DPDK 使用網卡ip/tcp/udp checksum offload

感謝我的老大馬大師，我們花了很多時間才找出第三個步。

網卡ip/tcp/udp checksum offload的功能就是將報文的ip，tcp， udp校驗和的計算交給網卡計算，節省cpu的消耗。

第一步

在初始化網卡時打開網卡的checksum offload功能

.hw_ip_checksum = 1, /**< IP checksum offload disabled */

調用rte_eth_dev_configure()函數進行網卡配置;

static struct rte_eth_conf default_port_conf = {
    .rxmode = {
        .mq_mode = ETH_MQ_RX_RSS,
        .max_rx_pkt_len = ETHER_MAX_LEN,
        .split_hdr_size = 0, /**< hdr buf size */
        .header_split   = 0, /**< Header Split disabled */
        .hw_ip_checksum = 1, /**< IP checksum offload enabled */
        .hw_vlan_filter = 0, /**< VLAN filtering disabled */
        .hw_vlan_strip  = 0, /**< VLAN strip disabled. */
        .hw_vlan_extend = 0, /**< Extended VLAN disabled. */
        .jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
        .hw_strip_crc   = 0, /**< CRC stripped by hardware */
        .enable_lro     = 0, /**< LRO disabled */
    },
    .rx_adv_conf = {
        .rss_conf = {
            .rss_key = default_rsskey_40bytes,
            .rss_key_len = 40,
            .rss_hf = ETH_RSS_PROTO_MASK,
        },
    },
    .txmode = {
        .mq_mode = ETH_MQ_TX_NONE,
    },
};

第二步

文檔出處：http://dpdk.org/doc/api-2.2/rte__mbuf_8h.html

#define PKT_TX_L4_NO_CKSUM   (0ULL << 52)

Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved, 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware L4 checksum offload, the user needs to:

• fill l2_len and l3_len in mbuf
• set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
• set the flag PKT_TX_IPV4 or PKT_TX_IPV6
• calculate the pseudo header checksum and set it in the L4 header (only for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum(). For SCTP, set the crc field to 0.Disable L4 cksum of TX pkt.
  

上面的最後一個操作是計算l3頭部的校驗和設置到l4頭部校驗和裏面，是爲了讓網卡知道tcp僞頭部校驗和。

操作函數如下，操作函數主要關注的是上面提到的幾個操作，

flags是struct rte_mbuf *m裏的uint64_t ol_flags;        /**< Offload features. */

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);
}

process_inner_cksums函數只是dpdk例子中的，他的步驟就是完成上面幾步

static uint64_t
process_inner_cksums(struct ether_hdr *eth_hdr, union tunnel_offload_info *info)
{
        void *l3_hdr = NULL;
        uint8_t l4_proto;
        uint16_t ethertype;
        struct ipv4_hdr *ipv4_hdr;
        struct ipv6_hdr *ipv6_hdr;
        struct udp_hdr *udp_hdr;
        struct tcp_hdr *tcp_hdr;
        struct sctp_hdr *sctp_hdr;
        uint64_t ol_flags = 0;
        info->l2_len = sizeof(struct ether_hdr);
        ethertype = rte_be_to_cpu_16(eth_hdr->ether_type);
        if (ethertype == ETHER_TYPE_VLAN) {
                struct vlan_hdr *vlan_hdr = (struct vlan_hdr *)(eth_hdr + 1);
                info->l2_len  += sizeof(struct vlan_hdr);
                ethertype = rte_be_to_cpu_16(vlan_hdr->eth_proto);
        }
        l3_hdr = (char *)eth_hdr + info->l2_len;
        if (ethertype == ETHER_TYPE_IPv4) {
                ipv4_hdr = (struct ipv4_hdr *)l3_hdr;
                ipv4_hdr->hdr_checksum = 0;
                ol_flags |= PKT_TX_IPV4;
                ol_flags |= PKT_TX_IP_CKSUM;
                info->l3_len = sizeof(struct ipv4_hdr);
                l4_proto = ipv4_hdr->next_proto_id;
        } else if (ethertype == ETHER_TYPE_IPv6) {
                ipv6_hdr = (struct ipv6_hdr *)l3_hdr;
                info->l3_len = sizeof(struct ipv6_hdr);
                l4_proto = ipv6_hdr->proto;
                ol_flags |= PKT_TX_IPV6;
        } else
                return 0; /* packet type not supported, nothing to do */
        if (l4_proto == IPPROTO_UDP) {
                udp_hdr = (struct udp_hdr *)((char *)l3_hdr + info->l3_len);
                ol_flags |= PKT_TX_UDP_CKSUM;
                udp_hdr->dgram_cksum = get_psd_sum(l3_hdr, ol_flags);
        } else if (l4_proto == IPPROTO_TCP) {
                tcp_hdr = (struct tcp_hdr *)((char *)l3_hdr + info->l3_len);
                ol_flags |= PKT_TX_TCP_CKSUM;
                tcp_hdr->cksum = get_psd_sum(l3_hdr, ol_flags);
                if (tso_segsz != 0) {
                        ol_flags |= PKT_TX_TCP_SEG;
                        info->tso_segsz = tso_segsz;
                        info->l4_len = sizeof(struct tcp_hdr);
                }
        } else if (l4_proto == IPPROTO_SCTP) {
                sctp_hdr = (struct sctp_hdr *)((char *)l3_hdr + info->l3_len);
                sctp_hdr->cksum = 0;
                ol_flags |= PKT_TX_SCTP_CKSUM;
        }
        return ol_flags;
}

第三步：

通過rte_eth_dev_info_get(port_id, &dev_info)獲取設備信息，如果不支持CKSUM，要清除對應位，然後初始化發送隊列的時候傳入dev_info.default_txconf

        /* Disable features that are not supported by port's HW */
        if (!(dev_info.tx_offload_capa & DEV_TX_OFFLOAD_UDP_CKSUM)) {
            dev_info.default_txconf.txq_flags |= ETH_TXQ_FLAGS_NOXSUMUDP;
        }


        if (!(dev_info.tx_offload_capa & DEV_TX_OFFLOAD_TCP_CKSUM)) {
            dev_info.default_txconf.txq_flags |= ETH_TXQ_FLAGS_NOXSUMTCP;
        }


        if (!(dev_info.tx_offload_capa & DEV_TX_OFFLOAD_SCTP_CKSUM)) {
            dev_info.default_txconf.txq_flags |= ETH_TXQ_FLAGS_NOXSUMSCTP;
        }
        rte_eth_tx_queue_setup(port_id, q, TX_QUEUE_SIZE,
                socketid, &dev_info.default_txconf);

參考例子：dpdk源碼裏app裏的test-pmd程序，裏面的process_inner_cksums有相關操作。

備註：以上是實現這個功能的步驟，理解爲主。