更简单一点的描述：

6.5 守护进程

6.5.1 基本概念
守护进程就是在后台执行, 没有控制终端或登录 Shell 的进程

6.5.2 实现方法
-〉关闭文件描述符: for (i=0; i < NOFILE; i++) close (i);
-〉甩开控制终端: 打开 /dev/tty, 然后用 ioctl 该文件上执行 TIOCNOTTY 命令
-〉脱离会话和进程组
===============================================================================
* 使用 setsid 设置为新会话的领头进程
* 因为 setsid 要求调用进程不是会话的领头进程时才有效.
* 因此, 首先可用 fork 派生, 并结束父进程.
* 之后调用 setsid, 子进程成为新会话的领头进程, 从而与原有会话, 进程组, tty 脱离.
* 但是, 为避免作为新会话领头进程的子进程在打开一个终端设备时, 拥有一个控制终端.
* 再次派生, 并结束父进程, 可让子进程成为非会话领头进程.

CODE

   if (fork ())
       exit (0);
   
   setsid ();

   if (fork ())
       exit (0);

-〉改变工作目录到根目录: chdir ("/")
-〉进程的 umask 设置为 0: umask (0)
-〉处理 SIGCHLD 信号, 为避免形成僵尸进程

from: http://263.aka.org.cn/Lectures/002/Lecture...1.6/daemon.html

今天看 agetty 的源程序，为了知道 setsid 是干什么用的，查到一大堆守护进程的东西 —— 还是不明白 setsid 是怎么回事

setsid
把一个进程设置为一个会话的过程组的组长 组长没有控制终端

这样当终端关闭时 进程就不会退出

很实用，谢了。

看到一篇英文的

from: http://www.ucdot.org/article.pl?sid=03/12/...219&mode=thread

Sometimes it is useful to start an application which is backgrounded and no longer associated with the starting terminal. Usually this is referred to as daemonizing.
Stevens in his book "Advanced programming in the unix environment" presents the classic daemonizing code. I have modified it here to create an application which when run from the command line, detaches from the command line and writes a message to syslog every 5 seconds.



#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#include <sys/types.h>
#include <syslog.h>


int main(int argc, char *argv[])
{
char c;
int _argc = 0;
char *_argv[3];
pid_t pid;

if ((pid = fork()) < 0) {
fprintf(stderr, "fork failed\n");
exit(1);
} else if (pid != 0) {
exit(0);
}

setsid();
chdir("/");
umask(0);
close(0);
close(1);
close(2);

for (;;) {
sleep(5);
syslog(LOG_INFO, "Still sleeping\n");
}
}

The trouble with the above is that it doesn't work under uClinux because there is no fork, only vfork. If we replace the fork call above with vfork, then the parent application never terminates. The child won't background until exec is called.

We can thus modify the classic daemonizing program to exec a program which can then be the background process while the parent process terminates. The trick is to exec the original program itself with an argument which indicates that it is the child process. The code below illustrates.

/*
* © Matthew Natalier, 2003
*
* If you want to use this, tell me that you love me :-)
*
*/
#include <getopt.h>
#include <stdio.h>
#include <stdlib.h>
#include <syslog.h>
#include <sys/types.h>
#include <unistd.h>


int execed = 0;

int main(int argc, char *argv[])
{
char c;
int _argc = 0;
char *_argv[3];
pid_t pid;

while ((c=getopt(argc, argv, "D")) > 0) {
switch© {
case 'D':
execed = 1;
break;
default:
fprintf(stderr, "You probably don't want to pass "
"options to this\n");
exit(1);
}
}

if (!execed) {
if ((pid = vfork()) < 0) {
fprintf(stderr, "vfork failed\n");
exit(1);
} else if (pid != 0) {
exit(0);
}

_argv[_argc++] = argv[0];
_argv[_argc++] = "-D";
_argv[_argc++] = NULL;
execv(_argv[0], _argv);
/* Not reached */
fprintf(stderr, "Couldn't exec\n");
_exit(1);

} else {
setsid();
chdir("/");
umask(0);
close(0);
close(1);
close(2);
}

for (;;) {
sleep(5);
syslog(LOG_INFO, "Still sleeping\n");
}
}

The best thing about the above code is that it will work on uClinux or standard Linux systems without modification.

资料很全 不错啊

sbsbsb