Re: [Qemu-devel] Bug with smp ppc guest.

[Andreas Färber]

Am 03.04.2014 20:26, schrieb Alexander Graf:
> 
> On 03.04.2014, at 19:00, Frederic Konrad <address@hidden> wrote:
> 
>> On 03/04/2014 17:29, Andreas Färber wrote:
>>> Hi Fred,
>>>
>>> Am 03.04.2014 17:19, schrieb Frederic Konrad:
>>>> I tried to boot a mpc85xx smp image with a new platform inside qemu.
>>>>
>>>> This command line reproduce the issue:
>>>> ./ppc-softmmu/qemu-system-ppc -M mpc8544ds -kernel zImage --smp 2
>>>>
>>>> This use to work but since this commit:
>>>> d197fdbc3b83655f3c145722805f0998c04dce16
>>>>
>>>>
>>>> target-ppc: Reset SPRs on CPU reset
>>>>
>>>>     
>>>>     This resets SPR values to defaults on CPU reset. This should help
>>>>
>>>>     with little-endian guests reboot issues.
>>>>
>>>> I get some errors inside the guest:
>>> Is this after today's fix?
>>>
>>> http://git.qemu-project.org/?p=qemu.git;a=commit;h=87d8354de3560eb705bcee83142d775f703177a6
>>>
>>> Regards,
>>> Andreas
>>>
>> Hi Andreas,
>>
>> I just tried seems it doesn't make any difference for me ;).
> 
> That patch should only have an effect when you're running with KVM which you 
> probably don't :).
> 
> I can see the same breakage here, but only when my guest kernel is e500v2. 
> With an e500mc kernel (even for -M mpc8544ds) it works fine. The secondary 
> core also does get woken up. So it's probably something related to SPE 
> emulation - hrm.

I think I found it:

hw/ppc/e500.c:
        env->spr[SPR_BOOKE_PIR] = cs->cpu_index = i;

With this patch this gets overwritten to spr->default_value, so the CPUs
end up having the same value on reset.

http://git.qemu-project.org/?p=qemu.git;a=blobdiff;f=target-ppc/translate_init.c;h=954dee3ec27b6a00559c878b8248ec6a07f35d79;hp=6084f40f28a62903e59db40771f33dd3e2c2996d;hb=d197fdbc3b83655f3c145722805f0998c04dce16;hpb=7aaf4957efad2bf6ac449d0caa9a1535f61006bc

Regards,
Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg