From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S262126AbTKWQEx (ORCPT ); Sun, 23 Nov 2003 11:04:53 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262192AbTKWQEx (ORCPT ); Sun, 23 Nov 2003 11:04:53 -0500 Received: from mail3.ithnet.com ([217.64.64.7]:50561 "HELO heather-ng.ithnet.com") by vger.kernel.org with SMTP id S262126AbTKWQEw (ORCPT ); Sun, 23 Nov 2003 11:04:52 -0500 X-Sender-Authentication: net64 Date: Sun, 23 Nov 2003 17:04:46 +0100 From: Stephan von Krawczynski To: linux-kernel Subject: Problem Report: 2.4.23-rc2, ide, dma settings Message-Id: <20031123170446.7c70439e.skraw@ithnet.com> Organization: ith Kommunikationstechnik GmbH X-Mailer: Sylpheed version 0.9.7 (GTK+ 1.2.10; i686-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hello all, I just found out that either I did not understand what the "keepsettings" flag in hdparm is meant to be, or it does not work as expected. Situation is as follows: Trying to record DVDs with a SONY DRU-500A connected to 00:0f.1 IDE interface: ServerWorks CSB5 IDE Controller (rev 93) I do "hdparm -k1 -d1 /dev/hdc" and see: /dev/hdc: setting using_dma to 1 (on) setting keep_settings to 1 (on) using_dma = 1 (on) keepsettings = 1 (on) Then I try recording, but something bad happens: Nov 23 16:36:50 box kernel: scsi : aborting command due to timeout : pid 5977459, scsi3, channel 0, id 0, lun 0 UNKNOWN(0x54) 01 00 00 00 00 00 00 00 00 Nov 23 16:36:50 box kernel: hdc: irq timeout: status=0xd0 { Busy } Nov 23 16:36:50 box kernel: hdc: ATAPI reset complete And after that: # hdparm /dev/hdc /dev/hdc: HDIO_GET_MULTCOUNT failed: Invalid argument IO_support = 0 (default 16-bit) unmaskirq = 0 (off) using_dma = 0 (off) keepsettings = 1 (on) readonly = 0 (off) BLKRAGET failed: Invalid argument HDIO_GETGEO failed: Invalid argument DMA is off. Is this expected? Recording btw _continues_, only without DMA. This happens during high load situations... Regards, Stephan