From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261575AbTLHTcy (ORCPT ); Mon, 8 Dec 2003 14:32:54 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261660AbTLHTcy (ORCPT ); Mon, 8 Dec 2003 14:32:54 -0500 Received: from sccrmhc13.comcast.net ([204.127.202.64]:24471 "EHLO sccrmhc13.comcast.net") by vger.kernel.org with ESMTP id S261575AbTLHTcv (ORCPT ); Mon, 8 Dec 2003 14:32:51 -0500 Message-ID: <3FD4D103.2080007@nucleodyne.com> Date: Mon, 08 Dec 2003 11:29:07 -0800 From: "NucleoDyne Systems Inc." Organization: www.NucleoDyne.com User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.0) Gecko/20020623 Debian/1.0.0-0.woody.1 X-Accept-Language: en MIME-Version: 1.0 To: linux-kernel@vger.kernel.org Subject: lost SCSI IO 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 have been developing a SCSI based application and facing hardship due to lack of better debugging support in SCSI domain on linux. May be I am not very familiar with linux scsi subsystem. I have a lost IO, sitting somewhere in some queue. The scsi logging facility has been turned on with : echo "scsi log all" > /proc/scsi/scsi. The syslog shows that the request has started: scsi_do_req (host = 0, channel = 0 target = 3, buffer =00000000, bufflen = 0, d)command : 00 20 00 00 00 00 Leaving scsi_do_req() The SCSI bus trace does not show any activity. I guess only way to find out the state of the IO is to put printf and recompile the kernel. HP-UX had facilities like lkcd, linux crash dump analyzer. I used to be called q4. A perl script could be written to navigate kernel data structures and extract information from them on a running system. If we had that kind of tool already into linux then debugging the live system would be easier. Is there a plan to include lkcd into default kernel?