zebra: re-install nhg on interface up

Intermittently zebra and kernel are out of sync
when interface flaps and the add's/dels are in
same processing queue and zebra assumes no change in nexthop.
Hence we need to bring in a reinstall to kernel
of the nexthops and routes to sync their states.

Upon interface flap kernel would have deleted NHGs
associated to a interface (the one flapped),
zebra retains NHGs for 3 mins even though upper
layer protocol removes the nexthops (associated NHG).
As part of interface address add ,
re-add singleton NHGs associated to interface.

Ticket: #3173663
Issue: 3173663

Signed-off-by: Ashwini Reddy <ashred@nvidia.com>
Signed-off-by: Chirag Shah <chirag@nvidia.com>

3 files changed, 51 insertions, 0 deletions

diff --git a/zebra/redistribute.c b/zebra/redistribute.c
index d2fa85eb6..6767000f3 100644
--- a/zebra/redistribute.c
+++ b/zebra/redistribute.c
@@ -548,6 +548,10 @@ void zebra_interface_address_add_update(struct interface *ifp,
 						client, ifp, ifc);
 		}
 	}
+	/* interface associated NHGs may have been deleted,
+	 * re-sync zebra -> dplane NHGs
+	 */
+	zebra_interface_nhg_reinstall(ifp);
 }
 
 /* Interface address deletion. */
diff --git a/zebra/zebra_nhg.c b/zebra/zebra_nhg.c
index 8616b1405..2888a9ac6 100644
--- a/zebra/zebra_nhg.c
+++ b/zebra/zebra_nhg.c
@@ -3080,6 +3080,12 @@ void zebra_nhg_install_kernel(struct nhg_hash_entry *nhe)
 	/* Resolve it first */
 	nhe = zebra_nhg_resolve(nhe);
 
+	if (zebra_nhg_set_valid_if_active(nhe)) {
+		if (IS_ZEBRA_DEBUG_NHG_DETAIL)
+			zlog_debug("%s: valid flag set for nh %pNG", __func__,
+				   nhe);
+	}
+
 	/* Make sure all depends are installed/queued */
 	frr_each(nhg_connected_tree, &nhe->nhg_depends, rb_node_dep) {
 		zebra_nhg_install_kernel(rb_node_dep->nhe);
@@ -3651,3 +3657,43 @@ static ssize_t printfrr_nhghe(struct fbuf *buf, struct printfrr_eargs *ea,
 	ret += bputs(buf, "]");
 	return ret;
 }
+
+/*
+ * On interface add the nexthop that resolves to this intf needs
+ * a re-install. There are following scenarios when the nexthop group update
+ * gets skipped:
+ * 1. When upper level protocol sends removal of NHG, there is
+ * timer running to keep NHG for 180 seconds, during this interval, same route
+ * with same set of nexthops installation is given , the same NHG is used
+ * but since NHG is not reinstalled on interface address add, it is not aware
+ * in Dplan/Kernel.
+ * 2. Due to a quick port flap due to interface add and delete
+ * to be processed in same queue one after another. Zebra believes that
+ * there is no change in nhg in this case. Hence this re-install will
+ * make sure the nexthop group gets updated to Dplan/Kernel.
+ */
+void zebra_interface_nhg_reinstall(struct interface *ifp)
+{
+	struct nhg_connected *rb_node_dep = NULL;
+	struct zebra_if *zif = ifp->info;
+	struct nexthop *nh;
+
+	if (IS_ZEBRA_DEBUG_NHG_DETAIL)
+		zlog_debug(
+			"%s: Installing interface %s associated NHGs into kernel",
+			__func__, ifp->name);
+
+	frr_each (nhg_connected_tree, &zif->nhg_dependents, rb_node_dep) {
+		nh = rb_node_dep->nhe->nhg.nexthop;
+		if (zebra_nhg_set_valid_if_active(rb_node_dep->nhe)) {
+			if (IS_ZEBRA_DEBUG_NHG_DETAIL)
+				zlog_debug(
+					"%s: Setting the valid flag for nhe %pNG, interface: %s",
+					__func__, rb_node_dep->nhe, ifp->name);
+		}
+		/* Check for singleton NHG associated to interface */
+		if (nexthop_is_ifindex_type(nh) &&
+		    zebra_nhg_depends_is_empty(rb_node_dep->nhe))
+			zebra_nhg_install_kernel(rb_node_dep->nhe);
+	}
+}
diff --git a/zebra/zebra_nhg.h b/zebra/zebra_nhg.h
index b178007b4..6179be344 100644
--- a/zebra/zebra_nhg.h
+++ b/zebra/zebra_nhg.h
@@ -358,6 +358,7 @@ extern uint8_t zebra_nhg_nhe2grp(struct nh_grp *grp, struct nhg_hash_entry *nhe,
 /* Dataplane install/uninstall */
 extern void zebra_nhg_install_kernel(struct nhg_hash_entry *nhe);
 extern void zebra_nhg_uninstall_kernel(struct nhg_hash_entry *nhe);
+extern void zebra_interface_nhg_reinstall(struct interface *ifp);
 
 /* Forward ref of dplane update context type */
 struct zebra_dplane_ctx;