[linux-elitists] Dual bore - a question on diagnosis

[Phil Mayers]

Dave Jones wrote:
> On Fri, Dec 22, 2006 at 02:21:30PM +0000, Phil Mayers wrote:
> 
>  > A combination of arguments seems to help things along:
>  > 
>  > mem=512m notsc ide0=noprobe ide1=noprobe
> 
> I'm surprised it worked at all with these, as the only one I've
> been able to track down breaks horribly and doesn't boot unless
> you pass acpi=off.  It seems to have some issues parsing the PCI bus
> topology.  AFAIK people at HP are poking at the acpi code hoping
> to get this resolved.
> 
> This bit 2.6.18 and 2.6.20rc1 kernels, so it should affect
> all the Fedora/RHEL5-beta releases, so I'm quite puzzled how
> it wasn't needed for you.

Ah, I think I ran into that this morning - my machine has an nvidia 
quadro (using the free drivers before anyone gets excited) in the PCI 
Express slot. Removing that means it failed to boot completely, though I 
only discovered that a few hours ago.

Presumably with the nv card in there, the bus topology is changed 
significantly enough that it boots (but does then suffer performance 
problems, unless you pass the right kernel arguments). I had 
hypothesised that the incorrect MTRR setup was related to the intel 
graphics, but it was just a hunch.

There's no sign of the intel video in the lscpi output at the moment 
with it booted on the nv card, so I guess maybe the hardware gets 
disabled if a plug-in gfx adapter is present.

Sounds like HP are working on it - I look forward to them delivering us 
an xmas present.

Cheers, and thanks for taking the time to look at it,

Happy holiday all,
Phil