Re: [PATCH] pata_amd: remove contamination added during cable_detect conversion

[Tejun Heo <htejun@gmail.com>]

Hello,

Alan Cox wrote:
> On Mon, 23 Apr 2007 02:04:38 +0900
> Tejun Heo <htejun@gmail.com> wrote:
> 
>> This is added by added by cff63dfceb52c564fe1ba5394d50ab7d599a11b9
>>  - pata: cable methods.
> 
> The Cable method fix for the Nvidia hardware if ACPI drivers are present
> and usable is to use the ACPI driver, so its not a "contamination" its a
> fix for the fact the Nvidia code can't do cable detect any other way

But pata_amd modifications should be merged together with pata_acpi not
separately.

Anyways, I've been testing pata_acpi and it doesn't work on my nforce
ultra machine.  Whether the cable is 40c or not, _STM happily sets UDMA
mode above 2 resulting in a series of transfer errors.  EH kicks in
eventually and drops transfer speed.

Also, ata_acpi_gtm() implementation was broken in that it skipped one
level of indirection (acpi_object, please take a look at the last patch
of just posted patchset), so it was basically returning garbage values.

I have patches to update pata_amd to use _STM/_GTM for cable detection
(a lot are taken from pata_acpi) but ACPI cable detection is just as
broken as PCI config one.

Thanks.

-- 
tejun