From: Ondrej Zajicek (work) <santiago@crfreenet.org>
Date: Wed, 4 Jul 2018 14:42:42 +0000 (+0200)
Subject: BGP: Long-lived graceful restart enhancements
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=e31e36814790f4a0047fc8eb1727be3ad307eaf2;p=thirdparty%2Fbird.git

BGP: Long-lived graceful restart enhancements

When use of LLGR is negotiated, handle hold timeout by LLGR instead of by
hard restart. Allow to configure whether BFD session down event should be
handled by GR/LLGR or by hard restart.
---

diff --git a/doc/bird.sgml b/doc/bird.sgml
index c66e8c725..7cbb4de69 100644
--- a/doc/bird.sgml
+++ b/doc/bird.sgml
@@ -1977,13 +1977,16 @@ using the following configuration parameters:
 	immediately shut down. Note that this option cannot be used with
 	multihop BGP. Default: disabled.
 
-	<tag><label id="bgp-bfd">bfd <M>switch</M></tag>
+	<tag><label id="bgp-bfd">bfd <M>switch</M>|graceful</tag>
 	BGP could use BFD protocol as an advisory mechanism for neighbor
 	liveness and failure detection. If enabled, BIRD setups a BFD session
 	for the BGP neighbor and tracks its liveness by it. This has an
 	advantage of an order of magnitude lower detection times in case of
-	failure. Note that BFD protocol also has to be configured, see
-	<ref id="bfd" name="BFD"> section for details. Default: disabled.
+	failure. When a neighbor failure is detected, the BGP session is
+	restarted. Optionally, it can be configured (by <cf/graceful/ argument)
+	to trigger graceful restart instead of regular restart.  Note that BFD
+	protocol also has to be configured, see <ref id="bfd" name="BFD">
+	section for details. Default: disabled.
 
 	<tag><label id="bgp-ttl-security">ttl security <m/switch/</tag>
 	Use GTSM (<rfc id="5082"> - the generalized TTL security mechanism). GTSM
diff --git a/proto/bgp/bgp.c b/proto/bgp/bgp.c
index 0932051ab..369d78f57 100644
--- a/proto/bgp/bgp.c
+++ b/proto/bgp/bgp.c
@@ -704,6 +704,12 @@ bgp_hold_timeout(timer *t)
 
   if (sk_rx_ready(conn->sk) > 0)
     bgp_start_timer(conn->hold_timer, 10);
+  else if ((conn->state == BS_ESTABLISHED) && p->gr_ready && conn->peer_llgr_able)
+  {
+    BGP_TRACE(D_EVENTS, "Hold timer expired");
+    bgp_handle_graceful_restart(p);
+    bgp_conn_enter_idle_state(conn);
+  }
   else
     bgp_error(conn, 4, 0, NULL, 0);
 }
@@ -1046,13 +1052,29 @@ bgp_bfd_notify(struct bfd_request *req)
   int ps = p->p.proto_state;
 
   if (req->down && ((ps == PS_START) || (ps == PS_UP)))
+  {
+    BGP_TRACE(D_EVENTS, "BFD session down");
+
+    /* Ignore if already in GR */
+    if (p->gr_active && (p->cf->bfd == BGP_BFD_GRACEFUL))
+      return;
+
+    if (p->conn && (p->conn->state == BS_ESTABLISHED) &&
+	p->gr_ready && (p->cf->bfd == BGP_BFD_GRACEFUL))
+    {
+      /* Trigger graceful restart */
+      bgp_handle_graceful_restart(p);
+      bgp_conn_enter_idle_state(p->conn);
+    }
+    else
     {
-      BGP_TRACE(D_EVENTS, "BFD session down");
+      /* Trigger session down */
       bgp_store_error(p, NULL, BE_MISC, BEM_BFD_DOWN);
       if (ps == PS_UP)
 	bgp_update_startup_delay(p);
       bgp_stop(p, 0, NULL, 0);
     }
+  }
 }
 
 static void
diff --git a/proto/bgp/bgp.h b/proto/bgp/bgp.h
index 53194fb62..7703fb6e9 100644
--- a/proto/bgp/bgp.h
+++ b/proto/bgp/bgp.h
@@ -102,6 +102,8 @@ struct bgp_config {
 #define BGP_GRS_LLGR_1		2	/* Long-lived GR phase 1 (restart time) */
 #define BGP_GRS_LLGR_2		3	/* Long-lived GR phase 2 (stale time) */
 
+#define BGP_BFD_GRACEFUL	2	/* BFD down triggers graceful restart */
+
 
 struct bgp_conn {
   struct bgp_proto *bgp;
diff --git a/proto/bgp/config.Y b/proto/bgp/config.Y
index 2ed1944df..95ceccb5f 100644
--- a/proto/bgp/config.Y
+++ b/proto/bgp/config.Y
@@ -172,6 +172,7 @@ bgp_proto:
  | bgp_proto TTL SECURITY bool ';' { BGP_CFG->ttl_security = $4; }
  | bgp_proto CHECK LINK bool ';' { BGP_CFG->check_link = $4; }
  | bgp_proto BFD bool ';' { BGP_CFG->bfd = $3; cf_check_bfd($3); }
+ | bgp_proto BFD GRACEFUL ';' { BGP_CFG->bfd = BGP_BFD_GRACEFUL; cf_check_bfd(1); }
  ;
 
 CF_ADDTO(dynamic_attr, BGP_ORIGIN