[coyotos-dev] Coyotos SMP?

[Valerio Bellizzomi]

On 22/10/2007, at 11.04, Jonathan S. Shapiro wrote:

>On Sun, 2007-10-21 at 02:35 +0200, Valerio Bellizzomi wrote:
>> This is the current status of my tests:
>> 
>> 1. Mainboards that trigger the cpu_ncpu assert:
>> 	- ASUS P4C800 mainboard with one P4 with HT enabled.
>> 	- GIGABYTE P4 TITAN mainboard with one P4 with HT enabled.
>> 
>> 2. Mainboards that bootstrap successfully:
>> 	- ABIT BP-6 mainboard with *two* old Celeron CPUs mounted on-board.
>> 	- Laptop ACER Travelmate with Pentium III.
>> 	- ECS K7S5A mainboard with Athlon.
>> 
>> So, the cpu_ncpu assert is triggered on P4 with HT, where the ACPI
>reports
>> 2 CPUs. It doesn't trigger on Dual Celeron, but I have not checked how
>> many CPUs are reported by ACPI on the BP-6 mainboard.
>
>I think that I know what is happening. Your newer motherboards [1] have
>ACPI bios, so Coyotos is detecting multiple CPUs. Your older
>motherboards [2] probably have the intel SMP extensions but not ACPI. We
>don't support the older SMP tables (yet), so the second CPU is never
>detected. Does this seem like a plausible explanation?

Yes and no. I have checked and found something that I had completely
forgot: the BP-6 mainboard has the option to enable/disable the "ACPI
function" in the BIOS setup.
In my previous tests it was *disabled*. I just tested the grubdisk on BP-6
with ACPI *enabled*, and *two* CPUs are detected, and it triggers the
cpu_ncpu assert.

val

>
>I have switched us to use IOAPIC by default, and I have removed the
>asserts from various places. That should help.
>-- 
>Jonathan S. Shapiro, Ph.D.
>Managing Director
>The EROS Group, LLC
>www.coyotos.org, www.eros-os.org