Re: kvm guest for 486 sx

[Andre Przywara <andre.przywara@amd.com>]

Avi Kivity wrote:
> On 07/07/2009 04:15 PM, Jerry Geis wrote:
>>>> Notice FPU indicates yes and I requested -fpu on the command line.
>>>
>>> I think it detects the host fpu as a 487 coprocessor.  Anything in 
>>> dmesg about it?  Note there's no fpu in the flags: line.
>>>
>>> That's a kvm kernel bug.
>>>
>>>> also is this a 486 as it says AMD?
>>>
>>> You can override it if you like, -cpu 486,vendor=AuthenticAMD (I think).
>>>
>>
>> Sorry - I thought the " -cpu 486,-fpu " was the way to turn off the fpu.
>> I 'm trying to emulute a 486 sx that has no fpu.
> 
> I thought so too, but in fact it emulates a 486sx+487.
-fpu does not help here, as the injected 486 CPUID limits the level to 
0, so effectively no other leaf than leaf 0 is checked by the guest. 
That's why the FPU bit, which is 0000_0001.EDX:0, is not honored.
I bumped up the level to 1 and set the model to 2, which should mimic a 
later 486SX chip, together with -fpu Linux (2.6.21) still insisted on 
fpu: yes (probably determined by probing FPU instructions).
 >  We can make it
> emulate a 486sx without a 487, but that will take a kernel patch.
What are you thinking of? Forcing CR0.EM to 1 (and don't let the guest 
change it), then inject #NM into the guest?

Regards,
Andre.

-- 
Andre Przywara
AMD-Operating System Research Center (OSRC), Dresden, Germany
Tel: +49 351 448 3567 12
----to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Thomas M. McCoy; Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632