David S. Miller wrote:
From: Jens David <dg1kjd@afthd.tu-darmstadt.de>
Date: Thu, 28 Nov 2002 10:11:08 +0100
Now how do I make sure that between my last read() and the close()
system call no more data arrives? Sounds like a race condition to me.
The client and server must agree on what is to be sent/received.
Or do I have to make sure on higher protocol layers that this situation
does not occur?
That's right.
I disagree. The two directions of a TCP connection are independent. Data can flow in one direction reliably even if the other direction is completely ignored. If I don't want rx data I shouldn't have to read it just to keep the kernel from losing the tx data I give it. But... to work around the problem and avoid a race or an assumption about what the other side will do, you just have to make sure the tx queue is empty before you close. Is there a better way to do that than the "shutdown(SHUT_WR) then read() until EOF" method? Jim - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
