Linux Kernel 学习笔记9：内核与用户层通信之netlink

(本章基于：Linux-4.4.0-37)

内核与用户空间通信有很多种通信方式，netlink是其中一种，其余的还有/proc、ioctl、sockopt、共享内存等等。netlink的特点是异步全双工。

netlink使用32位端口寻址，称为pid（与进程号没有关系），其中内核的pid地址为0,。netlink主要特性如下：

1 支持全双工、异步通信(当然同步也支持)

2 用户空间可使用标准的BSD socket接口(但netlink并没有屏蔽掉协议包的构造与解析过程，推荐使用libnl等第三方库)

3 在内核空间使用专用的内核API接口

4 支持多播(因此支持“总线”式通信，可实现消息订阅)

5 在内核端可用于进程上下文与中断上下文

基本数据结构

struct msghdr {
void         *msg_name;       /* optional address */
socklen_t     msg_namelen;    /* size of address */
struct iovec *msg_iov;        /* scatter/gather array */
size_t        msg_iovlen;     /* # elements in msg_iov */
void         *msg_control;    /* ancillary data, see below */
size_t        msg_controllen; /* ancillary data buffer len */
int           msg_flags;      /* flags (unused) */
};

struct sockaddr_nl
{
sa_family_t nl_family; /*该字段总是为AF_NETLINK */
unsigned short nl_pad; /* 目前未用到，填充为0*/
__u32 nl_pid; /* process pid */
__u32 nl_groups; /* multicast groups mask */
};

struct sockaddr_nl是netlink通信地址，和我们通常socket编程中的sockaddr_in作用一样。pid表示通信端口，groups表示组，注意这里为希望加入多播组号的掩码，也就是说最多只支持32个组。

struct nlmsghdr
{
__u32 nlmsg_len; /* Length of message including header */
__u16 nlmsg_type; /* Message content */
__u16 nlmsg_flags; /* Additional flags */
__u32 nlmsg_seq; /* Sequence number */
__u32 nlmsg_pid; /* Sending process PID */
};

Netlink报文的数据区由消息头和消息体构成，struct nlmsghdr即为消息头，消息体接在消息头后。

内核层操作

创建socket

static inline struct sock *
netlink_kernel_create(struct net *net, int unit, struct netlink_kernel_cfg *cfg);

net:  一般直接填&init_net
unit：协议类型，可自定义，如#define NETLINK_TEST 25

cfg：配置结构，类型如下：

/* optional Netlink kernel configuration parameters */
struct netlink_kernel_cfg {
unsigned int	groups;
unsigned int	flags;
void		(*input)(struct sk_buff *skb);
struct mutex	*cb_mutex;
int		(*bind)(struct net *net, int group);
void		(*unbind)(struct net *net, int group);
bool		(*compare)(struct net *net, struct sock *sk);
};

groups：组编号；
input：接收回调函数，接收一个sk_buff结构，数据包含一个nlmsghdr协议头；

return：返回一个sock结构，返回NULL表示创建失败；

单播发送接口：

extern int netlink_unicast(struct sock *ssk, struct sk_buff *skb, __u32 portid, int nonblock);

(1) ssk：为函数 netlink_kernel_create()返回的socket。

(2) skb：存放消息，它的data字段指向要发送的netlink消息结构，而 skb的控制块保存了消息的地址信息，宏NETLINK_CB(skb)就用于方便设置该控制块。

(3) portid：pid端口。

(4) nonblock：表示该函数是否为非阻塞，如果为1，该函数将在没有接收缓存可利用时立即返回；而如果为0，该函数在没有接收缓存可利用定时睡眠。

多播发送接口：

extern int netlink_broadcast(struct sock *ssk, struct sk_buff *skb, __u32 portid,
__u32 group, gfp_t allocation);

group：接收消息的多播组，该参数的每一个位代表一个多播组，因此如果发送给多个多播组；

allocation：内存分配类型，一般地为GFP_ATOMIC或GFP_KERNEL，GFP_ATOMIC用于原子的上下文（即不可以睡眠），而GFP_KERNEL用于非原子上下文。

释放socket

extern void netlink_kernel_release(struct sock *sk);

用户层操作

nlmsghdr结构常见操作：

NLMSG_SPACE(len): 将len加上nlmsghdr头长度，并按4字节对齐；

NLMSG_DATA(nlh): 返回数据区首地址；

创建socke

int netlink_create_socket(void)
{
//create a socket
return socket(AF_NETLINK, SOCK_RAW, NETLINK_TEST);
}

bind

int netlink_bind(int sock_fd)
{
struct sockaddr_nl addr;

memset(&addr, 0, sizeof(struct sockaddr_nl));
addr.nl_family = AF_NETLINK;
addr.nl_pid = TEST_PID;
addr.nl_groups = 0;

return bind(sock_fd, (struct sockaddr *)&addr, sizeof(struct sockaddr_nl));
}

发送接收：

使用sendmsg、recvmsg发送接收数据

ssize_t sendmsg(int sockfd, const struct msghdr *msg, int flags);
ssize_t recvmsg(int sockfd, struct msghdr *msg, int flags);

使用sendto、recvfrom发送接收数据

ssize_t sendto(int sockfd, const void *buf, size_t len, int flags,
const struct sockaddr *dest_addr, socklen_t addrlen);
ssize_t recvfrom(int sockfd, void *buf, size_t len, int flags,
struct sockaddr *src_addr, socklen_t *addrlen);

例：

说明：用户层pid设置为100，应用层发送一条信息到内核，内核回复同样的信息；


内核层：

#include <linux/init.h>
#include <linux/module.h>
#include <linux/stat.h>
#include <linux/kdev_t.h>
#include <linux/fs.h>
#include <linux/device.h>
#include <linux/cdev.h>
#include <asm/uaccess.h>

#include <net/netlink.h>
#include <net/sock.h>

#define NETLINK_TEST (25)

static dev_t devId;
static struct class *cls = NULL;
struct sock *nl_sk = NULL;

static void
hello_cleanup(void)
{
netlink_kernel_release(nl_sk);
device_destroy(cls, devId);
class_destroy(cls);
unregister_chrdev_region(devId, 1);
}

static void
netlink_send(int pid, uint8_t *message, int len)
{
struct sk_buff *skb_1;
struct nlmsghdr *nlh;

if(!message || !nl_sk) {
return;
}

skb_1 = alloc_skb(NLMSG_SPACE(len), GFP_KERNEL);
if( !skb_1 ) {
printk(KERN_ERR "alloc_skb error!\n");
}

nlh = nlmsg_put(skb_1, 0, 0, 0, len, 0);
NETLINK_CB(skb_1).portid = 0;
NETLINK_CB(skb_1).dst_group = 0;
memcpy(NLMSG_DATA(nlh), message, len);
netlink_unicast(nl_sk, skb_1, pid, MSG_DONTWAIT);
}

static void
netlink_input(struct sk_buff *__skb)
{
struct sk_buff *skb;
char str[100];
struct nlmsghdr *nlh;

if( !__skb ) {
return;
}

skb = skb_get(__skb);
if( skb->len < NLMSG_SPACE(0)) {
return;
}

nlh = nlmsg_hdr(skb);
memset(str, 0, sizeof(str));
memcpy(str, NLMSG_DATA(nlh), sizeof(str));
printk(KERN_INFO "receive message (pid:%d):%s\n", nlh->nlmsg_pid, str);
printk(KERN_INFO "space:%d\n", NLMSG_SPACE(0));
printk(KERN_INFO "size:%d\n", nlh->nlmsg_len);
netlink_send(nlh->nlmsg_pid, NLMSG_DATA(nlh), nlh->nlmsg_len - NLMSG_SPACE(0));

return;
}

static __init int netlink_init(void)
{
int result;
struct netlink_kernel_cfg nkc;

printk(KERN_WARNING "netlink init start!\n");

//动态注册设备号
if(( result = alloc_chrdev_region(&devId, 0, 1, "stone-alloc-dev") ) != 0) {
printk(KERN_WARNING "register dev id error:%d\n", result);
goto err;
} else {
printk(KERN_WARNING "register dev id success!\n");
}
//动态创建设备节点
cls = class_create(THIS_MODULE, "stone-class");
if(IS_ERR(cls)) {
printk(KERN_WARNING "create class error!\n");
goto err;
}
if(device_create(cls, NULL, devId, "", "hello%d", 0) == NULL) {
printk(KERN_WARNING "create device error!\n");
goto err;
}

//初始化netlink
nkc.groups = 0;
nkc.flags = 0;
nkc.input = netlink_input;
nkc.cb_mutex = NULL;
nkc.bind = NULL;
nkc.unbind = NULL;
nkc.compare = NULL;
nl_sk = netlink_kernel_create(&init_net, NETLINK_TEST, &nkc);
if( !nl_sk ) {
printk(KERN_ERR "[netlink] create netlink socket error!\n");
goto err;
}

printk(KERN_ALERT "netlink init success!\n");
return 0;
err:
hello_cleanup();
return -1;
}

static __exit void netlink_exit(void)
{
hello_cleanup();
printk(KERN_WARNING "netlink exit!\n");
}

module_init(netlink_init);
module_exit(netlink_exit);

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Stone");

应用层1：

#include <sys/stat.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/socket.h>
#include <string.h>
#include <asm/types.h>
#include <linux/netlink.h>
#include <linux/socket.h>
#include <errno.h>

#define NETLINK_TEST    (25)
#define MAX_PAYLOAD     (1024)
#define TEST_PID        (100)

int netlink_create_socket(void)
{
//create a socket
return socket(AF_NETLINK, SOCK_RAW, NETLINK_TEST);
}
int netlink_bind(int sock_fd)
{
struct sockaddr_nl addr;

memset(&addr, 0, sizeof(struct sockaddr_nl));
addr.nl_family = AF_NETLINK;
addr.nl_pid = TEST_PID;
addr.nl_groups = 0;

return bind(sock_fd, (struct sockaddr *)&addr, sizeof(struct sockaddr_nl));
}
int
netlink_send_message(int sock_fd, const unsigned char *message, int len,
unsigned int pid, unsigned int group)
{
struct nlmsghdr *nlh = NULL;
struct sockaddr_nl dest_addr;
struct iovec iov;
struct msghdr msg;

if( !message ) {
return -1;
}

//create message
nlh = (struct nlmsghdr *)malloc(NLMSG_SPACE(len));
if( !nlh ) {
perror("malloc");
return -2;
}
nlh->nlmsg_len = NLMSG_SPACE(len);
nlh->nlmsg_pid = TEST_PID;
nlh->nlmsg_flags = 0;
memcpy(NLMSG_DATA(nlh), message, len);

iov.iov_base = (void *)nlh;
iov.iov_len = nlh->nlmsg_len;
memset(&dest_addr, 0, sizeof(struct sockaddr_nl));
dest_addr.nl_family = AF_NETLINK;
dest_addr.nl_pid = pid;
dest_addr.nl_groups = group;

memset(&msg, 0, sizeof(struct msghdr));
msg.msg_name = (void *)&dest_addr;
msg.msg_namelen = sizeof(struct sockaddr_nl);
msg.msg_iov = &iov;
msg.msg_iovlen = 1;

//send message
if( sendmsg(sock_fd, &msg, 0) < 0 )
{
printf("send error!\n");
free(nlh);
return -3;
}

free(nlh);
return 0;
}

int
netlink_recv_message(int sock_fd, unsigned char *message, int *len)
{
struct nlmsghdr *nlh = NULL;
struct sockaddr_nl source_addr;
struct iovec iov;
struct msghdr msg;

if( !message || !len ) {
return -1;
}

//create message
nlh = (struct nlmsghdr *)malloc(NLMSG_SPACE(MAX_PAYLOAD));
if( !nlh ) {
perror("malloc");
return -2;
}
iov.iov_base = (void *)nlh;
iov.iov_len = NLMSG_SPACE(MAX_PAYLOAD);
memset(&source_addr, 0, sizeof(struct sockaddr_nl));
memset(&msg, 0, sizeof(struct msghdr));
msg.msg_name = (void *)&source_addr;
msg.msg_namelen = sizeof(struct sockaddr_nl);
msg.msg_iov = &iov;
msg.msg_iovlen = 1;

if ( recvmsg(sock_fd, &msg, 0) < 0 ) {
printf("recvmsg error!\n");
return -3;
}
*len = nlh->nlmsg_len - NLMSG_SPACE(0);
memcpy(message, (unsigned char *)NLMSG_DATA(nlh), *len);

free(nlh);
return 0;
}

int
main(int argc, char **argv)
{
int sock_fd;
char buf[MAX_PAYLOAD];
int len;

if( argc < 2) {
printf("enter message!\n");
exit(EXIT_FAILURE);
}

sock_fd = netlink_create_socket();
if(sock_fd == -1) {
printf("socket error!\n");
return -1;
}

if( netlink_bind(sock_fd) < 0 ) {
perror("bind");
close(sock_fd);
exit(EXIT_FAILURE);
}

netlink_send_message(sock_fd, argv[1], strlen(argv[1]) + 1, 0, 0);
if( netlink_recv_message(sock_fd, buf, &len) == 0 ) {
printf("recv:%s len:%d\n", buf, len);
}

close(sock_fd);
return 0;
}

上面例子讲的是使用sendmsg、recvmsg。在应用层我们同样可以使用sendto、recvfrom发送接收数据，操作模式和UDP通信非常相似，区别是为了能成功接收数据，我们同样需要使用bind()绑定自身地址，但对于UDP这不是必须的；

应用层2：

#include <sys/stat.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/socket.h>
#include <string.h>
#include <asm/types.h>
#include <linux/netlink.h>
#include <linux/socket.h>
#include <errno.h>

#define NETLINK_TEST (25)
#define MAX_PAYLOAD (1024)
#define TEST_PID (100)

int netlink_create_socket(void) { //create a socket return socket(AF_NETLINK, SOCK_RAW, NETLINK_TEST); }
int netlink_bind(int sock_fd) { struct sockaddr_nl addr; memset(&addr, 0, sizeof(struct sockaddr_nl)); addr.nl_family = AF_NETLINK; addr.nl_pid = TEST_PID; addr.nl_groups = 0; return bind(sock_fd, (struct sockaddr *)&addr, sizeof(struct sockaddr_nl)); }
int
netlink_send_message(int sock_fd, const unsigned char *message, int len,
unsigned int pid, unsigned int group)
{
struct nlmsghdr *nlh = NULL;
struct sockaddr_nl dest_addr;

if( !message ) {
return -1;
}

//create message
nlh = (struct nlmsghdr *)malloc(NLMSG_SPACE(len));
if( !nlh ) {
perror("malloc");
return -2;
}
nlh->nlmsg_len = NLMSG_SPACE(len);
nlh->nlmsg_pid = TEST_PID;
nlh->nlmsg_flags = 0;
memcpy(NLMSG_DATA(nlh), message, len);

memset(&dest_addr, 0, sizeof(struct sockaddr_nl));
dest_addr.nl_family = AF_NETLINK;
dest_addr.nl_pid = pid;
dest_addr.nl_groups = group;

//send message
if( sendto(sock_fd, nlh, nlh->nlmsg_len, 0, (struct sockaddr *)&dest_addr, sizeof(struct sockaddr_nl)) != nlh->nlmsg_len ) {
printf("send error!\n");
free(nlh);
return -3;
}

free(nlh);
return 0;
}

int
netlink_recv_message(int sock_fd, unsigned char *message, int *len)
{
struct nlmsghdr *nlh = NULL;
struct sockaddr_nl src_addr;
socklen_t addrlen = sizeof(struct sockaddr_nl);

if( !message || !len ) {
return -1;
}

//create message
nlh = (struct nlmsghdr *)malloc(NLMSG_SPACE(MAX_PAYLOAD));
if( !nlh ) {
perror("malloc");
return -2;
}
memset(&src_addr, 0, sizeof(struct sockaddr_nl));
if( recvfrom(sock_fd, nlh, NLMSG_SPACE(MAX_PAYLOAD), 0, (struct sockaddr *)&src_addr, (socklen_t *)&addrlen) < 0 ) {
printf("recvmsg error!\n");
return -3;
}
*len = nlh->nlmsg_len - NLMSG_SPACE(0);
memcpy(message, (unsigned char *)NLMSG_DATA(nlh), *len);

free(nlh);
return 0;
}

int
main(int argc, char **argv)
{
int sock_fd;
char buf[MAX_PAYLOAD];
int len;

if( argc < 2) {
printf("enter message!\n");
exit(EXIT_FAILURE);
}

sock_fd = netlink_create_socket();
if(sock_fd == -1) {
printf("socket error!\n");
return -1;
}

if( netlink_bind(sock_fd) < 0 ) {
perror("bind");
close(sock_fd);
exit(EXIT_FAILURE);
}

netlink_send_message(sock_fd, argv[1], strlen(argv[1]) + 1, 0, 0);
if( netlink_recv_message(sock_fd, buf, &len) == 0 ) {
printf("recv:%s len:%d\n", buf, len);
}

close(sock_fd);
return 0;
}