5高并发服务器:socket IPC


  1. socket IPC

socket API原本是为网络通讯设计的,但后来在socket的框架上发展出一种IPC机制,就是UNIX Domain Socket。虽然网络socket也可用于同一台主机的进程间通讯(通过loopback地址127.0.0.1),但是UNIX Domain Socket用于IPC更有效率:不需要经过网络协议栈,不需要打包拆包、计算校验和、维护序号和应答等,只是将应用层数据从一个进程拷贝到另一个进程。这是因为,IPC机制本质上是可靠的通讯,而网络协议是为不可靠的通讯设计的。UNIX Domain Socket也提供面向流和面向数据包两种API接口,类似于TCPUDP,但是面向消息的UNIX Domain Socket也是可靠的,消息既不会丢失也不会顺序错乱。

UNIX Domain Socket是全双工的,API接口语义丰富,相比其它IPC机制有明显的优越性,目前已成为使用最广泛的IPC机制,比如X Window服务器和GUI程序之间就是通过UNIX Domain Socket通讯的。

使用UNIX DomainSocket的过程和网络socket十分相似,也要先调用socket()创建一个socket文件描述符,address family指定为AF_UNIXtype可以选择SOCK_DGRAMSOCK_STREAMprotocol参数仍然指定为0即可。

UNIX Domain Socket与网络socket编程最明显的不同在于地址格式不同,用结构体sockaddr_un表示,网络编程的socket地址是IP地址加端口号,而UNIX Domain Socket的地址是一个socket类型的文件在文件系统中的路径,这个socket文件由bind()调用创建,如果调用bind()时该文件已存在,则bind()错误返回。

   以下程序将UNIXDomain socket绑定到一个地址。

size =offsetof(struct sockaddr_un,sun_path) + strlen(un.sun_path);

offsetof查看第二个参数在前者内的偏移量

案例说明:

server.c

#include <stdlib.h>

#include <stdio.h>

#include <stddef.h>

#include <sys/socket.h>

#include <sys/un.h>

#include <sys/types.h>

#include <sys/stat.h>

#include <unistd.h>

#include <errno.h>

#include <ctype.h>

 

#define QLEN 10

/*

 * Create a server endpoint of a connection.

 * Returns fd if all OK, <0 on error.

 */

int serv_listen(const char *name)

{

   int fd, len, err, rval;

   struct sockaddr_un un;

   /* create a UNIX domain stream socket */

   if ((fd = socket(AF_UNIX, SOCK_STREAM, 0)) < 0)

       return(-1);

   unlink(name); /* in case it already exists */

   /* fill in socket address structure */

   memset(&un, 0, sizeof(un));

   un.sun_family = AF_UNIX;

   strcpy(un.sun_path, name);

   len = offsetof(struct sockaddr_un, sun_path) + strlen(name);

   /* bind the name to the descriptor */

   if (bind(fd, (struct sockaddr *)&un, len) < 0) {

       rval = -2;

       goto errout;

   }

   if (listen(fd, QLEN) < 0) { /* tell kernel we're a server */

       rval = -3;

       goto errout;

   }

   return(fd);

errout:

   err = errno;

   close(fd);

   errno = err;

   return(rval);

}

int serv_accept(int listenfd, uid_t *uidptr)

{

   int clifd, len, err, rval;

   time_t staletime;

   struct sockaddr_un un;

   struct stat statbuf;

   len = sizeof(un);

   if ((clifd = accept(listenfd, (struct sockaddr *)&un, &len)) < 0)

       return(-1); /* often errno=EINTR, if signal caught */

   /* obtain the client's uid from its calling address */

   len -= offsetof(struct sockaddr_un, sun_path); /* len of pathname */

   un.sun_path[len] = 0; /* null terminate */

   if (stat(un.sun_path, &statbuf) < 0) {

       rval = -2;

       goto errout;

   }

   if (S_ISSOCK(statbuf.st_mode) == 0) {

       rval = -3; /* not a socket */

       goto errout;

   }

   if (uidptr != NULL)

       *uidptr = statbuf.st_uid; /* return uid of caller */

   unlink(un.sun_path); /* we're done with pathname now */

   return(clifd);

errout:

   err = errno;

   close(clifd);

   errno = err;

   return(rval);

}

int main(void)

{

   int lfd, cfd, n, i;

   uid_t cuid;

   char buf[1024];

   lfd = serv_listen("foo.socket");

   if (lfd < 0) {

       switch (lfd) {

           case -3:perror("listen"); break;

           case -2:perror("bind"); break;

           case -1:perror("socket"); break;

       }

       exit(-1);

   }

   cfd = serv_accept(lfd, &cuid);

   if (cfd < 0) {

       switch (cfd) {

           case -3:perror("not a socket"); break;

           case -2:perror("a bad filename"); break;

           case -1:perror("accept"); break;

       }

       exit(-1);

   }

   while (1) {

r_again:

       n = read(cfd, buf, 1024);

       if (n == -1) {

           if (errno == EINTR)

               goto r_again;

       }

       else if (n == 0) {

           printf("the other side has been closed.\n");

           break;

       }

       for (i = 0; i < n; i++)

           buf[i] = toupper(buf[i]);

       write(cfd, buf, n);

   }

   close(cfd);

   close(lfd);

   return 0;

}

client.c

#include <stdio.h>

#include <stdlib.h>

#include <stddef.h>

#include <sys/stat.h>

#include <fcntl.h>

#include <unistd.h>

#include <sys/socket.h>

#include <sys/un.h>

#include <errno.h>

#define CLI_PATH "/var/tmp/" /* +5 for pid = 14 chars */

/*

 * Create a client endpoint and connect to a server.

 * Returns fd if all OK, <0 on error.

 */

int cli_conn(const char *name)

{

   int fd, len, err, rval;

   struct sockaddr_un un;

   /* create a UNIX domain stream socket */

   if ((fd = socket(AF_UNIX, SOCK_STREAM, 0)) < 0)

       return(-1);

   /* fill socket address structure with our address */

   memset(&un, 0, sizeof(un));

   un.sun_family = AF_UNIX;

   sprintf(un.sun_path, "%s%05d", CLI_PATH, getpid());

   len = offsetof(struct sockaddr_un, sun_path) + strlen(un.sun_path);

   unlink(un.sun_path); /* in case it already exists */

   if (bind(fd, (struct sockaddr *)&un, len) < 0) {

       rval = -2;

       goto errout;

   }

   /* fill socket address structure with server's address */

   memset(&un, 0, sizeof(un));

   un.sun_family = AF_UNIX;

   strcpy(un.sun_path, name);

   len = offsetof(struct sockaddr_un, sun_path) + strlen(name);

   if (connect(fd, (struct sockaddr *)&un, len) < 0) {

       rval = -4;

       goto errout;

   }

   return(fd);

errout:

   err = errno;

   close(fd);

   errno = err;

   return(rval);

}

int main(void)

{

   int fd, n;

   char buf[1024];

   fd = cli_conn("foo.socket");

   if (fd < 0) {

       switch (fd) {

           case -4:perror("connect"); break;

           case -3:perror("listen"); break;

           case -2:perror("bind"); break;

           case -1:perror("socket"); break;

       }

       exit(-1);

   }

   while (fgets(buf, sizeof(buf), stdin) != NULL) {

       write(fd, buf, strlen(buf));

       n = read(fd, buf, sizeof(buf));

       write(STDOUT_FILENO, buf, n);

   }

   close(fd);

   return 0;

}

 

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

涂作权的博客

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付:¥1
获取中
扫码支付

您的余额不足,请更换扫码支付或充值

打赏作者

实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值