[apue] 作为 daemon 启动, Unix Domain Socket 侦听失败?

前段时间写一个传递文件句柄的小 demo，有 server 端、有 client 端，之间通过 Unix Domain Socket 通讯。

在普通模式下，双方可以正常建立连接，当server端作为daemon启动时，则第一次启动成功，之后再启动， listen 会连接报 ENOTSUPP 错误，导致启动失败。

spipe.c

int cli_conn(const char *name)
{
    int fd, len, err, rval; 
    struct sockaddr_un un; 

    if ((fd = socket (AF_UNIX, SOCK_STREAM, 0)) < 0) { 
        printf ("create socket failed\n"); 
        return -1; 
    }

    printf ("create socket ok\n"); 
    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) {
        err = errno; 
        printf ("connect failed\n"); 
        rval = -4; 
        goto errout; 
    }

    printf ("connect to server ok\n"); 
    return fd;
errout:
    close (fd); 
    errno = err; 
    return rval;
}


int serv_listen (const char *name)
{
    int fd, len, err, rval; 
    struct sockaddr_un un; 

    if ((fd = socket (AF_UNIX, SOCK_STREAM, 0)) < 0) {
        printf ("socket failed\n"); 
        return -1; 
    }

    printf ("create socket ok\n"); 
    unlink (name); 
    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 (bind (fd, (struct sockaddr *)&un, len) < 0) {
        err = errno; 
        printf ("bind failed\n");  
        rval = -2; 
        goto errout; 
    }

    printf ("bind socket to path ok\n"); 
    if (listen (fd, QLEN) < 0) { 
        err = errno; 
        printf ("listen failed, errno %d\n", errno); 
        rval = -3; 
        goto errout; 
    }

    printf ("start listen on socket ok\n"); 
    return fd; 
errout:
    close (fd); 
    errno = err; 
    return rval; 
}

int serv_accept (int listenfd, uid_t *uidptr)
{
    int clifd, err, rval; 
    time_t staletime; 
    struct sockaddr_un un; 
    struct stat statbuf; 

    size_t len = sizeof (un); 
    if ((clifd = accept (listenfd, (struct sockaddr *)&un, &len)) < 0) { 
        printf ("accept failed\n"); 
        return -1; 
    }

    len -= offsetof (struct sockaddr_un, sun_path); 
    un.sun_path[len] = 0; 
    printf ("accept %s ok\n", un.sun_path); 

    unlink (un.sun_path); 
    return clifd; 

errout:
    close (clifd); 
    errno = err; 
    return rval; 
}

 

出错的位置在 serv_listen (line 57) 处，出错时的 server 端输出为：

Jan 17 00:24:44 localhost opend: create socket ok
Jan 17 00:24:44 localhost opend: bind socket to path ok
Jan 17 00:24:44 localhost opend: listen failed, errno 95
Jan 17 00:24:44 localhost opend: serv_listen error: Operation not supported

 

errno 95 为 ENOTSUPP。不以 daemon 运行时正常的输出如下：

create socket ok
bind socket to path ok
start listen on socket ok
accept  ok
new connection: uid 0, fd 4

 

可能细心的读者会觉得，以 daemon 方式运行 printf 怎么还可以输出呢，是有以下宏定义做了处理：

#ifdef USE_APUE
#include "../apue.h"
#define printf log_msg
#endif

 

以 daemon 运行时会定义 USE_APUE 宏，从而将 printf 重定义为 log_msg 输出到 syslog。

下面是 server 端的代码：

csopend2.c

int main (int argc, char *argv[])
{
    int c = 0; 
    log_open ("open.serv", LOG_PID, LOG_USER); 

    opterr = 0; // don't want getopt() writting to stderr !
    while ((c = getopt (argc, argv, "d")) != EOF) {
        switch (c) { 
            case 'd':
                debug = log_to_stderr = 1; 
                break; 
            case '?':
                err_quit ("unrecongnized option: -%c", optopt); 
        }
    }

    if (debug == 0)
    {
        log_to_stderr = 0; 
        daemonize ("opend"); 
    }

    loop (); 
    return 0; 
}

 

不使用 -d 时表示 daemon 运行(与常识相反?)，上面标黄的代码就是。

对应的 client 端代码：

csopenc.c

 

一开始怀疑是用于 listen 的本地 socket 文件已经存在，于是去 /tmp 目录看了下，果然有 opend 这个文件，删除之，再运行，不行；

然后怀疑是没有复用端口(?)所致，于是在 listen 之前添加了以下代码段：

    int opt = 1; 
    if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (const void *)&opt, sizeof(opt)) < 0) {
        err = errno; 
        printf ("setsockopt failed\n"); 
        rval = -3; 
        goto errout; 
    }

 

设置端口复用。编译、运行，输出如下：

Jan 17 00:43:11 localhost opend: create socket ok
Jan 17 00:43:11 localhost opend: bind socket to path ok
Jan 17 00:43:11 localhost opend: set socket option ok
Jan 17 00:43:11 localhost opend: listen failed, errno 95
Jan 17 00:43:11 localhost opend: serv_listen error: Operation not supported

 

设置成功了，但还是不行

难道 daemon 与普通进程使用 Unix 域套接字还有什么区别么?

暂时存疑……