2001-04-30 13:49:15

by Charl P. Botha

[permalink] [raw]
Subject: pci/quirks.c - VIA PCI latency in 2.4.4

You're right, this is a problem, your solution is not entirely correct
though (the south bridge has to be checked, but the patch is to the config
registers of the pci-host bridge). Please see my patch posted on this list
with subject "Re: 2.4.4 Sound corruption [PATCH] NEW, ignore previous
patch".

Regards,

--
charl p. botha | computer graphics and cad/cam
http://cpbotha.net/ | http://www.cg.its.tudelft.nl/


2001-04-30 13:59:07

by Charl P. Botha

[permalink] [raw]
Subject: Re: pci/quirks.c - VIA PCI latency in 2.4.4

I should probably clarify. The fix is *only* valid for the VIA VT8363 host to
pci bridge, therefore the test for the 686B south bridge only gets done if a
8363 is found.

No fix is known for intel pci-to-host bridges with the 686B south bridge,
and in the case of the AMD-761 chipset, there are certain BIOS settings you
can change. See: http://home.tiscalinet.de/au-ja/review-kt133a-4-en.html

On Mon, Apr 30, 2001 at 03:48:52PM +0200, Charl P. Botha wrote:
> You're right, this is a problem, your solution is not entirely correct
> though (the south bridge has to be checked, but the patch is to the config
> registers of the pci-host bridge). Please see my patch posted on this list
> with subject "Re: 2.4.4 Sound corruption [PATCH] NEW, ignore previous
> patch".

--
charl p. botha | computer graphics and cad/cam
http://cpbotha.net/ | http://www.cg.its.tudelft.nl/

2001-04-30 14:33:40

by Alan

[permalink] [raw]
Subject: Re: pci/quirks.c - VIA PCI latency in 2.4.4

> I should probably clarify. The fix is *only* valid for the VIA VT8363 host to
> pci bridge, therefore the test for the 686B south bridge only gets done if a
> 8363 is found.
>
> No fix is known for intel pci-to-host bridges with the 686B south bridge,
> and in the case of the AMD-761 chipset, there are certain BIOS settings you
> can change. See: http://home.tiscalinet.de/au-ja/review-kt133a-4-en.html

The -ac tree has the ability to kill IDE DMA across the entire system. It may
be this is what should be done with all hybrid setups where there is no known
fix.

2001-04-30 15:39:31

by Charl P. Botha

[permalink] [raw]
Subject: Re: pci/quirks.c - VIA PCI latency in 2.4.4

On Mon, Apr 30, 2001 at 03:33:49PM +0100, Alan Cox wrote:
> > No fix is known for intel pci-to-host bridges with the 686B south bridge,
> > and in the case of the AMD-761 chipset, there are certain BIOS settings you
> > can change. See: http://home.tiscalinet.de/au-ja/review-kt133a-4-en.html
>
> The -ac tree has the ability to kill IDE DMA across the entire system. It may
> be this is what should be done with all hybrid setups where there is no known
> fix.

Alternatively, too much time shouldn't be spent on this specific bug. VIA
has recognised it, and motherboard vendors are already starting to release
updated BIOSes that sport the required settings (e.g. Abit, see
http://www.viahardware.com/faq/kt7/faqbios.html). I'm just afraid if the
kernel starts disabling IDE DMA on certain of these setups, there'll have to
be checks (in the kernel) for possible future updated BIOSes as well (else
we disable IDE DMA where it's not necessary).

Selective application of the fix (as in my patch) does no harm on the known
combinations though.

My very humble 2c,

--
charl p. botha | computer graphics and cad/cam
http://cpbotha.net/ | http://www.cg.its.tudelft.nl/