8.1-RELEASE - fatal trap 12 during install

[Andriy Gapon]

on 16/10/2010 16:40 Jacob Lambert said the following:
> On Saturday, October 16, 2010, Andriy Gapon <avg at icyb.net.ua> wrote:
>> [restored CC]
>>
>> on 16/10/2010 03:37 Jacob Lambert said the following:
>>> On Friday, October 15, 2010, Andriy Gapon <avg at icyb.net.ua> wrote:
>>>> on 15/10/2010 22:20 Jacob Lambert said the following:
>>>>> Hello everyone,
>>>>>
>>>>>  I've got a shiny new Supermicro H8QG6-F w/ 4x Opteron 6168 12 core
>>>>> CPUs and 128GB of DDR3 1333 ECC RAM onto which I've been trying to
>>>>
>>>> Do I read it correct that that's 48 logical CPUs total?
>>>>
>>> Yup, 48.  :)
>>
>> http://article.gmane.org/gmane.os.freebsd.current/128099
>>
>> In addition to that cpumask_t is only 32-bit even on amd64 and it's used all
>> over the place, so currently you won't have any luck with > 32 CPUs.
>> :-(
> Thanks for the link!
> BTW, I've got a box for testing if anybody wants to tackle >32 CPUs. :)

I've just followed up to that post with a patch.
It should allow you to boot without panics, although not more than 32 CPUs will
actually be used.
Could you please test the patch and post your results in _that_ thread?
Thanks!

>> You might want to try booting with ACPI enabled and only one socket populated.
> Already tried that and I get the same error as if it's fully populated.

That's probably because the LAPICs are still marked as enabled and we check
whether a CPU is enabled after UMA initialization.

-- 
Andriy Gapon