zebra: fix for unexpected fdb entry showing up during ifdown/ifup events

Ticket: 2674793
Testing Done:  precommit, evpn-min and evpn-smoke

The problem in this case is whenever we are triggering ifdown
followed by ifup of bridge, we see that remote mac entries
are programmed with vlan-1 in the fdb from zebra and never cleaned up.
bridge has vlan_default_pvid 1 which means any port that gets added
will initially have vlan 1 which then gets deleted by ifupdown2 and
the proper vlan gets added.

The problem lies in zebra where we are not cleaning up the remote
macs during vlan change.

Fix is to uninstall the remote macs and then install them
during vlan change.

Signed-off-by: Stephen Worley <sworley@nvidia.com>

1 files changed, 41 insertions, 30 deletions

diff --git a/zebra/zebra_vxlan_if.c b/zebra/zebra_vxlan_if.c
index 57fdc1953..31cef767d 100644
--- a/zebra/zebra_vxlan_if.c
+++ b/zebra/zebra_vxlan_if.c
@@ -175,9 +175,11 @@ static int zebra_vxlan_if_del_vni(struct interface *ifp,
 
 static int zebra_vxlan_if_update_vni(struct interface *ifp,
 				     struct zebra_vxlan_vni *vnip,
-				     uint16_t chgflags)
+				     struct zebra_vxlan_if_update_ctx *ctx)
 {
 	vni_t vni;
+	uint16_t chgflags;
+	vlanid_t access_vlan;
 	struct zebra_if *zif;
 	struct zebra_l2info_vxlan *vxl;
 	struct zebra_evpn *zevpn;
@@ -193,6 +195,7 @@ static int zebra_vxlan_if_update_vni(struct interface *ifp,
 	assert(zif);
 	vxl = &zif->l2info.vxl;
 	vni = vnip->vni;
+	chgflags = ctx->chgflags;
 
 	zl3vni = zl3vni_lookup(vni);
 	if (zl3vni) {
@@ -299,8 +302,12 @@ static int zebra_vxlan_if_update_vni(struct interface *ifp,
 			/* Remove all existing local neigh and MACs for this VNI
 			 * (including from BGP)
 			 */
+			access_vlan = vnip->access_vlan;
+			vnip->access_vlan = ctx->old_vni.access_vlan;
 			zebra_evpn_neigh_del_all(zevpn, 0, 1, DEL_LOCAL_MAC);
 			zebra_evpn_mac_del_all(zevpn, 0, 1, DEL_LOCAL_MAC);
+			zebra_evpn_rem_mac_uninstall_all(zevpn);
+			vnip->access_vlan = access_vlan;
 		}
 
 		if (zevpn->local_vtep_ip.s_addr != vxl->vtep_ip.s_addr ||
@@ -347,15 +354,11 @@ static int zebra_vxlan_if_update_vni(struct interface *ifp,
 		if (chgflags & ZEBRA_VXLIF_MASTER_CHANGE)
 			zebra_evpn_read_mac_neigh(zevpn, ifp);
 		else if (chgflags & ZEBRA_VXLIF_VLAN_CHANGE) {
-			struct mac_walk_ctx m_wctx;
 			struct neigh_walk_ctx n_wctx;
 
 			zebra_evpn_read_mac_neigh(zevpn, ifp);
 
-			memset(&m_wctx, 0, sizeof(m_wctx));
-			m_wctx.zevpn = zevpn;
-			hash_iterate(zevpn->mac_table,
-				     zebra_evpn_install_mac_hash, &m_wctx);
+			zebra_evpn_rem_mac_install_all(zevpn);
 
 			memset(&n_wctx, 0, sizeof(n_wctx));
 			n_wctx.zevpn = zevpn;
@@ -493,15 +496,15 @@ static int zebra_vxlan_if_add_update_vni(struct zebra_if *zif,
 					 struct zebra_vxlan_vni *vni,
 					 void *ctxt)
 {
-	struct hash *old_vni_table;
 	struct zebra_vxlan_vni vni_tmp;
+	struct zebra_vxlan_if_update_ctx *ctx;
 	struct zebra_vxlan_vni *old_vni = NULL;
 
-	old_vni_table = (struct hash *)ctxt;
+	ctx = (struct zebra_vxlan_if_update_ctx *)ctxt;
 	memcpy(&vni_tmp, vni, sizeof(*vni));
 
-	old_vni = hash_release(old_vni_table, &vni_tmp);
-	if (!old_vni) {
+	if ((hashcount(ctx->old_vni_table) == 0) ||
+	    !(old_vni = hash_release(ctx->old_vni_table, &vni_tmp))) {
 		if (IS_ZEBRA_DEBUG_VXLAN)
 			zlog_debug("vxlan %s adding vni(%d, %d)",
 				   zif->ifp->name, vni->vni, vni->access_vlan);
@@ -510,6 +513,9 @@ static int zebra_vxlan_if_add_update_vni(struct zebra_if *zif,
 		return 0;
 	}
 
+	ctx->old_vni = *old_vni;
+	ctx->chgflags = ZEBRA_VXLIF_VLAN_CHANGE;
+
 	/* copy mcast group from old_vni as thats not being changed here */
 	vni->mcast_grp = old_vni->mcast_grp;
 
@@ -524,8 +530,7 @@ static int zebra_vxlan_if_add_update_vni(struct zebra_if *zif,
 		zebra_evpn_vl_vxl_deref(old_vni->access_vlan, old_vni->vni,
 					zif);
 		zebra_evpn_vl_vxl_ref(vni->access_vlan, vni->vni, zif);
-		zebra_vxlan_if_update_vni(zif->ifp, vni,
-					  ZEBRA_VXLIF_VLAN_CHANGE);
+		zebra_vxlan_if_update_vni(zif->ifp, vni, ctx);
 		zebra_vxlan_vni_free(old_vni);
 	}
 
@@ -536,10 +541,10 @@ static int zebra_vxlan_if_vni_entry_update_callback(struct zebra_if *zif,
 						    struct zebra_vxlan_vni *vni,
 						    void *ctxt)
 {
-	uint16_t *chgflags;
+	struct zebra_vxlan_if_update_ctx *ctx;
 
-	chgflags = (uint16_t *)ctxt;
-	return zebra_vxlan_if_update_vni(zif->ifp, vni, *chgflags);
+	ctx = (struct zebra_vxlan_if_update_ctx *)ctxt;
+	return zebra_vxlan_if_update_vni(zif->ifp, vni, ctx);
 }
 
 static int zebra_vxlan_if_vni_entry_del_callback(struct zebra_if *zif,
@@ -770,31 +775,32 @@ int zebra_vxlan_if_vni_table_add_update(struct interface *ifp,
 					struct hash *vni_table)
 {
 	struct zebra_if *zif;
-	struct hash *old_vni_table = NULL;
 	struct zebra_vxlan_vni_info *vni_info;
+	struct zebra_vxlan_if_update_ctx ctx;
 
 	zif = (struct zebra_if *)ifp->info;
 
 	vni_info = VNI_INFO_FROM_ZEBRA_IF(zif);
 
-	old_vni_table = vni_info->vni_table;
+	memset(&ctx, 0, sizeof(ctx));
+	ctx.old_vni_table = vni_info->vni_table;
 	vni_info->vni_table = vni_table;
 
-	zebra_vxlan_if_vni_iterate(zif, zebra_vxlan_if_add_update_vni,
-				   old_vni_table);
+	zebra_vxlan_if_vni_iterate(zif, zebra_vxlan_if_add_update_vni, &ctx);
 
 	/* release kernel deleted vnis */
-	if (old_vni_table) {
-		if (hashcount(old_vni_table)) {
+	if (ctx.old_vni_table) {
+		if (hashcount(ctx.old_vni_table)) {
 			/* UGLY HACK: Put back the old table so that delete of
 			 * MACs goes through and then flip back.
 			 */
-			vni_info->vni_table = old_vni_table;
-			hash_iterate(old_vni_table, zebra_vxlan_if_vni_clean,
-				     zif);
+			vni_info->vni_table = ctx.old_vni_table;
+			hash_iterate(ctx.old_vni_table,
+				     zebra_vxlan_if_vni_clean, zif);
 			vni_info->vni_table = vni_table;
 		}
-		zebra_vxlan_vni_table_destroy(old_vni_table);
+		zebra_vxlan_vni_table_destroy(ctx.old_vni_table);
+		ctx.old_vni_table = NULL;
 	}
 
 	return 0;
@@ -805,6 +811,7 @@ int zebra_vxlan_if_vni_mcast_group_update(struct interface *ifp, vni_t vni_id,
 {
 	struct zebra_if *zif;
 	struct zebra_vxlan_vni *vni;
+	struct zebra_vxlan_if_update_ctx ctx;
 
 	zif = (struct zebra_if *)ifp->info;
 
@@ -815,13 +822,16 @@ int zebra_vxlan_if_vni_mcast_group_update(struct interface *ifp, vni_t vni_id,
 	if (!vni)
 		return 0;
 
+	memset(&ctx, 0, sizeof(ctx));
+	ctx.old_vni.mcast_grp = vni->mcast_grp;
+	ctx.chgflags = ZEBRA_VXLIF_MCAST_GRP_CHANGE;
+
 	if (mcast_group)
 		vni->mcast_grp = *mcast_group;
 	else
 		memset(&vni->mcast_grp, 0, sizeof(vni->mcast_grp));
 
-	return zebra_vxlan_if_update_vni(ifp, vni,
-					 ZEBRA_VXLIF_MCAST_GRP_CHANGE);
+	return zebra_vxlan_if_update_vni(ifp, vni, &ctx);
 }
 
 int zebra_vxlan_if_vni_down(struct interface *ifp, struct zebra_vxlan_vni *vnip)
@@ -1061,7 +1071,8 @@ int zebra_vxlan_if_del(struct interface *ifp)
 /*
  * Handle VxLAN interface update - change to tunnel IP, master or VLAN.
  */
-int zebra_vxlan_if_update(struct interface *ifp, uint16_t chgflags)
+int zebra_vxlan_if_update(struct interface *ifp,
+			  struct zebra_vxlan_if_update_ctx *ctx)
 {
 	struct zebra_if *zif;
 	struct zebra_vxlan_vni_info *vni_info;
@@ -1071,11 +1082,11 @@ int zebra_vxlan_if_update(struct interface *ifp, uint16_t chgflags)
 
 	if (IS_ZEBRA_VXLAN_IF_VNI(zif)) {
 		vni_info = VNI_INFO_FROM_ZEBRA_IF(zif);
-		return zebra_vxlan_if_update_vni(ifp, &vni_info->vni, chgflags);
+		return zebra_vxlan_if_update_vni(ifp, &vni_info->vni, ctx);
 	}
 
 	zebra_vxlan_if_vni_iterate(
-		zif, zebra_vxlan_if_vni_entry_update_callback, &chgflags);
+		zif, zebra_vxlan_if_vni_entry_update_callback, ctx);
 
 	return 0;
 }