linux-cp: handle ipv4 routes when interface is disabled

Type: improvement

Currently, when an interface is brought down administratively, IPv4
routes that resolve through that interface remain in the FIB. However,
the kernel removes those routes but doesn't send any notifications about
that. Desynchronization between the kernel and VPP happens.

With this change, when a notification received from the kernel
indicating that an interface was brought down, in addition to bringing
the VPP interface down, walk the IPv4 FIB bound to that interface and
remove any entries that resolve through that interface and were added
with one of the linux-cp FIB sources.

Signed-off-by: Alexander Chernavin <achernavin@netgate.com>
Change-Id: I0cd14bb63c9e6616ae1c5739b17c3bf33b186bc2

1 files changed, 88 insertions, 2 deletions

diff --git a/src/plugins/linux-cp/lcp_router.c b/src/plugins/linux-cp/lcp_router.c
index 7542aa06614..34d0500b0c5 100644
--- a/src/plugins/linux-cp/lcp_router.c
+++ b/src/plugins/linux-cp/lcp_router.c
@@ -307,6 +307,9 @@ lcp_router_link_addr (struct rtnl_link *rl, lcp_itf_pair_t *lip)
 					  lip->lip_phy_adjs.adj_index[AF_IP6]);
 }
 
+static void lcp_router_table_flush (lcp_router_table_t *nlt, u32 sw_if_index,
+				    fib_source_t source);
+
 static void
 lcp_router_link_add (struct rtnl_link *rl, void *ctx)
 {
@@ -320,6 +323,8 @@ lcp_router_link_add (struct rtnl_link *rl, void *ctx)
   if (INDEX_INVALID != lipi)
     {
       lcp_itf_pair_t *lip;
+      u32 sw_if_flags;
+      u32 sw_if_up;
 
       lip = lcp_itf_pair_get (lipi);
       if (!vnet_get_sw_interface (vnm, lip->lip_phy_sw_if_index))
@@ -328,16 +333,48 @@ lcp_router_link_add (struct rtnl_link *rl, void *ctx)
       if (lcp_router_lip_ts_check ((nl_msg_info_t *) ctx, lip))
 	return;
 
-      if (up)
+      sw_if_flags =
+	vnet_sw_interface_get_flags (vnm, lip->lip_phy_sw_if_index);
+      sw_if_up = (sw_if_flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP);
+
+      if (!sw_if_up && up)
 	{
 	  vnet_sw_interface_admin_up (vnet_get_main (),
 				      lip->lip_phy_sw_if_index);
 	}
-      else
+      else if (sw_if_up && !up)
 	{
 	  vnet_sw_interface_admin_down (vnet_get_main (),
 					lip->lip_phy_sw_if_index);
+
+	  /* When an interface is brought down administratively, the kernel
+	   * removes routes which resolve through that interface. For IPv4
+	   * routes, the kernel will not send any explicit RTM_DELROUTE
+	   * messages about removing them. In order to synchronize with the
+	   * kernel, affected IPv4 routes need to be manually removed from the
+	   * FIB. The behavior is different for IPv6 routes. Explicit
+	   * RTM_DELROUTE messages are sent about IPv6 routes being removed.
+	   */
+	  u32 fib_index;
+	  lcp_router_table_t *nlt;
+
+	  fib_index = fib_table_get_index_for_sw_if_index (
+	    FIB_PROTOCOL_IP4, lip->lip_phy_sw_if_index);
+
+	  pool_foreach (nlt, lcp_router_table_pool)
+	    {
+	      if (fib_index == nlt->nlt_fib_index &&
+		  FIB_PROTOCOL_IP4 == nlt->nlt_proto)
+		{
+		  lcp_router_table_flush (nlt, lip->lip_phy_sw_if_index,
+					  lcp_rt_fib_src);
+		  lcp_router_table_flush (nlt, lip->lip_phy_sw_if_index,
+					  lcp_rt_fib_src_dynamic);
+		  break;
+		}
+	    }
 	}
+
       LCP_ROUTER_DBG ("link: %s (%d) -> %U/%U %s", rtnl_link_get_name (rl),
 		      rtnl_link_get_ifindex (rl), format_vnet_sw_if_index_name,
 		      vnm, lip->lip_phy_sw_if_index,
@@ -1108,6 +1145,55 @@ lcp_router_route_sync_end (void)
     }
 }
 
+typedef struct lcp_router_table_flush_ctx_t_
+{
+  fib_node_index_t *lrtf_entries;
+  u32 lrtf_sw_if_index;
+  fib_source_t lrtf_source;
+} lcp_router_table_flush_ctx_t;
+
+static fib_table_walk_rc_t
+lcp_router_table_flush_cb (fib_node_index_t fib_entry_index, void *arg)
+{
+  lcp_router_table_flush_ctx_t *ctx = arg;
+
+  if (fib_entry_get_resolving_interface_for_source (
+	fib_entry_index, ctx->lrtf_source) == ctx->lrtf_sw_if_index)
+    {
+      vec_add1 (ctx->lrtf_entries, fib_entry_index);
+    }
+  return (FIB_TABLE_WALK_CONTINUE);
+}
+
+static void
+lcp_router_table_flush (lcp_router_table_t *nlt, u32 sw_if_index,
+			fib_source_t source)
+{
+  fib_node_index_t *fib_entry_index;
+  lcp_router_table_flush_ctx_t ctx = {
+    .lrtf_entries = NULL,
+    .lrtf_sw_if_index = sw_if_index,
+    .lrtf_source = source,
+  };
+
+  fib_table_walk (nlt->nlt_fib_index, nlt->nlt_proto,
+		  lcp_router_table_flush_cb, &ctx);
+
+  LCP_ROUTER_DBG (
+    "Flush table: proto %U, fib-index %u, resolved via %U for source %U",
+    format_fib_protocol, nlt->nlt_proto, nlt->nlt_fib_index,
+    format_vnet_sw_if_index_name, vnet_get_main (), sw_if_index,
+    format_fib_source, source);
+
+  vec_foreach (fib_entry_index, ctx.lrtf_entries)
+    {
+      fib_table_entry_delete_index (*fib_entry_index, source);
+      lcp_router_table_unlock (nlt);
+    }
+
+  vec_free (ctx.lrtf_entries);
+}
+
 const nl_vft_t lcp_router_vft = {
   .nvl_rt_link_add = { .is_mp_safe = 0, .cb = lcp_router_link_add },
   .nvl_rt_link_del = { .is_mp_safe = 0, .cb = lcp_router_link_del },