Linux下非轮询方式监控网卡连接状态

在Linux下监控网卡的连接状态有多种方式，我想要的方式，不是以轮询方式定时查询或主动获取某个值，而是在网卡连接状态变化时我的程序能收到通知。

通过搜索找到stackoverflow上的一帖子（地址见后面的参考）得到了线索，做法是：
1. 使用AF_NETLINK socket
2. 绑定到RTMGRP_LINK组
3. 等待接收RTM_NEWLINK和RTM_DELLINK类型的message
4. 解析收到的消息中ifinfomsg结构体的ifi_flags成员是否被设置了IFF_RUNNING

太多文字段落看着容易吃力啊，就让文字尽可能地作为代码注释吧，理论与实践相结合啊，

端上代码：

#include <stdio.h>#include <stdlib.h>#include <unistd.h>#include <signal.h>#include <errno.h>#include <string.h>#include <asm/types.h> /* 根据`man 7 rtnetlink`，包含此头文件。（此头文件不包含也能编译通过，但为了兼容性，或者说平台移植性，包含上吧，万一去掉后放到非Linux系统下编不过了呢） */#include <sys/socket.h>#include <sys/ioctl.h> /* `man 7 rtnetlink`中有"ifi_flags  contains  the  device  flags,  see  netdevice(7)"这句话，`man 7 netdevice`中提到包含此头文件。 *///#include <net/if.h> /* use the following header instead of this one */#include <linux/if.h> /* this one is suite for the Linux, more related to interact with Linux kernel's network interfaces */#include <linux/if_arp.h> /* 包含ARPHRD_LOOPBACK的定义 */#include <linux/netlink.h> /* 根据`man 7 rtnetlink`，包含此头文件 */#include <linux/rtnetlink.h> /* 根据`man 7 rtnetlink`，包含此头文件 */#define PRINTF(fmt, ...) printf("%s:%d: " fmt, __FUNCTION__, __LINE__, ## __VA_ARGS__)#define BUF_SIZE 4096 /* 看到`man 7 netlink`中reading netlink message那个EXAMPLE的buf的大小是4096，我也就写这么大了。 */static int init(struct sockaddr_nl *psa);static int deinit();static int msg_req();static int msg_loop(struct msghdr *pmh);static void sig_handler(int sig);static int sfd = -1;int main (int argc, char *argv[]){int ret = 0;/* 下面4行变量声明参考`man 7 netlink`中reading netlink message那个EXAMPLE，和`man 2 recvmsg`中"The recvmsg() call uses a msghdr structure to minimize the number of directly supplied arguments"这行之后的结构体说明。 */char buf[BUF_SIZE];struct iovec iov = {buf, sizeof(buf)};struct sockaddr_nl sa;struct msghdr msg = {(void *)&sa, sizeof(sa), &iov, 1, NULL, 0, 0};/* 程序的主要流程体现在下面4个函数调用： * 1. 初始化socket * 2. 请求获取当前link状态 * 3. 循环等待接收link状态的消息 * 4. 反初始化 */ret = init(&sa);if (!ret) {ret = msg_req();}if (!ret) {ret = msg_loop(&msg);}ret = deinit();return ret;}static int init(struct sockaddr_nl *psa){int ret = 0;struct sigaction sigact;sigact.sa_handler = sig_handler;if (!ret && -1 == sigemptyset(&sigact.sa_mask)) {PRINTF("ERROR! sigemptyset\n");ret = -1;}if (!ret && -1 == sigaction(SIGINT, &sigact, NULL)) { /* SIGINT用于保证按Ctrl+C时进程能够在退出前执行反初始化（资源释放） */PRINTF("ERROR! sigaction SIGINT\n");ret = -1;}if (!ret && -1 == sigaction(SIGTERM, &sigact, NULL)) { /* SIGINT用于保证按Ctrl+C时进程能够在退出前执行反初始化（资源释放） */PRINTF("ERROR! sigaction SIGTERM\n");ret = -1;}if (!ret) {/* `man 7 netlink`中有"the netlink protocol does not distinguish between datagram and raw sockets"这样一句话，所以socket函数的第2个参数填SOCK_RAW和SOCK_DGRAM均可。 *//* 根据`man 7 netlink`中netlink_family参数的解释，NETLINK_ROUTE作为netlink_family参数的值能够获取link状态信息。 */sfd = socket(AF_NETLINK, SOCK_DGRAM, NETLINK_ROUTE);if (-1 == sfd) {PRINTF("ERROR! socket: %s\n", strerror(errno));ret = -1;}}/* bind代码参考`man 7 netlink`中EXAMPLE部分的第1段代码 */memset(psa, 0, sizeof(*psa));psa->nl_family = AF_NETLINK;psa->nl_groups = RTMGRP_LINK; /* 绑定到RTMGRP_LINK组 */if (!ret && bind(sfd, (struct sockaddr *)psa, sizeof(*psa))) {PRINTF("ERROR! bind: %s\n", strerror(errno));ret = -1;}/* deinit if init failed */if (0 != ret) {deinit();}return ret;}static int deinit(){int ret = 0;if (-1 != sfd) {if (-1 == close(sfd)) {PRINTF("ERROR! close: %s\n", strerror(errno));ret = -1;}sfd = -1;}return ret;}static int msg_req(){int ret = 0;struct {struct nlmsghdr  nh;struct ifinfomsg ifimsg;} req;/* 发送消息来请求获取当前link状态，这是为了当程序启动后能够显示一下当前的网卡连接状态，否则程序运行时网卡连接状态没有变化的话，程序就一直没有任何输出 *//* 代码参考`man 3 rtnetlink`中的EXAMPLE，主要参考了http://fossies.org/linux/misc/open-fcoe-3.19.tar.gz/open-fcoe-3.19/fcoe-utils/lib/rtnetlink.c中的send_getlink_dump函数 */memset(&req, 0, sizeof(req));req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg)); /* NLMSG_LENGTH会返回包含nlmsghdr这个header和ifinfomsg这个payload加起来的size，返回的是对齐后的值 */req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP; /* 根据`man 7 netlink`中NLM_F_REQUEST的解释，所有请求类型的消息都要设置NLM_F_REQUEST；设置了NLM_F_DUMP就能够获取link状态，根据`man 7 netlink`，NLM_F_DUMP相当于NLM_F_ROOT|NLM_F_MATCH，`man 7 netlink`中说NLM_F_MATCH是"Not implemented yet"，但去掉NLM_F_DUMP，只设置NLM_F_MATCH也是有效果的。 */req.nh.nlmsg_type = RTM_GETLINK; /* `man 7 rtnetlink`中说RTM_GETLINK用于"get information about a specific network interface"，`man 7 rtnetlink`中的"These messages contain an ifinfomsg structure followed by a series of rtattr structures"这句话解释了RTM_GETLINK的消息结构 */req.ifimsg.ifi_family = AF_UNSPEC; /* 根据`man 7 rtnetlink`中ifinfomsg结构体的定义，恒为AF_UNSPEC */req.ifimsg.ifi_index = 0; /* http://man7.org/linux/man-pages/man7/rtnetlink.7.html中有"ifi_index is the unique interface index (since Linux 3.7, it is possible to feed a nonzero value with the RTM_NEWLINK message, thus creating a link with the given ifindex)"这句话，说明这个成员设置0没问题 */req.ifimsg.ifi_change = 0xFFFFFFFF; /* `man 7 rtnetlink`中有"ifi_change is reserved for future use and should be always set to 0xFFFFFFFF"这句话，所以设置为0xFFFFFFFF */if (-1 == send(sfd, &req, req.nh.nlmsg_len, 0)) {PRINTF("ERROR! send: %s\n", strerror(errno));ret = -1;}return ret;}static int msg_loop(struct msghdr *pmh){int ret = 0;ssize_t nread = -1;char *buf = (char *)(pmh->msg_iov->iov_base);struct nlmsghdr *nh;struct ifinfomsg *ifimsg;struct rtattr *rta;int attrlen;/* 循环体参考`man 7 netlink`中reading netlink message那个EXAMPLE，连注释都照搬过来了 :D */while (!ret) {nread = recvmsg(sfd, pmh, 0);if (-1 == nread) {PRINTF("ERROR! recvmsg: %s\n", strerror(errno));ret = -1;}for (nh = (struct nlmsghdr *)buf; /* `man 7 netlink`中"Netlink messages consist of a byte stream with one or multiple nlmsghdr headers and associated payload"这句话的周围解释了netlink消息的结构，所以消息的开头是nlmsghdr类型的header */!ret && NLMSG_OK(nh, nread); /* 根据`man 3 netlink`，NLMSG_OK用来检查收到的netlink消息是否OK */nh = NLMSG_NEXT(nh, nread)) { /* 根据`man 3 netlink`，NLMSG_NEXT用于定位到下一个nlmsghdr header的开头，注意这个宏会改变nread的值 *//* `man 3 netlink`中有"The caller must check if the current nlmsghdr didn't have the NLMSG_DONE set—this function doesn't return NULL on end"这句话，所以要检查NLMSG_DONE *//* The end of multipart message. */if (NLMSG_DONE == nh->nlmsg_type) {break;}if (NLMSG_ERROR == nh->nlmsg_type) {/* Do some error handling. */PRINTF("ERROR! NLMSG_ERROR\n");ret = -1;}/* Continue with parsing payload. */if (!ret && (RTM_NEWLINK == nh->nlmsg_type || RTM_DELLINK == nh->nlmsg_type)) { /* `man 7 rtnetlink`中的"These messages contain an ifinfomsg structure followed by a series of rtattr structures"这句话解释了RTM_NEWLINK和RTM_DELLINK的消息结构，前一句说明了RTM_NEWLINK和RTM_DELLINK分别用于"Create"和"remove" "information about a specific network interface"，link的"Create"和"remove"动作产生的消息是程序要解析的。 */ifimsg = (struct ifinfomsg *)NLMSG_DATA(nh); /* 根据`man 3 netlink`，NLMSG_DATA用于定位nlmsghdr header后面的payload的内存地址 */if (ARPHRD_LOOPBACK != ifimsg->ifi_type) { /* 排除loopback网卡适配器 *//* `man 7 rtnetlink`中的"These messages contain an ifinfomsg structure followed by a series of rtattr structures"这句话说明ifinfomsg之后的内容就是属性 */attrlen = nh->nlmsg_len - NLMSG_LENGTH(sizeof(struct ifinfomsg)); /* NLMSG_LENGTH返回的值是nlmsghdr和ifinfomsg加起来的长度，减去后，剩下的就是属性的长度 */for (rta = IFLA_RTA(ifimsg); /* IFLA_RTA用于定位ifinfomsg payload后面的属性的内存地址 */RTA_OK(rta, attrlen) && rta->rta_type <= IFLA_MAX; /* 根据`man 3 rtnetlink`，RTA_OK用来检查rta指向的属性是否有效 */rta = RTA_NEXT(rta, attrlen)) { /* 根据`man 3 rtnetlink`，RTA_NEXT用于定位到下一个属性的开头，注意这个宏会改变attrlen的值 */if (IFLA_IFNAME == rta->rta_type) { /* 当rta_type为IFLA_IFNAME时，属性包含的数据是网卡名字符串 */printf("%s: ", (char*)RTA_DATA(rta)); /* 打印网卡名，如eth0 */}}/* 实际上，当link up时，IFF_RUNNING和IFF_LOWER_UP都会被设置，link down时都会被清除，以IFF_LOWER_UP作为判断依据也行，但根据`man 7 netdevice`，IFF_LOWER_UP是从内核2.6.17版本才有的，所以使用IFF_RUNNING兼容性可能好点 */if (IFF_RUNNING & ifimsg->ifi_flags)printf("link up\n");elseprintf("link down\n");}}}}return ret;}static void sig_handler(int sig){exit(deinit());}

代码和那些man命令的运行环境为debian wheezy 7.7。

参考：
http://stackoverflow.com/questions/10340145/get-event-for-nic-ethernet-card-link-status-on-linux
man 3 netlink
man 7 netlink
man 3 rtnetlink
man 7 rtnetlink
man 7 netdevice
http://centaur.sch.bme.hu/~leait/projects/openwrt/
http://fossies.org/linux/misc/open-fcoe-3.19.tar.gz/open-fcoe-3.19/fcoe-utils/lib/rtnetlink.c