y->weight) ||
-	  (x->label_count != y->label_count))
+	  (!x->mls != !y->mls) ||
+	  ((x->mls) && (y->mls) && (x->mls->len != y->mls->len)))
 	return 0;
 
-      for (int i = 0; i < x->label_count; i++)
-	if (x->label_stack[i] != y->label_stack[i])
+      if (!x->mls)
+	continue;
+
+      for (uint i = 0; i < x->mls->len; i++)
+	if (x->mls->stack[i] != y->mls->stack[i])
 	  return 0;
     }
     return !x && !y;
 
   case RTDX_RECURSIVE:
-    return ipa_equal(x->via, y->via);
+    if (!ipa_equal(x->via, y->via) ||
+	(!x->mls != !y->mls) ||
+	((x->mls) && (y->mls) && (x->mls->len != y->mls->len)))
+      return 0;
+
+    if (!x->mls)
+      return 1;
+
+    for (uint i = 0; i < x->mls->len; i++)
+      if (x->mls->stack[i] != y->mls->stack[i])
+	return 0;
+
+    return 1;
 
   default:
     return 1;
diff --git a/proto/static/static.h b/proto/static/static.h
index bfcbd8c3..0976a9c9 100644
--- a/proto/static/static.h
+++ b/proto/static/static.h
@@ -42,9 +42,8 @@ struct static_route {
   byte active;				/* Next hop is active (nbr/iface/BFD available) */
   byte weight;				/* Multipath next hop weight */
   byte use_bfd;				/* Configured to use BFD */
-  byte label_count;			/* Number of labels in stack */
   struct bfd_request *bfd_req;		/* BFD request, if BFD is used */
-  u32 *label_stack;			/* Label stack if label_count > 0 */
+  mpls_label_stack *mls;		/* MPLS label stack; may be NULL */
 };
 
 /*
-- 
cgit v1.2.3