net/enic: add VLAN and csum offloads to simple Tx handler

Currently the simple Tx handler supports no offloads, which makes it
usable only for a small number of benchmarks. Add vlan and checksum
offloads to the handler, as cycles/packet increases only by about 3
cycles, and applications commonly use those offloads.

Signed-off-by: Hyong Youb Kim <hyonkim@cisco.com>
Reviewed-by: John Daley <johndale@cisco.com>

diff --git a/drivers/net/enic/enic_main.c b/drivers/net/enic/enic_main.c
index fd940c5..03c5ef7 100644 (file)
--- a/drivers/net/enic/enic_main.c
+++ b/drivers/net/enic/enic_main.c
@@ -534,6 +534,7 @@ int enic_enable(struct enic *enic)
        unsigned int index;
        int err;
        struct rte_eth_dev *eth_dev = enic->rte_dev;
+       uint64_t simple_tx_offloads;
 
        eth_dev->data->dev_link.link_speed = vnic_dev_port_speed(enic->vdev);
        eth_dev->data->dev_link.link_duplex = ETH_LINK_FULL_DUPLEX;
@@ -572,10 +573,17 @@ int enic_enable(struct enic *enic)
        }
 
        /*
-        * Use the simple TX handler if possible. All offloads must be
-        * disabled.
+        * Use the simple TX handler if possible. Only checksum offloads
+        * and vlan insertion are supported.
         */
-       if (eth_dev->data->dev_conf.txmode.offloads == 0) {
+       simple_tx_offloads = enic->tx_offload_capa &
+               (DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
+                DEV_TX_OFFLOAD_VLAN_INSERT |
+                DEV_TX_OFFLOAD_IPV4_CKSUM |
+                DEV_TX_OFFLOAD_UDP_CKSUM |
+                DEV_TX_OFFLOAD_TCP_CKSUM);
+       if ((eth_dev->data->dev_conf.txmode.offloads &
+            ~simple_tx_offloads) == 0) {
                PMD_INIT_LOG(DEBUG, " use the simple tx handler");
                eth_dev->tx_pkt_burst = &enic_simple_xmit_pkts;
                for (index = 0; index < enic->wq_count; index++)

diff --git a/drivers/net/enic/enic_rxtx.c b/drivers/net/enic/enic_rxtx.c
index 8d57c41..276a2e5 100644 (file)
--- a/drivers/net/enic/enic_rxtx.c
+++ b/drivers/net/enic/enic_rxtx.c
@@ -842,12 +842,33 @@ static void enqueue_simple_pkts(struct rte_mbuf **pkts,
                                struct enic *enic)
 {
        struct rte_mbuf *p;
+       uint16_t mss;
 
        while (n) {
                n--;
                p = *pkts++;
                desc->address = p->buf_iova + p->data_off;
                desc->length = p->pkt_len;
+               /* VLAN insert */
+               desc->vlan_tag = p->vlan_tci;
+               desc->header_length_flags &=
+                       ((1 << WQ_ENET_FLAGS_EOP_SHIFT) |
+                        (1 << WQ_ENET_FLAGS_CQ_ENTRY_SHIFT));
+               if (p->ol_flags & PKT_TX_VLAN) {
+                       desc->header_length_flags |=
+                               1 << WQ_ENET_FLAGS_VLAN_TAG_INSERT_SHIFT;
+               }
+               /*
+                * Checksum offload. We use WQ_ENET_OFFLOAD_MODE_CSUM, which
+                * is 0, so no need to set offload_mode.
+                */
+               mss = 0;
+               if (p->ol_flags & PKT_TX_IP_CKSUM)
+                       mss |= ENIC_CALC_IP_CKSUM << WQ_ENET_MSS_SHIFT;
+               if (p->ol_flags & PKT_TX_L4_MASK)
+                       mss |= ENIC_CALC_TCP_UDP_CKSUM << WQ_ENET_MSS_SHIFT;
+               desc->mss_loopback = mss;
+
                /*
                 * The app should not send oversized
                 * packets. tx_pkt_prepare includes a check as