Re: Trying to get DMA working with IDE alim15x3 controller

From: Alan Cox <alan@lxorguk.ukuu.org.uk>
To: Art Haas <ahaas@airmail.net>
Cc: Bartlomiej Zolnierkiewicz <B.Zolnierkiewicz@elka.pw.edu.pl>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: Trying to get DMA working with IDE alim15x3 controller
Date: 16 Jul 2003 14:09:49 +0100	[thread overview]
Message-ID: <1058360981.6633.4.camel@dhcp22.swansea.linux.org.uk> (raw)
In-Reply-To: <20030715233202.GB17444@artsapartment.org>

On Mer, 2003-07-16 at 00:32, Art Haas wrote:
> I'd received mail from another person stating that the ALi DMA is broken
> from many old chips, but that it is only the UDMA stuff that is broken,

The older chips dont do UDMA. Then a lot of the next range of chips do
UDMA but not 48bit command so you can end up in PIO mode. Possibly our
logic for this needs tightening now we use lba28 commands whenever 
possible, so an LBA48 capable disk under LBA48 limit size is drivable
entirely in LBA28 safely.

> and the MW_DMA stuff works. Also the notes about the 2.6 transition
> indicate ali15x3 and DMA don't always play nicely together. Still,
> if the MW_DMA stuff works, some hdparm trickery like ...

Simplex DMA needs some fixing for one. That should only bite you if you
have a set up something like

hda -> unused
hdb -> unused
hdc -> DISK
hdd -> unused

in which case we assign the DMA engine to hda/hdb!