Re: [Qemu-devel] [PATCH] target/ppc/cpu-models: set POWER9_v1.0 as POWER9 DD1

[Thomas Huth <thuth@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2079 bytes --]

On 23.06.2017 11:21, David Gibson wrote:
> On Thu, Jun 22, 2017 at 01:31:24PM +0200, Thomas Huth wrote:
>> On 22.06.2017 13:26, Laurent Vivier wrote:
>>> CPU_POWERPC_POWER9_DD1 is 0x004E0100, so this is the POWER9 v1.0.
>>>
>>> When we run qemu on a POWER9 DD1 host, we must use either
>>> "-cpu host" or "-cpu POWER9", but in the latter case it fails with
>>>
>>>     Unable to find sPAPR CPU Core definition
>>>
>>> because POWER9 DD1 doesn't appear in the list of known CPUs.
>>>
>>> This patch fixes this by defining POWER9_v1.0 with POWER9 DD1
>>> PVR instead of CPU_POWERPC_POWER9_BASE.
>>>
>>> Signed-off-by: Laurent Vivier <lvivier@redhat.com>
>>> ---
>>>  target/ppc/cpu-models.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/target/ppc/cpu-models.c b/target/ppc/cpu-models.c
>>> index 4d3e635..a22363c 100644
>>> --- a/target/ppc/cpu-models.c
>>> +++ b/target/ppc/cpu-models.c
>>> @@ -1144,7 +1144,7 @@
>>>      POWERPC_DEF("970_v2.2",      CPU_POWERPC_970_v22,                970,
>>>                  "PowerPC 970 v2.2")
>>>  
>>> -    POWERPC_DEF("POWER9_v1.0",   CPU_POWERPC_POWER9_BASE,            POWER9,
>>> +    POWERPC_DEF("POWER9_v1.0",   CPU_POWERPC_POWER9_DD1,             POWER9,
>>>                  "POWER9 v1.0")
>>>  
>>>      POWERPC_DEF("970fx_v1.0",    CPU_POWERPC_970FX_v10,              970,
>>>
>>
>> I think this also makes sense for running in TCG mode to get a valid
>> real PVR there.
> 
> I'm not so convinced.
> 
> IIUC, this will make TCG default (for now) to a DD1 POWER9.  That's a)
> probably not what anyone wants - who'd select a buggy prototype and b)
> not accurate - TCG does not implement DD1's bugs.

But DD1 = v1.0. There is no real chip which is using
CPU_POWERPC_POWER9_BASE as PVR ... so using that is also not accurate,
and also likely not what users expect when the select "POWER9_v1.0", and
it just does not work as soon as KVM is enabled. So I think Laurent's
patch is the right way to go. Or do you have a better suggestion?

 Thomas


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]