原來我們實現(xiàn)connect()超時基本上都使用unix網(wǎng)絡(luò)編程一書的非阻塞方式（connect_nonb），今天在網(wǎng)上看到一篇文章，覺得很有意思，轉(zhuǎn)載如下：
讀Linux內(nèi)核源碼的時候偶然發(fā)現(xiàn)其connect的超時參數(shù)竟然和用SO_SNDTIMO操作的參數(shù)一致:
File: net/ipv4/af_inet.c
559 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
560
561 if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
562 /* Error code is set above */
563 if (!timeo || !inet_wait_for_connect(sk, timeo))
564 goto out;
565
566 err = sock_intr_errno(timeo);
567 if (signal_pending(current))
568 goto out;
569 }
這意味著：在Linux平臺下，可以通過在connect之前設(shè)置SO_SNDTIMO來達到控制連接超時的目的。簡單的寫了份測試代碼:
#include <stdlib.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <errno.h>
int main(int argc, char *argv[])
{
int fd;
struct sockaddr_in addr;
struct timeval timeo = {3, 0};
socklen_t len = sizeof(timeo);
fd = socket(AF_INET, SOCK_STREAM, 0);
if (argc == 4)
timeo.tv_sec = atoi(argv[3]);
setsockopt(fd, SOL_SOCKET, SO_SNDTIMEO, &timeo, len);
addr.sin_family = AF_INET;
addr.sin_addr.s_addr = inet_addr(argv[1]);
addr.sin_port = htons(atoi(argv[2]));
if (connect(fd, (struct sockaddr*)&addr, sizeof(addr)) == -1) {
if (errno == EINPROGRESS) {
fprintf(stderr, "timeout\n");
return -1;
}
perror("connect");
return 0;
}
printf("connected\n");
return 0;
}
