DPDK 简易应用开发之路 3：实现ping（ARP ICMP 协议）

本机环境为 Ubuntu20.04 ，dpdk-stable-20.11.10
网卡IP为192.168.131.153 mac地址 00 0c 29 00 04 4d

理论基础

机器A内部没有机器B的物理地址，则 A ping B 的时候需要先发 arp 请求，以获取机器 B 的 MAC 地址。

获取 MAC 地址

如果 A 和 B 在同一网段内，A 会直接广播询问 谁有 B的 MAC 地址请告知我，物理机ping 虚拟机地址通过抓包来看：

虚拟机（192.168.131.153）通过NAT路由器与外部网络通信，而这个路由器通常会使用一个内部IP（如192.168.131.1）作为网关，所以这里不是本机的地址，而是虚拟网关的地址。

B 会向 A发送一个 arp 应答，告知 A ，192.168.131.153的MAC地址是00 0c 29 00 04 4d。

如果 A 和 B 不在同一网段内，该请求会经路由器进入主机 B所在的网段，然后进行 ARP 广播查找主机 B 的物理地址。

至此主机 A 就获得了主机 B 的物理地址，并将其写入到了本地的 ARP 缓存表。

发送 ICMP 报文

ping 实际上就是 A 向 B 发送ICMP请求， B 向 A 发送ICMP应答，形如下图（已经实现ping通的效果）：

代码流程

使用dpdk接管了一个网卡，当用主机去 ping 虚拟机中被 dpdk 接管的网卡的时候，程序需要读取该网卡中的数据包，判断数据包中的目的地址是不是我的IP地址，然后判断它是ARP请求还是ICMP请求，然后相应的返回一个对应的数据包。

当从该网卡读取了一个数据包，发现目的地址是我的IP地址后：

如果是ARP请求：需要封装一个ARP响应包，把我的MAC地址写进去并发送出去。封装ARP协议之后，还需要封装成以太网帧，即还需要添加首部。然后通过 dpdk 的 API 将数据包发送给目的主机。

如果是ICMP请求：需要封装一个ICMP应答包来响应该ICMP请求并发送出去。封装协议的过程，其实就是在协议的首部填充内容，例如填充你的MAC地址、源IP地址、目的IP地址、报文类型、首部大小、校验和等。

ARP数据包发送

encode_arp_pkt函数：利用目标MAC地址、源IP和目标IP构建ARP包。它首先设置以太网头部，包括源和目标MAC地址，以及以太网类型为ARP。设置ARP头部的各种字段，如硬件类型、协议类型、操作码（ARP应答），以及源和目标的MAC/IP地址。

send_arp函数：用于发送ARP包。它首先分配一个mbuf（数据包缓冲区），检查分配是否成功，然后设置数据包的长度，调用encode_arp_pkt函数构建ARP包，最后通过rte_eth_tx_burst发送该包，并在发送后释放mbuf以避免内存泄漏。

// 编码并构建ARP包
static void encode_arp_pkt(uint8_t *msg, uint8_t *dst_mac, uint32_t sip, uint32_t dip) {struct rte_ether_hdr *eth = (struct rte_ether_hdr *)msg;rte_memcpy(eth->s_addr.addr_bytes, src_mac, RTE_ETHER_ADDR_LEN);  // 设置源MACrte_memcpy(eth->d_addr.addr_bytes, dst_mac, RTE_ETHER_ADDR_LEN);  // 设置目标MACeth->ether_type = htons(RTE_ETHER_TYPE_ARP);  // 设置以太类型为ARPstruct rte_arp_hdr *arp = (struct rte_arp_hdr *)(eth + 1);arp->arp_hardware = htons(1);  // 硬件类型，1表示以太网arp->arp_protocol = htons(RTE_ETHER_TYPE_IPV4);  // 协议类型，IPv4arp->arp_hlen = RTE_ETHER_ADDR_LEN;  // 硬件地址长度arp->arp_plen = sizeof(uint32_t);  // 协议地址长度arp->arp_opcode = htons(RTE_ARP_OP_REPLY);  // 设置为ARP应答rte_memcpy(arp->arp_data.arp_sha.addr_bytes, src_mac, RTE_ETHER_ADDR_LEN);  // 源MAC地址rte_memcpy(arp->arp_data.arp_tha.addr_bytes, dst_mac, RTE_ETHER_ADDR_LEN);  // 目标MAC地址arp->arp_data.arp_sip = sip;  // 源IP地址arp->arp_data.arp_tip = dip;  // 目标IP地址
}// 发送ARP包
static void send_arp(struct rte_mempool *mbuf_pool, uint8_t *dst_mac, uint32_t sip, uint32_t dip) {const unsigned total_len = sizeof(struct rte_ether_hdr) + sizeof(struct rte_arp_hdr);struct rte_mbuf *mbuf = rte_pktmbuf_alloc(mbuf_pool);  // 分配内存池中的mbufif (mbuf == NULL) {rte_exit(EXIT_FAILURE, "Failed to allocate mbuf for ARP\n");}mbuf->pkt_len = total_len;mbuf->data_len = total_len;uint8_t *pktdata = rte_pktmbuf_mtod(mbuf, uint8_t*);  // 获取数据指针encode_arp_pkt(pktdata, dst_mac, sip, dip);  // 构建ARP包rte_eth_tx_burst(gDpdkPortId, 0, &mbuf, 1);  // 发送ARP包rte_pktmbuf_free(mbuf);  // 释放mbuf
}

ICMP数据包发送

encode_icmp_pkt ：负责构造ICMP包。首先设置以太网头，包括源和目标MAC地址，以及以太网类型为IPv4。接着，构建IP头，设置源和目标IP地址、TTL、协议类型（ICMP）和校验和。最后，构造ICMP头，设置类型为回显应答、标识符、序列号和ICMP校验和。

send_icmp：分配一个mbuf（数据包缓冲区），然后调用encode_icmp_pkt填充数据包。通过rte_eth_tx_burst函数发送ICMP包，最后释放mbuf。

// 编码并构建ICMP包
static void encode_icmp_pkt(uint8_t *msg, uint8_t *dst_mac, uint32_t sip, uint32_t dip, uint16_t id, uint16_t seqnb) {struct rte_ether_hdr *eth = (struct rte_ether_hdr *)msg;rte_memcpy(eth->s_addr.addr_bytes, src_mac, RTE_ETHER_ADDR_LEN);  // 设置源MACrte_memcpy(eth->d_addr.addr_bytes, dst_mac, RTE_ETHER_ADDR_LEN);  // 设置目标MACeth->ether_type = htons(RTE_ETHER_TYPE_IPV4);  // 设置以太类型为IPv4struct rte_ipv4_hdr *ip = (struct rte_ipv4_hdr *)(eth + 1);ip->version_ihl = 0x45;ip->type_of_service = 0;ip->total_length = htons(sizeof(struct rte_icmp_hdr) + sizeof(struct rte_ipv4_hdr));  // IP报文长度ip->packet_id = 0;ip->fragment_offset = 0;ip->time_to_live = 64;ip->next_proto_id = IPPROTO_ICMP;  // 下一层协议为ICMPip->src_addr = sip;  // 设置源IP地址ip->dst_addr = dip;  // 设置目标IP地址ip->hdr_checksum = 0;ip->hdr_checksum = rte_ipv4_cksum(ip);  // 计算并设置IP头校验和struct rte_icmp_hdr *icmp = (struct rte_icmp_hdr *)(ip + 1);icmp->icmp_type = RTE_IP_ICMP_ECHO_REPLY;  // 设置ICMP类型为回显应答icmp->icmp_code = 0;icmp->icmp_ident = id;icmp->icmp_seq_nb = seqnb;icmp->icmp_cksum = 0;icmp->icmp_cksum = checksum((uint16_t *)icmp, sizeof(struct rte_icmp_hdr));  // 计算ICMP校验和
}
// 发送ICMP包
static void send_icmp(struct rte_mempool *mbuf_pool, uint8_t *dst_mac, uint32_t sip, uint32_t dip, uint16_t id, uint16_t seqnb) {const unsigned total_len = sizeof(struct rte_ether_hdr) + sizeof(struct rte_ipv4_hdr) + sizeof(struct rte_icmp_hdr);struct rte_mbuf *mbuf = rte_pktmbuf_alloc(mbuf_pool);  // 分配内存池中的mbufif (mbuf == NULL) {rte_exit(EXIT_FAILURE, "Failed to allocate mbuf for ICMP\n");}mbuf->pkt_len = total_len;mbuf->data_len = total_len;uint8_t *pktdata = rte_pktmbuf_mtod(mbuf, uint8_t*);  // 获取数据指针encode_icmp_pkt(pktdata, dst_mac, sip, dip, id, seqnb);  // 构建ICMP包rte_eth_tx_burst(gDpdkPortId, 0, &mbuf, 1);  // 发送ICMP包rte_pktmbuf_free(mbuf);  // 释放mbuf
}

应答处理

实现了ARP和ICMP协议的基本处理逻辑。handle_arp用于回应ARP请求，而handle_ipv4_icmp则处理ICMP回显请求。主要就是通过指针运算来解析头部数据。

// 处理接收到的ARP请求并发送ARP应答
static void handle_arp(struct rte_mbuf *mbuf, struct rte_mempool *mbuf_pool) {struct rte_ether_hdr *ehdr = rte_pktmbuf_mtod(mbuf, struct rte_ether_hdr*);struct rte_arp_hdr *ahdr = (struct rte_arp_hdr *)(ehdr + 1);if (ahdr->arp_data.arp_tip == LOCAL_IP) {send_arp(mbuf_pool, ahdr->arp_data.arp_sha.addr_bytes, ahdr->arp_data.arp_tip, ahdr->arp_data.arp_sip);  // 发送ARP应答}rte_pktmbuf_free(mbuf);  // 释放mbuf
}// 处理接收到的ICMP回显请求并发送回显应答
static void handle_ipv4_icmp(struct rte_mbuf *mbuf, struct rte_mempool *mbuf_pool) {struct rte_ether_hdr *ehdr = rte_pktmbuf_mtod(mbuf, struct rte_ether_hdr*);struct rte_ipv4_hdr *iphdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_ipv4_hdr*, sizeof(struct rte_ether_hdr));struct rte_icmp_hdr *icmphdr = (struct rte_icmp_hdr *)(iphdr + 1);if (icmphdr->icmp_type == RTE_IP_ICMP_ECHO_REQUEST) {send_icmp(mbuf_pool, ehdr->s_addr.addr_bytes, iphdr->dst_addr, iphdr->src_addr, icmphdr->icmp_ident, icmphdr->icmp_seq_nb);  // 发送ICMP应答}rte_pktmbuf_free(mbuf);  // 释放mbuf
}

rte_pktmbuf_mtod_offset用于将数据包的mbuf指针转换为特定类型的指针，同时可以指定一个偏移量。这个函数在处理多层协议时特别有用，因为它允许你直接访问mbuf中某个特定偏移位置的数据。

完整代码

完整代码也可见 github

#include <rte_eal.h>
#include <rte_ethdev.h>
#include <rte_mbuf.h>
#include <stdio.h>
#include <stdint.h>
#include <arpa/inet.h>#define NUM_MBUFS (8192)  // 增加了用于更大缓冲池的MBUF数量
#define BURST_SIZE 32  // 每次从接收队列中读取的包数量
#define MAKE_IPV4_ADDR(a, b, c, d) (a + (b<<8) + (c<<16) + (d<<24))static const uint32_t LOCAL_IP = MAKE_IPV4_ADDR(192, 168, 131, 153);  // 本地IP地址
static uint8_t src_mac[RTE_ETHER_ADDR_LEN];  // 本地MAC地址int gDpdkPortId = 0;  // 端口ID// 默认端口配置
static const struct rte_eth_conf port_conf_default = {.rxmode = {.max_rx_pkt_len = RTE_ETHER_MAX_LEN,},
};// 初始化DPDK端口
static void init_port(struct rte_mempool *mbuf_pool) {uint16_t nb_sys_ports = rte_eth_dev_count_avail();if (nb_sys_ports == 0) {rte_exit(EXIT_FAILURE, "No Ethernet ports available\n");}struct rte_eth_dev_info dev_info;rte_eth_dev_info_get(gDpdkPortId, &dev_info);const int num_rx_queues = 1;const int num_tx_queues = 1;struct rte_eth_conf port_conf = port_conf_default;rte_eth_dev_configure(gDpdkPortId, num_rx_queues, num_tx_queues, &port_conf);if (rte_eth_rx_queue_setup(gDpdkPortId, 0, 128, rte_eth_dev_socket_id(gDpdkPortId), NULL, mbuf_pool) < 0) {rte_exit(EXIT_FAILURE, "Failed to set up RX queue\n");}struct rte_eth_txconf txq_conf = dev_info.default_txconf;txq_conf.offloads = port_conf.rxmode.offloads;if (rte_eth_tx_queue_setup(gDpdkPortId, 0, 1024, rte_eth_dev_socket_id(gDpdkPortId), &txq_conf) < 0) {rte_exit(EXIT_FAILURE, "Failed to set up TX queue\n");}if (rte_eth_dev_start(gDpdkPortId) < 0) {rte_exit(EXIT_FAILURE, "Failed to start Ethernet device\n");}rte_eth_promiscuous_enable(gDpdkPortId);  // 启用混杂模式，接收所有包
}// 计算校验和
static uint16_t checksum(uint16_t *addr, int count) {register long sum = 0;while (count > 1) {sum += *(unsigned short*)addr++;count -= 2;}if (count > 0) {sum += *(unsigned char *)addr;}while (sum >> 16) {sum = (sum & 0xffff) + (sum >> 16);}return (uint16_t)~sum;
}// 编码并构建ICMP包
static void encode_icmp_pkt(uint8_t *msg, uint8_t *dst_mac, uint32_t sip, uint32_t dip, uint16_t id, uint16_t seqnb) {struct rte_ether_hdr *eth = (struct rte_ether_hdr *)msg;rte_memcpy(eth->s_addr.addr_bytes, src_mac, RTE_ETHER_ADDR_LEN);  // 设置源MACrte_memcpy(eth->d_addr.addr_bytes, dst_mac, RTE_ETHER_ADDR_LEN);  // 设置目标MACeth->ether_type = htons(RTE_ETHER_TYPE_IPV4);  // 设置以太类型为IPv4struct rte_ipv4_hdr *ip = (struct rte_ipv4_hdr *)(eth + 1);ip->version_ihl = 0x45;ip->type_of_service = 0;ip->total_length = htons(sizeof(struct rte_icmp_hdr) + sizeof(struct rte_ipv4_hdr));  // IP报文长度ip->packet_id = 0;ip->fragment_offset = 0;ip->time_to_live = 64;ip->next_proto_id = IPPROTO_ICMP;  // 下一层协议为ICMPip->src_addr = sip;  // 设置源IP地址ip->dst_addr = dip;  // 设置目标IP地址ip->hdr_checksum = 0;ip->hdr_checksum = rte_ipv4_cksum(ip);  // 计算并设置IP头校验和struct rte_icmp_hdr *icmp = (struct rte_icmp_hdr *)(ip + 1);icmp->icmp_type = RTE_IP_ICMP_ECHO_REPLY;  // 设置ICMP类型为回显应答icmp->icmp_code = 0;icmp->icmp_ident = id;icmp->icmp_seq_nb = seqnb;icmp->icmp_cksum = 0;icmp->icmp_cksum = checksum((uint16_t *)icmp, sizeof(struct rte_icmp_hdr));  // 计算ICMP校验和// 输出ICMP报文相关信息printf("ICMP Packet:\n");printf("  Source MAC: %02X:%02X:%02X:%02X:%02X:%02X\n", eth->s_addr.addr_bytes[0], eth->s_addr.addr_bytes[1], eth->s_addr.addr_bytes[2],eth->s_addr.addr_bytes[3], eth->s_addr.addr_bytes[4], eth->s_addr.addr_bytes[5]);printf("  Destination MAC: %02X:%02X:%02X:%02X:%02X:%02X\n", eth->d_addr.addr_bytes[0], eth->d_addr.addr_bytes[1], eth->d_addr.addr_bytes[2],eth->d_addr.addr_bytes[3], eth->d_addr.addr_bytes[4], eth->d_addr.addr_bytes[5]);printf("  Source IP: %s\n", inet_ntoa(*(struct in_addr *)&sip));printf("  Destination IP: %s\n", inet_ntoa(*(struct in_addr *)&dip));
}// 发送ICMP包
static void send_icmp(struct rte_mempool *mbuf_pool, uint8_t *dst_mac, uint32_t sip, uint32_t dip, uint16_t id, uint16_t seqnb) {const unsigned total_len = sizeof(struct rte_ether_hdr) + sizeof(struct rte_ipv4_hdr) + sizeof(struct rte_icmp_hdr);struct rte_mbuf *mbuf = rte_pktmbuf_alloc(mbuf_pool);  // 分配内存池中的mbufif (mbuf == NULL) {rte_exit(EXIT_FAILURE, "Failed to allocate mbuf for ICMP\n");}mbuf->pkt_len = total_len;mbuf->data_len = total_len;uint8_t *pktdata = rte_pktmbuf_mtod(mbuf, uint8_t*);  // 获取数据指针encode_icmp_pkt(pktdata, dst_mac, sip, dip, id, seqnb);  // 构建ICMP包rte_eth_tx_burst(gDpdkPortId, 0, &mbuf, 1);  // 发送ICMP包rte_pktmbuf_free(mbuf);  // 释放mbuf
}// 编码并构建ARP包
static void encode_arp_pkt(uint8_t *msg, uint8_t *dst_mac, uint32_t sip, uint32_t dip) {struct rte_ether_hdr *eth = (struct rte_ether_hdr *)msg;rte_memcpy(eth->s_addr.addr_bytes, src_mac, RTE_ETHER_ADDR_LEN);  // 设置源MACrte_memcpy(eth->d_addr.addr_bytes, dst_mac, RTE_ETHER_ADDR_LEN);  // 设置目标MACeth->ether_type = htons(RTE_ETHER_TYPE_ARP);  // 设置以太类型为ARPstruct rte_arp_hdr *arp = (struct rte_arp_hdr *)(eth + 1);arp->arp_hardware = htons(1);  // 硬件类型，1表示以太网arp->arp_protocol = htons(RTE_ETHER_TYPE_IPV4);  // 协议类型，IPv4arp->arp_hlen = RTE_ETHER_ADDR_LEN;  // 硬件地址长度arp->arp_plen = sizeof(uint32_t);  // 协议地址长度arp->arp_opcode = htons(RTE_ARP_OP_REPLY);  // 设置为ARP应答rte_memcpy(arp->arp_data.arp_sha.addr_bytes, src_mac, RTE_ETHER_ADDR_LEN);  // 源MAC地址rte_memcpy(arp->arp_data.arp_tha.addr_bytes, dst_mac, RTE_ETHER_ADDR_LEN);  // 目标MAC地址arp->arp_data.arp_sip = sip;  // 源IP地址arp->arp_data.arp_tip = dip;  // 目标IP地址// 输出ARP报文相关信息printf("ARP Packet:\n");printf("  Source MAC: %02X:%02X:%02X:%02X:%02X:%02X\n", eth->s_addr.addr_bytes[0], eth->s_addr.addr_bytes[1], eth->s_addr.addr_bytes[2],eth->s_addr.addr_bytes[3], eth->s_addr.addr_bytes[4], eth->s_addr.addr_bytes[5]);printf("  Destination MAC: %02X:%02X:%02X:%02X:%02X:%02X\n", eth->d_addr.addr_bytes[0], eth->d_addr.addr_bytes[1], eth->d_addr.addr_bytes[2],eth->d_addr.addr_bytes[3], eth->d_addr.addr_bytes[4], eth->d_addr.addr_bytes[5]);printf("  Source IP: %s\n", inet_ntoa(*(struct in_addr *)&sip));printf("  Destination IP: %s\n", inet_ntoa(*(struct in_addr *)&dip));
}// 发送ARP包
static void send_arp(struct rte_mempool *mbuf_pool, uint8_t *dst_mac, uint32_t sip, uint32_t dip) {const unsigned total_len = sizeof(struct rte_ether_hdr) + sizeof(struct rte_arp_hdr);struct rte_mbuf *mbuf = rte_pktmbuf_alloc(mbuf_pool);  // 分配内存池中的mbufif (mbuf == NULL) {rte_exit(EXIT_FAILURE, "Failed to allocate mbuf for ARP\n");}mbuf->pkt_len = total_len;mbuf->data_len = total_len;uint8_t *pktdata = rte_pktmbuf_mtod(mbuf, uint8_t*);  // 获取数据指针encode_arp_pkt(pktdata, dst_mac, sip, dip);  // 构建ARP包rte_eth_tx_burst(gDpdkPortId, 0, &mbuf, 1);  // 发送ARP包rte_pktmbuf_free(mbuf);  // 释放mbuf
}// 处理接收到的ARP请求并发送ARP应答
static void handle_arp(struct rte_mbuf *mbuf, struct rte_mempool *mbuf_pool) {struct rte_ether_hdr *ehdr = rte_pktmbuf_mtod(mbuf, struct rte_ether_hdr*);struct rte_arp_hdr *ahdr = (struct rte_arp_hdr *)(ehdr + 1);if (ahdr->arp_data.arp_tip == LOCAL_IP) {send_arp(mbuf_pool, ahdr->arp_data.arp_sha.addr_bytes, ahdr->arp_data.arp_tip, ahdr->arp_data.arp_sip);  // 发送ARP应答}rte_pktmbuf_free(mbuf);  // 释放mbuf
}// 处理接收到的ICMP回显请求并发送回显应答
static void handle_ipv4_icmp(struct rte_mbuf *mbuf, struct rte_mempool *mbuf_pool) {struct rte_ether_hdr *ehdr = rte_pktmbuf_mtod(mbuf, struct rte_ether_hdr*);struct rte_ipv4_hdr *iphdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_ipv4_hdr*, sizeof(struct rte_ether_hdr));struct rte_icmp_hdr *icmphdr = (struct rte_icmp_hdr *)(iphdr + 1);if (icmphdr->icmp_type == RTE_IP_ICMP_ECHO_REQUEST) {send_icmp(mbuf_pool, ehdr->s_addr.addr_bytes, iphdr->dst_addr, iphdr->src_addr, icmphdr->icmp_ident, icmphdr->icmp_seq_nb);  // 发送ICMP应答}rte_pktmbuf_free(mbuf);  // 释放mbuf
}int main(int argc, char *argv[]) {if (rte_eal_init(argc, argv) < 0) {rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");}struct rte_mempool *mbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", NUM_MBUFS,0, 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());if (mbuf_pool == NULL) {rte_exit(EXIT_FAILURE, "Could not create mbuf pool\n");}init_port(mbuf_pool);  // 初始化DPDK端口rte_eth_macaddr_get(gDpdkPortId, (struct rte_ether_addr *)src_mac);  // 获取本地MAC地址printf("DPDK initialized. Waiting for packets...\n");// 主循环，处理接收到的报文while (1) {struct rte_mbuf *mbufs[BURST_SIZE];unsigned num_recvd = rte_eth_rx_burst(gDpdkPortId, 0, mbufs, BURST_SIZE);  // 从接收队列中接收报文if (num_recvd == 0) {continue;}// 处理接收到的每个报文for (unsigned i = 0; i < num_recvd; i++) {struct rte_ether_hdr *ehdr = rte_pktmbuf_mtod(mbufs[i], struct rte_ether_hdr*);if (ehdr->ether_type == rte_cpu_to_be_16(RTE_ETHER_TYPE_ARP)) {handle_arp(mbufs[i], mbuf_pool);  // 处理ARP报文} else if (ehdr->ether_type == rte_cpu_to_be_16(RTE_ETHER_TYPE_IPV4)) {struct rte_ipv4_hdr *iphdr = rte_pktmbuf_mtod_offset(mbufs[i], struct rte_ipv4_hdr*, sizeof(struct rte_ether_hdr));if (iphdr->next_proto_id == IPPROTO_ICMP) {handle_ipv4_icmp(mbufs[i], mbuf_pool);  // 处理ICMP报文}} else {rte_pktmbuf_free(mbufs[i]);  // 如果报文类型不是ARP或ICMP，释放mbuf}}}return 0;
}

运行效果