Fix sending GARP/NA on Bonded Interface Active/Backup Link Up/Down

For bonded interface in Active/Backup mode (mode 1), we need to
send a GARP/NA packet, if IP address is present, on slave link
state change to up or down to help with route convergence. The
callback from DPDK happens in a separate thread so we need to make
sure RPC call is used to signal the send_garp_na process in the
main thread. Also need to fix DPDK polling so the slave links are
not polled.

Change-Id: If5fd8ea2d28c54dd28726ac403ad366386ce9651
Signed-off-by: John Lo <loj@cisco.com>

5 files changed, 101 insertions, 72 deletions

diff --git a/src/plugins/dpdk/device/common.c b/src/plugins/dpdk/device/common.c
index 2707b4d889c..aedc3f5227d 100644
--- a/src/plugins/dpdk/device/common.c
+++ b/src/plugins/dpdk/device/common.c
@@ -181,12 +181,69 @@ dpdk_device_stop (dpdk_device_t * xd)
     }
 }
 
+/* Even type for send_garp_na_process */
+enum
+{
+  SEND_GARP_NA = 1,
+} dpdk_send_garp_na_process_event_t;
+
+static vlib_node_registration_t send_garp_na_proc_node;
+
+static uword
+send_garp_na_process (vlib_main_t * vm,
+		      vlib_node_runtime_t * rt, vlib_frame_t * f)
+{
+  vnet_main_t *vnm = vnet_get_main ();
+  uword event_type, *event_data = 0;
+
+  while (1)
+    {
+      u32 i;
+      uword dpdk_port;
+      vlib_process_wait_for_event (vm);
+      event_type = vlib_process_get_events (vm, &event_data);
+      ASSERT (event_type == SEND_GARP_NA);
+      for (i = 0; i < vec_len (event_data); i++)
+	{
+	  dpdk_port = event_data[i];
+	  if (i < 5)		/* wait 0.2 sec for link to settle, max total 1 sec */
+	    vlib_process_suspend (vm, 0.2);
+	  dpdk_device_t *xd = &dpdk_main.devices[dpdk_port];
+	  u32 hw_if_index = xd->hw_if_index;
+	  vnet_hw_interface_t *hi = vnet_get_hw_interface (vnm, hw_if_index);
+	  dpdk_update_link_state (xd, vlib_time_now (vm));
+	  send_ip4_garp (vm, hi);
+	  send_ip6_na (vm, hi);
+	}
+      vec_reset_length (event_data);
+    }
+  return 0;
+}
+
+/* *INDENT-OFF* */
+VLIB_REGISTER_NODE (send_garp_na_proc_node, static) = {
+    .function = send_garp_na_process,
+    .type = VLIB_NODE_TYPE_PROCESS,
+    .name = "send-garp-na-process",
+};
+/* *INDENT-ON* */
+
+void vl_api_force_rpc_call_main_thread (void *fp, u8 * data, u32 data_length);
+
+static void
+garp_na_proc_callback (uword * dpdk_port)
+{
+  vlib_main_t *vm = vlib_get_main ();
+  ASSERT (vlib_get_thread_index () == 0);
+  vlib_process_signal_event
+    (vm, send_garp_na_proc_node.index, SEND_GARP_NA, *dpdk_port);
+}
+
 always_inline int
 dpdk_port_state_callback_inline (uint8_t port_id,
 				 enum rte_eth_event_type type, void *param)
 {
   struct rte_eth_link link;
-  vlib_main_t *vm = vlib_get_main ();
   dpdk_device_t *xd = &dpdk_main.devices[port_id];
 
   RTE_SET_USED (param);
@@ -201,32 +258,21 @@ dpdk_port_state_callback_inline (uint8_t port_id,
 
   if (xd->flags & DPDK_DEVICE_FLAG_BOND_SLAVE)
     {
-      u8 bd_port = xd->bond_port;
+      uword bd_port = xd->bond_port;
       int bd_mode = rte_eth_bond_mode_get (bd_port);
-
-      if ((link_up && !(xd->flags & DPDK_DEVICE_FLAG_BOND_SLAVE_UP)) ||
-	  (!link_up && (xd->flags & DPDK_DEVICE_FLAG_BOND_SLAVE_UP)))
+#if 0
+      clib_warning ("Port %d state to %s, "
+		    "slave of port %d BondEthernet%d in mode %d",
+		    port_id, (link_up) ? "UP" : "DOWN",
+		    bd_port, xd->port_id, bd_mode);
+#endif
+      if (bd_mode == BONDING_MODE_ACTIVE_BACKUP)
 	{
-	  clib_warning ("Port %d state to %s, "
-			"slave of port %d BondEthernet%d in mode %d",
-			port_id, (link_up) ? "UP" : "DOWN",
-			bd_port, xd->port_id, bd_mode);
-	  if (bd_mode == BONDING_MODE_ACTIVE_BACKUP)
-	    {
-	      rte_eth_link_get_nowait (bd_port, &link);
-	      if (link.link_status)	/* bonded interface up */
-		{
-		  u32 hw_if_index = dpdk_main.devices[bd_port].hw_if_index;
-		  vlib_process_signal_event
-		    (vm, send_garp_na_process_node_index, SEND_GARP_NA,
-		     hw_if_index);
-		}
-	    }
+	  vl_api_force_rpc_call_main_thread
+	    (garp_na_proc_callback, (u8 *) & bd_port, sizeof (uword));
 	}
-      if (link_up)		/* Update slave link status */
-	xd->flags |= DPDK_DEVICE_FLAG_BOND_SLAVE_UP;
-      else
-	xd->flags &= ~DPDK_DEVICE_FLAG_BOND_SLAVE_UP;
+      xd->flags |= link_up ?
+	DPDK_DEVICE_FLAG_BOND_SLAVE_UP : ~DPDK_DEVICE_FLAG_BOND_SLAVE_UP;
     }
   else				/* Should not happen as callback not setup for "normal" links */
     {
diff --git a/src/plugins/dpdk/device/node.c b/src/plugins/dpdk/device/node.c
index 74fb8da1f2e..cf8b9699206 100644
--- a/src/plugins/dpdk/device/node.c
+++ b/src/plugins/dpdk/device/node.c
@@ -661,6 +661,8 @@ dpdk_input (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * f)
   foreach_device_and_queue (dq, rt->devices_and_queues)
     {
       xd = vec_elt_at_index(dm->devices, dq->dev_instance);
+      if (PREDICT_FALSE (xd->flags & DPDK_DEVICE_FLAG_BOND_SLAVE))
+	continue; 	/* Do not poll slave to a bonded interface */
       if (xd->flags & DPDK_DEVICE_FLAG_MAYBE_MULTISEG)
         n_rx_packets += dpdk_device_input (dm, xd, node, thread_index, dq->queue_id, /* maybe_multiseg */ 1);
       else
diff --git a/src/vlibmemory/memory_vlib.c b/src/vlibmemory/memory_vlib.c
index b6b87529375..77959e6d366 100644
--- a/src/vlibmemory/memory_vlib.c
+++ b/src/vlibmemory/memory_vlib.c
@@ -1452,8 +1452,9 @@ vl_api_rpc_call_reply_t_handler (vl_api_rpc_call_reply_t * mp)
   clib_warning ("unimplemented");
 }
 
-void
-vl_api_rpc_call_main_thread (void *fp, u8 * data, u32 data_length)
+always_inline void
+vl_api_rpc_call_main_thread_inline (void *fp, u8 * data, u32 data_length,
+				    u8 force_rpc)
 {
   vl_api_rpc_call_t *mp;
   api_main_t *am = &api_main;
@@ -1461,7 +1462,7 @@ vl_api_rpc_call_main_thread (void *fp, u8 * data, u32 data_length)
   unix_shared_memory_queue_t *q;
 
   /* Main thread: call the function directly */
-  if (vlib_get_thread_index () == 0)
+  if ((force_rpc == 0) && (vlib_get_thread_index () == 0))
     {
       vlib_main_t *vm = vlib_get_main ();
       void (*call_fp) (void *);
@@ -1507,6 +1508,29 @@ vl_api_rpc_call_main_thread (void *fp, u8 * data, u32 data_length)
   pthread_mutex_unlock (&q->mutex);
 }
 
+/*
+ * Check if called from worker threads.
+ * If so, make rpc call of fp through shmem.
+ * Otherwise, call fp directly
+ */
+void
+vl_api_rpc_call_main_thread (void *fp, u8 * data, u32 data_length)
+{
+  vl_api_rpc_call_main_thread_inline (fp, data, data_length,	/*force_rpc */
+				      0);
+}
+
+/*
+ * Always make rpc call of fp through shmem, useful for calling from threads
+ * not setup as worker threads, such as DPDK callback thread
+ */
+void
+vl_api_force_rpc_call_main_thread (void *fp, u8 * data, u32 data_length)
+{
+  vl_api_rpc_call_main_thread_inline (fp, data, data_length,	/*force_rpc */
+				      1);
+}
+
 static void
 vl_api_trace_plugin_msg_ids_t_handler (vl_api_trace_plugin_msg_ids_t * mp)
 {
diff --git a/src/vnet/ethernet/arp.c b/src/vnet/ethernet/arp.c
index e974d2551e5..120a276cc0f 100644
--- a/src/vnet/ethernet/arp.c
+++ b/src/vnet/ethernet/arp.c
@@ -2482,7 +2482,7 @@ ethernet_arp_change_mac (u32 sw_if_index)
   /* *INDENT-ON* */
 }
 
-void static
+void
 send_ip4_garp (vlib_main_t * vm, vnet_hw_interface_t * hi)
 {
   ip4_main_t *i4m = &ip4_main;
@@ -2526,42 +2526,6 @@ send_ip4_garp (vlib_main_t * vm, vnet_hw_interface_t * hi)
     }
 }
 
-static vlib_node_registration_t send_garp_na_proc_node;
-
-static uword
-send_garp_na_process (vlib_main_t * vm,
-		      vlib_node_runtime_t * rt, vlib_frame_t * f)
-{
-  vnet_main_t *vnm = vnet_get_main ();
-  uword event_type, *event_data = 0;
-
-  send_garp_na_process_node_index = send_garp_na_proc_node.index;
-
-  while (1)
-    {
-      vlib_process_wait_for_event (vm);
-      event_type = vlib_process_get_events (vm, &event_data);
-      if ((event_type == SEND_GARP_NA) && (vec_len (event_data) >= 1))
-	{
-	  u32 hw_if_index = event_data[0];
-	  vnet_hw_interface_t *hi = vnet_get_hw_interface (vnm, hw_if_index);
-	  send_ip4_garp (vm, hi);
-	  send_ip6_na (vm, hi);
-	}
-      vec_reset_length (event_data);
-    }
-  return 0;
-}
-
-
-/* *INDENT-OFF* */
-VLIB_REGISTER_NODE (send_garp_na_proc_node, static) = {
-    .function = send_garp_na_process,
-    .type = VLIB_NODE_TYPE_PROCESS,
-    .name = "send-garp-na-process",
-};
-/* *INDENT-ON* */
-
 /*
  * fd.io coding-style-patch-verification: ON
  *
diff --git a/src/vnet/ethernet/arp_packet.h b/src/vnet/ethernet/arp_packet.h
index d740b844e12..661f33f93af 100644
--- a/src/vnet/ethernet/arp_packet.h
+++ b/src/vnet/ethernet/arp_packet.h
@@ -167,14 +167,7 @@ typedef struct
 ethernet_arp_ip4_entry_t *ip4_neighbor_entries (u32 sw_if_index);
 u8 *format_ethernet_arp_ip4_entry (u8 * s, va_list * va);
 
-/* Node index for send_garp_na_process */
-extern u32 send_garp_na_process_node_index;
-
-/* Even type for send_garp_na_process */
-enum
-{
-  SEND_GARP_NA = 1,
-} dpdk_send_garp_na_process_event_t;
+void send_ip4_garp (vlib_main_t * vm, vnet_hw_interface_t * hi);
 
 #endif /* included_ethernet_arp_packet_h */