examples/vhost: add virtio offload
authorJijiang Liu <jijiang.liu@intel.com>
Fri, 5 Feb 2016 07:31:41 +0000 (15:31 +0800)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Wed, 17 Feb 2016 21:56:44 +0000 (22:56 +0100)
Change the codes in vhost sample to test virtio offload feature.

These changes include,

1. add two test options: tx-csum and tso.

2. add virtio_tx_offload() function to test vhost TX offload feature
   for VM to NIC case;

however, for VM to VM case, it doesn't need to call this function,
  the reason is explained in patch 2.

Signed-off-by: Jijiang Liu <jijiang.liu@intel.com>
Acked-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
examples/vhost/main.c

index 4318397..e08f451 100644 (file)
@@ -51,6 +51,9 @@
 #include <rte_malloc.h>
 #include <rte_virtio_net.h>
 #include <rte_ip.h>
+#include <rte_tcp.h>
+#include <rte_udp.h>
+#include <rte_sctp.h>
 
 #include "main.h"
 
@@ -200,6 +203,13 @@ typedef enum {
 static uint32_t enable_stats = 0;
 /* Enable retries on RX. */
 static uint32_t enable_retry = 1;
+
+/* Disable TX checksum offload */
+static uint32_t enable_tx_csum;
+
+/* Disable TSO offload */
+static uint32_t enable_tso;
+
 /* Specify timeout (in useconds) between retries on RX. */
 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
 /* Specify the number of retries on RX. */
@@ -430,6 +440,14 @@ port_init(uint8_t port)
 
        if (port >= rte_eth_dev_count()) return -1;
 
+       if (enable_tx_csum == 0)
+               rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_CSUM);
+
+       if (enable_tso == 0) {
+               rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO4);
+               rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO6);
+       }
+
        rx_rings = (uint16_t)dev_info.max_rx_queues;
        /* Configure ethernet device. */
        retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
@@ -565,7 +583,9 @@ us_vhost_usage(const char *prgname)
        "               --rx-desc-num [0-N]: the number of descriptors on rx, "
                        "used only when zero copy is enabled.\n"
        "               --tx-desc-num [0-N]: the number of descriptors on tx, "
-                       "used only when zero copy is enabled.\n",
+                       "used only when zero copy is enabled.\n"
+       "               --tx-csum [0|1] disable/enable TX checksum offload.\n"
+       "               --tso [0|1] disable/enable TCP segment offload.\n",
               prgname);
 }
 
@@ -591,6 +611,8 @@ us_vhost_parse_args(int argc, char **argv)
                {"zero-copy", required_argument, NULL, 0},
                {"rx-desc-num", required_argument, NULL, 0},
                {"tx-desc-num", required_argument, NULL, 0},
+               {"tx-csum", required_argument, NULL, 0},
+               {"tso", required_argument, NULL, 0},
                {NULL, 0, 0, 0},
        };
 
@@ -645,6 +667,28 @@ us_vhost_parse_args(int argc, char **argv)
                                }
                        }
 
+                       /* Enable/disable TX checksum offload. */
+                       if (!strncmp(long_option[option_index].name, "tx-csum", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, 1);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for tx-csum [0|1]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else
+                                       enable_tx_csum = ret;
+                       }
+
+                       /* Enable/disable TSO offload. */
+                       if (!strncmp(long_option[option_index].name, "tso", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, 1);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for tso [0|1]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else
+                                       enable_tso = ret;
+                       }
+
                        /* Specify the retries delay time (in useconds) on RX. */
                        if (!strncmp(long_option[option_index].name, "rx-retry-delay", MAX_LONG_OPT_SZ)) {
                                ret = parse_num_opt(optarg, INT32_MAX);
@@ -1103,6 +1147,58 @@ find_local_dest(struct virtio_net *dev, struct rte_mbuf *m,
        return 0;
 }
 
+static uint16_t
+get_psd_sum(void *l3_hdr, uint64_t ol_flags)
+{
+       if (ol_flags & PKT_TX_IPV4)
+               return rte_ipv4_phdr_cksum(l3_hdr, ol_flags);
+       else /* assume ethertype == ETHER_TYPE_IPv6 */
+               return rte_ipv6_phdr_cksum(l3_hdr, ol_flags);
+}
+
+static void virtio_tx_offload(struct rte_mbuf *m)
+{
+       void *l3_hdr;
+       struct ipv4_hdr *ipv4_hdr = NULL;
+       struct tcp_hdr *tcp_hdr = NULL;
+       struct udp_hdr *udp_hdr = NULL;
+       struct sctp_hdr *sctp_hdr = NULL;
+       struct ether_hdr *eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
+
+       l3_hdr = (char *)eth_hdr + m->l2_len;
+
+       if (m->tso_segsz != 0) {
+               ipv4_hdr = (struct ipv4_hdr *)l3_hdr;
+               tcp_hdr = (struct tcp_hdr *)((char *)l3_hdr + m->l3_len);
+               m->ol_flags |= PKT_TX_IP_CKSUM;
+               ipv4_hdr->hdr_checksum = 0;
+               tcp_hdr->cksum = get_psd_sum(l3_hdr, m->ol_flags);
+               return;
+       }
+
+       if (m->ol_flags & PKT_TX_L4_MASK) {
+               switch (m->ol_flags & PKT_TX_L4_MASK) {
+               case PKT_TX_TCP_CKSUM:
+                       tcp_hdr = (struct tcp_hdr *)
+                                       ((char *)l3_hdr + m->l3_len);
+                       tcp_hdr->cksum = get_psd_sum(l3_hdr, m->ol_flags);
+                       break;
+               case PKT_TX_UDP_CKSUM:
+                       udp_hdr = (struct udp_hdr *)
+                                       ((char *)l3_hdr + m->l3_len);
+                       udp_hdr->dgram_cksum = get_psd_sum(l3_hdr, m->ol_flags);
+                       break;
+               case PKT_TX_SCTP_CKSUM:
+                       sctp_hdr = (struct sctp_hdr *)
+                                       ((char *)l3_hdr + m->l3_len);
+                       sctp_hdr->cksum = 0;
+                       break;
+               default:
+                       break;
+               }
+       }
+}
+
 /*
  * This function routes the TX packet to the correct interface. This may be a local device
  * or the physical port.
@@ -1145,7 +1241,7 @@ virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m, uint16_t vlan_tag)
                        (vh->vlan_tci != vlan_tag_be))
                        vh->vlan_tci = vlan_tag_be;
        } else {
-               m->ol_flags = PKT_TX_VLAN_PKT;
+               m->ol_flags |= PKT_TX_VLAN_PKT;
 
                /*
                 * Find the right seg to adjust the data len when offset is
@@ -1169,6 +1265,9 @@ virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m, uint16_t vlan_tag)
                m->vlan_tci = vlan_tag;
        }
 
+       if ((m->ol_flags & PKT_TX_L4_MASK) || (m->ol_flags & PKT_TX_TCP_SEG))
+               virtio_tx_offload(m);
+
        tx_q->m_table[len] = m;
        len++;
        if (enable_stats) {
@@ -1834,7 +1933,7 @@ virtio_tx_route_zcp(struct virtio_net *dev, struct rte_mbuf *m,
                mbuf->buf_physaddr = m->buf_physaddr;
                mbuf->buf_addr = m->buf_addr;
        }
-       mbuf->ol_flags = PKT_TX_VLAN_PKT;
+       mbuf->ol_flags |= PKT_TX_VLAN_PKT;
        mbuf->vlan_tci = vlan_tag;
        mbuf->l2_len = sizeof(struct ether_hdr);
        mbuf->l3_len = sizeof(struct ipv4_hdr);