Merge branch 'master' of https://gitlab.nic.cz/labs/bird

Makefile.in

@@ -219,7 +219,7 @@ install-docs:
 
 # Cleanup
 clean::
-	rm -f $(objdir)/sysdep/paths.h
+	rm -f $(objdir)/sysdep/paths.h $(objdir)/nest/proto-build.c
 	rm -f $(addprefix $(exedir)/,bird birdc birdcl)
 	find $(objdir) -name "*.[od]" -exec rm -f '{}' '+'
 

doc/bird.sgml

@@ -1816,6 +1816,14 @@ Common route attributes are:
 	creation/removal. Zero is returned for routes with undefined outgoing
 	interfaces. Read-only.
 
+	<tag><label id="rta-onlink"><m/bool/ onlink</tag>
+	Onlink flag means that the specified nexthop is accessible on the
+	interface regardless of IP prefixes configured on the interface.
+	The attribute can be used to configure such next hops by first setting
+	<cf/onlink = true/ and <cf/ifname/, and then setting <cf/gw/. Possible
+	use case for setting this flag is to automatically build overlay IP-IP
+	networks on linux.
+
 	<tag><label id="rta-weight"><m/int/ weight</tag>
 	Multipath weight of route next hops. Valid values are 1-256. Reading
 	returns the weight of the first next hop, setting it sets weights of all

filter/config.Y

@@ -305,7 +305,7 @@ CF_KEYWORDS(FUNCTION, PRINT, PRINTN, UNSET, RETURN,
 	IF, THEN, ELSE, CASE,
 	FOR, IN, DO,
 	TRUE, FALSE, RT, RO, UNKNOWN, GENERIC,
-	FROM, GW, NET, MASK, PROTO, SOURCE, SCOPE, DEST, IFNAME, IFINDEX, WEIGHT, GW_MPLS,
+	FROM, GW, NET, MASK, PROTO, SOURCE, SCOPE, DEST, IFNAME, IFINDEX, WEIGHT, GW_MPLS, ONLINK,
 	PREFERENCE,
 	ROA_CHECK, ASN, SRC, DST,
 	IS_V4, IS_V6,
@@ -795,6 +795,7 @@ static_attr:
  | WEIGHT  { $$ = f_new_static_attr(T_INT,        SA_WEIGHT,	0); }
  | PREFERENCE { $$ = f_new_static_attr(T_INT,	  SA_PREF,	0); }
  | GW_MPLS { $$ = f_new_static_attr(T_INT,        SA_GW_MPLS,	0); }
+ | ONLINK  { $$ = f_new_static_attr(T_BOOL,       SA_ONLINK,	0); }
  ;
 
 term:

filter/data.h

@@ -102,6 +102,7 @@ enum f_sa_code {
   SA_WEIGHT,
   SA_PREF,
   SA_GW_MPLS,
+  SA_ONLINK,
 } PACKED;
 
 /* Static attribute definition (members of struct rta) */

filter/f-inst.c

@@ -694,6 +694,7 @@
       case SA_WEIGHT:	RESULT(sa.f_type, i, rta->nh.weight + 1); break;
       case SA_PREF:	RESULT(sa.f_type, i, rta->pref); break;
       case SA_GW_MPLS:	RESULT(sa.f_type, i, rta->nh.labels ? rta->nh.label[0] : MPLS_NULL); break;
+      case SA_ONLINK:	RESULT(sa.f_type, i, rta->nh.flags & RNF_ONLINK ? 1 : 0); break;
 
       default:
 	bug("Invalid static attribute access (%u/%u)", sa.f_type, sa.sa_code);
@@ -720,8 +721,8 @@
       case SA_GW:
 	{
 	  ip_addr ip = v1.val.ip;
-	  struct iface *ifa = ipa_is_link_local(ip) ? rta->nh.iface : NULL;
-	  neighbor *n = neigh_find((*fs->rte)->src->proto, ip, ifa, 0);
+	  struct iface *ifa = ipa_is_link_local(ip) || (rta->nh.flags & RNF_ONLINK) ? rta->nh.iface : NULL;
+	  neighbor *n = neigh_find((*fs->rte)->src->proto, ip, ifa, (rta->nh.flags & RNF_ONLINK) ? NEF_ONLINK : 0);
 	  if (!n || (n->scope == SCOPE_HOST))
 	    runtime( "Invalid gw address" );
 
@@ -801,6 +802,15 @@
 	rta->pref = v1.val.i;
 	break;
 
+      case SA_ONLINK:
+	{
+	  if (v1.val.i)
+	    rta->nh.flags |= RNF_ONLINK;
+	  else
+	    rta->nh.flags &= ~RNF_ONLINK;
+	}
+	break;
+
       default:
 	bug("Invalid static attribute access (%u/%u)", sa.f_type, sa.sa_code);
       }

proto/mrt/mrt.c

@@ -113,13 +113,13 @@ mrt_buffer_flush(buffer *b)
 }
 
 #define MRT_DEFINE_TYPE(S, T)						\
-  static inline void mrt_put_##S##_(buffer *b, T x)			\
+  static inline void UNUSED mrt_put_##S##_(buffer *b, T x)		\
   {									\
     put_##S(b->pos, x);							\
     b->pos += sizeof(T);						\
   }									\
 									\
-  static inline void mrt_put_##S(buffer *b, T x)			\
+  static inline void UNUSED mrt_put_##S(buffer *b, T x)			\
   {									\
     mrt_buffer_need(b, sizeof(T));					\
     put_##S(b->pos, x);							\

sysdep/bsd/krt-sock.Y

@@ -10,7 +10,7 @@ CF_HDR
 
 CF_DECLS
 
-CF_KEYWORDS(KERNEL, TABLE)
+CF_KEYWORDS(KERNEL, TABLE, METRIC)
 
 CF_GRAMMAR
 
@@ -25,6 +25,14 @@ kern_sys_item:
 
 	THIS_KRT->sys.table_id = $3;
    }
+ | METRIC expr {
+     if ($2 && !krt_max_metric)
+       cf_error("Kernel route metric not supported");
+     if ($2 > krt_max_metric)
+       cf_error("Kernel table id must be in range 0-%u", krt_max_metric);
+
+     THIS_KRT->sys.metric = $2;
+   }
  ;
 
 CF_CODE

sysdep/bsd/krt-sock.c

@@ -47,6 +47,11 @@ const int rt_default_ecmp = 0;
  * table_id is specified explicitly as sysctl scan argument, while in FreeBSD it
  * is handled implicitly by changing default table using setfib() syscall.
  *
+ * OpenBSD allows to use route metric. The behavior is controlled by these macro
+ * KRT_USE_METRIC, which enables use of rtm_priority in route send/recevive.
+ * There is also KRT_DEFAULT_METRIC and KRT_MAX_METRIC for default and maximum
+ * metric values.
+ *
  * KRT_SHARED_SOCKET	- use shared kernel socked instead of one for each krt_proto
  * KRT_USE_SETFIB_SCAN	- use setfib() for sysctl() route scan
  * KRT_USE_SETFIB_SOCK	- use SO_SETFIB socket option for kernel sockets
@@ -63,6 +68,9 @@ const int rt_default_ecmp = 0;
 
 #ifdef __OpenBSD__
 #define KRT_MAX_TABLES (RT_TABLEID_MAX+1)
+#define KRT_USE_METRIC
+#define KRT_MAX_METRIC 255
+#define KRT_DEFAULT_METRIC 56
 #define KRT_SHARED_SOCKET
 #define KRT_USE_SYSCTL_7
 #endif
@@ -71,6 +79,14 @@ const int rt_default_ecmp = 0;
 #define KRT_MAX_TABLES 1
 #endif
 
+#ifndef KRT_MAX_METRIC
+#define KRT_MAX_METRIC 0
+#endif
+
+#ifndef KRT_DEFAULT_METRIC
+#define KRT_DEFAULT_METRIC 0
+#endif
+
 
 /* Dynamic max number of tables */
 
@@ -143,6 +159,10 @@ static struct krt_proto *krt_table_map[KRT_MAX_TABLES][2];
 #endif
 
 
+/* Make it available to parser code */
+const uint krt_max_metric = KRT_MAX_METRIC;
+
+
 /* Route socket message processing */
 
 int
@@ -186,7 +206,7 @@ struct ks_msg
     memcpy(p, body, (l > sizeof(*p) ? sizeof(*p) : l));\
     body += l;}
 
-static inline void
+static inline void UNUSED
 sockaddr_fill_dl(struct sockaddr_dl *sa, struct iface *ifa)
 {
   uint len = OFFSETOF(struct sockaddr_dl, sdl_data);
@@ -231,6 +251,10 @@ krt_send_route(struct krt_proto *p, int cmd, rte *e)
   msg.rtm.rtm_tableid = KRT_CF->sys.table_id;
 #endif
 
+#ifdef KRT_USE_METRIC
+  msg.rtm.rtm_priority = KRT_CF->sys.metric;
+#endif
+
 #ifdef RTF_REJECT
   if(a->dest == RTD_UNREACHABLE)
     msg.rtm.rtm_flags |= RTF_REJECT;
@@ -586,7 +610,7 @@ krt_read_route(struct ks_msg *msg, struct krt_proto *p, int scan)
   e = rte_get_temp(&a, p->p.main_source);
   e->net = net;
 
-  ea_list *ea = alloca(sizeof(ea_list) + 1 * sizeof(eattr));
+  ea_list *ea = alloca(sizeof(ea_list) + 2 * sizeof(eattr));
   *ea = (ea_list) { .count = 1, .next = e->attrs->eattrs };
   e->attrs->eattrs = ea;
 
@@ -596,6 +620,15 @@ krt_read_route(struct ks_msg *msg, struct krt_proto *p, int scan)
     .u.data = src2,
   };
 
+#ifdef KRT_USE_METRIC
+  ea->count++;
+  ea->attrs[1] = (eattr) {
+    .id = EA_KRT_METRIC,
+    .type = EAF_TYPE_INT,
+    .u.data = msg->rtm.rtm_priority,
+  };
+#endif
+
   if (scan)
     krt_got_route(p, e, src);
   else
@@ -1155,7 +1188,7 @@ krt_sys_shutdown(struct krt_proto *p)
 int
 krt_sys_reconfigure(struct krt_proto *p UNUSED, struct krt_config *n, struct krt_config *o)
 {
-  return n->sys.table_id == o->sys.table_id;
+  return (n->sys.table_id == o->sys.table_id) && (n->sys.metric == o->sys.metric);
 }
 
 void
@@ -1168,11 +1201,13 @@ krt_sys_preconfig(struct config *c UNUSED)
 void krt_sys_init_config(struct krt_config *c)
 {
   c->sys.table_id = 0; /* Default table */
+  c->sys.metric = KRT_DEFAULT_METRIC;
 }
 
 void krt_sys_copy_config(struct krt_config *d, struct krt_config *s)
 {
   d->sys.table_id = s->sys.table_id;
+  d->sys.metric = s->sys.metric;
 }
 
 

sysdep/bsd/krt-sys.h

@@ -32,9 +32,11 @@ static inline void kif_sys_copy_config(struct kif_config *d UNUSED, struct kif_c
 /* Kernel routes */
 
 extern uint krt_max_tables;
+extern const uint krt_max_metric;
 
 struct krt_params {
   int table_id;				/* Kernel table ID we sync with */
+  u32 metric;				/* Kernel metric used for all routes */
 };
 
 struct krt_state {

sysdep/cf/README

@@ -14,3 +14,6 @@ CONFIG_DONTROUTE_UNICAST Use MSG_DONTROUTE flag for unicast packets (def for Fre
 CONFIG_USE_HDRINCL	Use IP_HDRINCL instead of control messages for source address on raw IP sockets.
 
 CONFIG_RESTRICTED_PRIVILEGES	Implements restricted privileges using drop_uid()
+
+CONFIG_MADV_DONTNEED_TO_FREE	To free pages, use MADV_DONTNEED instead of MADV_FREE (linux)
+CONFIG_DISABLE_THP	Disable transparent huge pages (linux)

sysdep/cf/linux.h

@@ -24,6 +24,7 @@
 #define CONFIG_INCLUDE_SYSPRIV_H "sysdep/linux/syspriv.h"
 
 #define CONFIG_MADV_DONTNEED_TO_FREE
+#define CONFIG_DISABLE_THP
 
 #ifndef AF_MPLS
 #define AF_MPLS 28

sysdep/unix/alloc.c

@@ -19,6 +19,10 @@
 #include <sys/mman.h>
 #endif
 
+#ifdef CONFIG_DISABLE_THP
+#include <sys/prctl.h>
+#endif
+
 long page_size = 0;
 
 #ifdef HAVE_MMAP
@@ -222,6 +226,12 @@ global_free_pages_cleanup_event(void *data UNUSED)
 void
 resource_sys_init(void)
 {
+#ifdef CONFIG_DISABLE_THP
+  /* Disable transparent huge pages, they do not work properly with madvice(MADV_DONTNEED) */
+  if (prctl(PR_SET_THP_DISABLE,  (unsigned long) 1,  (unsigned long) 0,  (unsigned long) 0,  (unsigned long) 0) < 0)
+    die("prctl(PR_SET_THP_DISABLE) failed: %m");
+#endif
+
 #ifdef HAVE_MMAP
   ASSERT_DIE(global_free_pages.cnt == 0);