From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261291AbTIOMGE (ORCPT ); Mon, 15 Sep 2003 08:06:04 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261300AbTIOMGE (ORCPT ); Mon, 15 Sep 2003 08:06:04 -0400 Received: from catv-d5deb78c.bp13catv.broadband.hu ([213.222.183.140]:9606 "EHLO dap.index") by vger.kernel.org with ESMTP id S261291AbTIOMGC (ORCPT ); Mon, 15 Sep 2003 08:06:02 -0400 Subject: reproducable lockup on changing DMA state (2.4.22, cmd649) From: Pallai Roland To: linux-kernel@vger.kernel.org Content-Type: text/plain Content-Transfer-Encoding: 7bit Organization: Message-Id: <1063627245.493.97.camel@localhost> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.2.3 Date: 15 Sep 2003 14:00:45 +0200 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org changing DMA state (hdparm -d0/1) always lockups my system (2.4.21/2.4.22) if another drive does heavy I/O on same channel in DMA mode.. there's no problem if the another drive's in PIO mode or there's no I/O.. the biggest problem, that the kernel itself tries to change DMA state if DMA timeout occurs and it's lockups the system too.. I've a machine with 28 IDE disks on plain CMD649 controllers, and lots of lockups happens when a disk going to die soon.. :( # kernel 2.4.22 # cmd649 ctrl # maxtor, samsung, quantum disks, nevermind -- DaP