From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261468AbTIXUdX (ORCPT ); Wed, 24 Sep 2003 16:33:23 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261515AbTIXUdW (ORCPT ); Wed, 24 Sep 2003 16:33:22 -0400 Received: from dhcp024-209-039-102.neo.rr.com ([24.209.39.102]:32655 "EHLO neo.rr.com") by vger.kernel.org with ESMTP id S261468AbTIXUdV (ORCPT ); Wed, 24 Sep 2003 16:33:21 -0400 Date: Wed, 24 Sep 2003 16:25:47 +0000 From: Adam Belay To: Alan Cox Cc: linux-kernel@vger.kernel.org Subject: Re: [PATCH] PnP Fixes for 2.6.0-test5 Message-ID: <20030924162547.GA25424@neo.rr.com> Mail-Followup-To: Adam Belay , Alan Cox , linux-kernel@vger.kernel.org References: <20030921200935.GB24897@neo.rr.com> <20030921201133.GE24897@neo.rr.com> <1064230334.8592.7.camel@dhcp23.swansea.linux.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1064230334.8592.7.camel@dhcp23.swansea.linux.org.uk> User-Agent: Mutt/1.4.1i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Sep 22, 2003 at 12:32:15PM +0100, Alan Cox wrote: > On Sul, 2003-09-21 at 21:11, Adam Belay wrote: > > # -------------------------------------------- > > # 03/09/21 ambx1@neo.rr.com 1.1357 > > # [PNP] remove DMA 0 restrictions > > # > > # The original argument for blocking DMA 0 was to avoid conflicts with > > # "memory refresh" but such configurations are only found on very old > > # 8-bit systems that are likely not supported by the linux kernel. > > DMA0 is used by lots of 386/486 era systems for memory refresh. It is > also "borrowed" by some other systems that know it isnt available to the > OS. There are a couple of heuristics I've seen suggested by vendors of > things like sound cards > > 1. Check the PnPBIOS information (never looked into this myself) Assuming the PnPBIOS provides this information through device nodes, the current pnp code will ensure that dma 0 is not used. > 2. Assume DMA 0 is free if the machine has a PCI bus detected > 3. Read the DMA 0 counter a few times. If it is continually changing > don't use DMA 0 > > #2 is certainly a good idea IMHO, I don't know how well the others work. Thanks, I'll look into these further. Regards, Adam