Re: Hard lockup with 4.0 RC5

[Pierre Pronchery <khorben%defora.org@localhost>]

                        Hey,

On 2007-12-04, Pierre Pronchery <khorben%defora.org@localhost> wrote:
> John Nemeth wrote:
>> On Apr 25,  6:23pm, Pierre Pronchery wrote:
>> } 
>> } I have here a machine, running 3.1_STABLE so far, that I just tried to
>> } run with a 4.0_RC5 kernel (compiled here). It locks up after "Kernelized
>> } RAIDframe activated". The incriminated dmesg is "dmesg-4.0_RC5.txt",
>> } attached to this mail.
>> } [...]
>> } The machine is based on an Asus CUV4X-D:
>> } http://usa.asus.com/100/html/share/3/txt/60/pdf/cuv4x-d/cuv4x-d.zip
>> } It is SMP-capable, and I have two CPUs, but for some reason it only
>> } works with one (BIOS never saw more than one). Maybe the board is b0rk3d...
>> 
>>     Or, perhaps the BIOS is "b0rk3d".  Have you checked for a BIOS
>> update?  It is fairly rare that Asus doesn't post updates for a
>> motherboard.  Taking a look, I found three BIOS updates posted for your
>> motherboard.
>
> Sure, but I checked already and the BIOS version is the latest (1016). I
> could try to re-flash it, but I'm not sure if it can really help...

I just spent some more time on it:
- booting with "disable acpi" works
- -current hangs with ACPI at "cpu0 at mainbus0 apid3"
- -current shows the same issues as _RC5, namely PCI cards can't map
  device registers (i/o space accesses are off)

Given this, jakllsch (from IRC) pointed out this message:
http://mail-index.netbsd.org/tech-kern/2002/11/19/0008.html

I disabled the BIOS "PNP OS" option and the system then works normally
without ACPI (still hangs at the same point).

Interestingly, Linux works by default on the same hardware, regardless
of the BIOS settings (vanilla 2.4.36.2 as Debian etch 2.6.18-6-686).
Should it be the case on NetBSD as well?

(ie, should this be fixed, or better PR'd as "wontfix"?)

HTH,
-- 
khorben