getsockopt() returns EINPROGRESS in non blocking connect()+select() flow

Question

Update:

My bad. The error I am getting is ECONNREFUSED and not EINPROGRESS. After checking the error variable I've found that it is greater than 0, I printfed errno instead of error. Of course errno is EINPROGRESS because it value didn't change since the call to connect().

Question answered. Thanks folks.

I am using the the the same piece of code as in Stevens' UNIX Network Programming non >blocking connect() example:

1. Setting socket to nonblocking
2. Initiate nonblocking connect()
3. Check for immediate completion
4. Call select() with timeout and wait for read or write readiness
5. When select() returns with value greater than 0 do getsockopt(socket, SOL_SOCKET, SO_ERROR, &error, &len).

The error I am getting is EINPROGRESS. The code is executed on rhel5 server.

Any ideas why I am getting this error?

Code snippet:

 flags = fcntl(sockfd, F_GETFL, 0);
 fcntl(sockfd, F_SETFL, flags | O_NONBLOCK);

 if ((retVal = connect(sockfd, saptr, salen)) < 0)
     if (errno != EINPROGRESS)
  return (-1);

 if (retVal == 0)
 {
     // restore file status flags
     fcntl(sockfd, F_SETFL, flags);  
     return 0;
 }

 FD_ZERO(&rset);
 FD_SET(sockfd, &rset);
 wset = rset;
 tval.tv_sec = nsec;
 tval.tv_usec = 0;

 if ((retVal = select(sockfd + 1, &rset, &wset, NULL, &tval)) == 0) 
 {
     // timeout
     close(sockfd);          
     errno = ETIMEDOUT;
     return (-1);
 }

 if (retVal < 0) 
 {
     // select() failed
     return (-1);
 }

 if (FD_ISSET(sockfd, &rset) || FD_ISSET(sockfd, &wset)) 
 {
     len = sizeof(error);
     error = 0;
     if (getsockopt(sockfd, SOL_SOCKET, SO_ERROR, &error, &len) < 0)
  return (-1);     

     if (error > 0) //<<<<< error == EINPROGRESS >>>
     { 
  close(sockfd);
  errno = error;
  return (-1);
     }
 } 
 else
 {
     return (-1);
 }

 // restore file status flags
 fcntl(sockfd, F_SETFL, flags);  

Answer 1

Shucks....I give up. I tried and tried but could not find a problem with your code. So, all I offer are some suggestions and hypothesis.

• Use FD_COPY to copy rset to wset.

• Can you do a getsockopt when the first connect fails. I am suspecting that Select is returning 0 and because of above, somehow your writefd set is set and doing a getsockopt is returning you the stale error from the previous connect. IMO, a subsquent connect should return EALREADY (although it could be platform dependant)

Please do the above (and downvote me if I am wrong about using FD_COPY) and share the results

Answer 2

EINPROGRESS indicates that the non-blocking connect() is still... in progress.

So I would ask - are you checking that after the select() returns, the connect()ing file descriptor is still set in the writefds fd_set? If it isn't, then that means select() has returned for some other reason.