examples/distributor: fix Rx thread logic for zero packet
authorReshma Pattan <reshma.pattan@intel.com>
Thu, 14 Jul 2016 15:48:45 +0000 (16:48 +0100)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Fri, 15 Jul 2016 21:42:48 +0000 (23:42 +0200)
Zero packets can be returned by rte_eth_rx_burst() and
rte_distributor_returned_pkts() inside lcore_rx(), so
for zero packet scenario instead of proceeding to
next operations we should continue to the next iteration of the
loop to avoid unnecessary processing overhead which is causing
rx packets to be dropped and hence distributor failing to forward the
packets.

Fixes: 07db4a97 ("examples/distributor: new sample app")

Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
examples/distributor/main.c

index 24857f2..537cee1 100644 (file)
@@ -221,14 +221,22 @@ lcore_rx(struct lcore_params *p)
                struct rte_mbuf *bufs[BURST_SIZE*2];
                const uint16_t nb_rx = rte_eth_rx_burst(port, 0, bufs,
                                BURST_SIZE);
+               if (unlikely(nb_rx == 0)) {
+                       if (++port == nb_ports)
+                               port = 0;
+                       continue;
+               }
                app_stats.rx.rx_pkts += nb_rx;
 
                rte_distributor_process(d, bufs, nb_rx);
                const uint16_t nb_ret = rte_distributor_returned_pkts(d,
                                bufs, BURST_SIZE*2);
                app_stats.rx.returned_pkts += nb_ret;
-               if (unlikely(nb_ret == 0))
+               if (unlikely(nb_ret == 0)) {
+                       if (++port == nb_ports)
+                               port = 0;
                        continue;
+               }
 
                uint16_t sent = rte_ring_enqueue_burst(r, (void *)bufs, nb_ret);
                app_stats.rx.enqueued_pkts += sent;