From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264031AbUDQUDm (ORCPT ); Sat, 17 Apr 2004 16:03:42 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264037AbUDQUDm (ORCPT ); Sat, 17 Apr 2004 16:03:42 -0400 Received: from picton.eecg.toronto.edu ([128.100.10.141]:39652 "EHLO picton.eecg.toronto.edu") by vger.kernel.org with ESMTP id S264031AbUDQUDl (ORCPT ); Sat, 17 Apr 2004 16:03:41 -0400 Message-ID: <40818D9A.9030002@acm.org> Date: Sat, 17 Apr 2004 16:03:38 -0400 From: Ashvin Goel User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040124 X-Accept-Language: en-us, en MIME-Version: 1.0 To: linux-kernel@vger.kernel.org Subject: Question regarding do_tcp_sendpages, tcp_current_mss and eff_sacks Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org I am trying to use do_tcp_sendpages() to stream some data from within the Linux Redhat kernel 2.4.20 (it is an smp kernel running on a uniprocessor). After a while this function returns an error value of -512. This error value makes no sense to me. A little debugging showed that tcp_current_mss() returns negative values. In particular, for the socket to which do_tcp_sendpages() is sending data, tp->eff_sacks has the value 200. This makes the value returned by tcp_current_mss be -156 (1448 - 4 - 200 * 8). Is this value reasonable? I think this value is causing do_tcp_sendpages() to return a large negative value. Thanks for any help/suggestions. Please CC your post to me personally. Thanks Ashvin Goel