possibly a dumb question about sb600/700

possibly a dumb question about sb600/700

[Tejun Heo]

Hello,

I've been looking at PATA support for sb600 and 700 and found something
weird.

* IDE atiixp.c has a separate entry for sb600 such that it only probes
the first port but sb700 doesn't use the entry.  So, does sb600 has one
PATA channel but sb700 has two?

* libata pata_atiixp.c doesn't have special handling for sb600's
single-channeldness.  Is this okay?

Thanks.

-- 
tejun

Re: possibly a dumb question about sb600/700

[Alan Cox]

On Mon, 12 Nov 2007 18:05:44 +0900
Tejun Heo <htejun@gmail.com> wrote:

> Hello,
> 
> I've been looking at PATA support for sb600 and 700 and found something
> weird.
> 
> * IDE atiixp.c has a separate entry for sb600 such that it only probes
> the first port but sb700 doesn't use the entry.  So, does sb600 has one
> PATA channel but sb700 has two?
> 
> * libata pata_atiixp.c doesn't have special handling for sb600's
> single-channeldness.  Is this okay?

Should be just fine as I understand it all anyway - the second channel
will have no resources assigned if absent which is handled by libata core
already.

Alan

Re: possibly a dumb question about sb600/700

[Tejun Heo]

Alan Cox wrote:
> On Mon, 12 Nov 2007 18:05:44 +0900
> Tejun Heo <htejun@gmail.com> wrote:
> 
>> Hello,
>>
>> I've been looking at PATA support for sb600 and 700 and found something
>> weird.
>>
>> * IDE atiixp.c has a separate entry for sb600 such that it only probes
>> the first port but sb700 doesn't use the entry.  So, does sb600 has one
>> PATA channel but sb700 has two?
>>
>> * libata pata_atiixp.c doesn't have special handling for sb600's
>> single-channeldness.  Is this okay?
> 
> Should be just fine as I understand it all anyway - the second channel
> will have no resources assigned if absent which is handled by libata core
> already.

Alright, then.  I was just worried about the asymmetry.  Oh, this
_reminds me of another problem regarding enable bits.  There's a system
(wyse thin client) with pata_amd controller where the enable bit isn't
set by the BIOS and there's no reliable way to identify the system (no
DMI).  Do you know how important is the enabled bits test for these
controllers?

-- 
tejun

Re: possibly a dumb question about sb600/700

[Alan Cox]

> Alright, then.  I was just worried about the asymmetry.  Oh, this
> _reminds me of another problem regarding enable bits.  There's a system
> (wyse thin client) with pata_amd controller where the enable bit isn't
> set by the BIOS and there's no reliable way to identify the system (no
> DMI).  Do you know how important is the enabled bits test for these
> controllers?

If we ignore the tests we get more problems with 30 second timeouts on
other hardware that has incorrect termination or a port which is simply
not wired on the mainboard.

I bet we can identify the Wyse somehow and I'd rather do that even if it
means looking for magic strings in the BIOS ROM image.

Alan

Re: possibly a dumb question about sb600/700

[Tejun Heo]

Alan Cox wrote:
>> Alright, then.  I was just worried about the asymmetry.  Oh, this
>> _reminds me of another problem regarding enable bits.  There's a system
>> (wyse thin client) with pata_amd controller where the enable bit isn't
>> set by the BIOS and there's no reliable way to identify the system (no
>> DMI).  Do you know how important is the enabled bits test for these
>> controllers?
> 
> If we ignore the tests we get more problems with 30 second timeouts on
> other hardware that has incorrect termination or a port which is simply
> not wired on the mainboard.
> 
> I bet we can identify the Wyse somehow and I'd rather do that even if it
> means looking for magic strings in the BIOS ROM image.

Thanks.  I don't have direct access to the machine so I guess they'll
have to live with temporary workaround for the time being.  :-(

-- 
tejun

Re: possibly a dumb question about sb600/700

[Alan Cox]

On Mon, 12 Nov 2007 23:44:30 +0900
Tejun Heo <htejun@gmail.com> wrote:

> Alan Cox wrote:
> >> Alright, then.  I was just worried about the asymmetry.  Oh, this
> >> _reminds me of another problem regarding enable bits.  There's a system
> >> (wyse thin client) with pata_amd controller where the enable bit isn't
> >> set by the BIOS and there's no reliable way to identify the system (no
> >> DMI).  Do you know how important is the enabled bits test for these
> >> controllers?
> > 
> > If we ignore the tests we get more problems with 30 second timeouts on
> > other hardware that has incorrect termination or a port which is simply
> > not wired on the mainboard.
> > 
> > I bet we can identify the Wyse somehow and I'd rather do that even if it
> > means looking for magic strings in the BIOS ROM image.
> 
> Thanks.  I don't have direct access to the machine so I guess they'll
> have to live with temporary workaround for the time being.  :-(

cp biosrom file
strings file
grep product name

not approximate offset