Re: [Qemu-devel] [PATCH v2] target-ppc: move POWER7+ to a separate family

[Alexey Kardashevskiy]

On 11/09/2013 03:59 AM, Andreas Färber wrote:
> Am 08.11.2013 15:54, schrieb Alexey Kardashevskiy:
>> On 11/09/2013 12:44 AM, Andreas Färber wrote:
>>> Am 08.11.2013 03:37, schrieb Alexey Kardashevskiy:
>>>> So far POWER7+ was a part of POWER7 family. However it has a different
>>>> PVR base value so in order to support PVR masks, it needs a separate
>>>> family class.
>>>>
>>>
>>> Alexey,
>>>
>>>> Another reason to make a POWER7+ family is that its name in the device
>>>> tree (/proc/device-tree/cpus/cpu*) should be "Power7+" but not "Power7"
>>>> and this cannot be easily fixed without a new family class.
>>>>
>>>> This adds a new family class, PVR base and mask values and moves
>>>> Power7+ v2.1 CPU to a new family. The class init function is copied
>>>> from the POWER7 family.
>>>>
>>>> Signed-off-by: Alexey Kardashevskiy <address@hidden>
>>>> ---
>>>> Changes:
>>>> v2:
>>>> * added VSX enable bit
>>>> ---
>>>>  target-ppc/cpu-models.c     |  2 +-
>>>>  target-ppc/cpu-models.h     |  2 ++
>>>>  target-ppc/translate_init.c | 38 ++++++++++++++++++++++++++++++++++++++
>>>>  3 files changed, 41 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/target-ppc/cpu-models.c b/target-ppc/cpu-models.c
>>>> index 04d88c5..7c9466f 100644
>>>> --- a/target-ppc/cpu-models.c
>>>> +++ b/target-ppc/cpu-models.c
>>>> @@ -1140,7 +1140,7 @@
>>>>                  "POWER7 v2.1")
>>>>      POWERPC_DEF("POWER7_v2.3",   CPU_POWERPC_POWER7_v23,             
>>>> POWER7,
>>>>                  "POWER7 v2.3")
>>>> -    POWERPC_DEF("POWER7+_v2.1",  CPU_POWERPC_POWER7P_v21,            
>>>> POWER7,
>>>> +    POWERPC_DEF("POWER7+_v2.1",  CPU_POWERPC_POWER7P_v21,            
>>>> POWER7P,
>>>>                  "POWER7+ v2.1")
>>>>      POWERPC_DEF("POWER8_v1.0",   CPU_POWERPC_POWER8_v10,             
>>>> POWER8,
>>>>                  "POWER8 v1.0")
>>>> diff --git a/target-ppc/cpu-models.h b/target-ppc/cpu-models.h
>>>> index 731ec4a..49ba4a4 100644
>>>> --- a/target-ppc/cpu-models.h
>>>> +++ b/target-ppc/cpu-models.h
>>>> @@ -558,6 +558,8 @@ enum {
>>>>      CPU_POWERPC_POWER7_v20         = 0x003F0200,
>>>>      CPU_POWERPC_POWER7_v21         = 0x003F0201,
>>>>      CPU_POWERPC_POWER7_v23         = 0x003F0203,
>>>> +    CPU_POWERPC_POWER7P_BASE       = 0x004A0000,
>>>> +    CPU_POWERPC_POWER7P_MASK       = 0xFFFF0000,
>>>>      CPU_POWERPC_POWER7P_v21        = 0x004A0201,
>>>>      CPU_POWERPC_POWER8_BASE        = 0x004B0000,
>>>>      CPU_POWERPC_POWER8_MASK        = 0xFFFF0000,
>>>> diff --git a/target-ppc/translate_init.c b/target-ppc/translate_init.c
>>>> index 35d1389..c030a20 100644
>>>> --- a/target-ppc/translate_init.c
>>>> +++ b/target-ppc/translate_init.c
>>>> @@ -7253,6 +7253,44 @@ POWERPC_FAMILY(POWER7)(ObjectClass *oc, void *data)
>>>>      pcc->l1_icache_size = 0x8000;
>>>>  }
>>>>  
>>>> +POWERPC_FAMILY(POWER7P)(ObjectClass *oc, void *data)
>>>> +{
>>>> +    DeviceClass *dc = DEVICE_CLASS(oc);
>>>> +    PowerPCCPUClass *pcc = POWERPC_CPU_CLASS(oc);
>>>> +
>>>> +    dc->fw_name = "PowerPC,POWER7+";
>>>
>>> Apart from the commit message differing from the code...
>>
>>
>> In what part?
> 
> The spelling of POWER7. You write it should be "Power7+" but implement
> it as upper-case "POWER7+" (ignoring the "PowerPC," prefix, that is).


Ah. Sorry.


>>> We've had this discussion before: Jacques reported that on his POWER7+
>>> box only "POWER7" is shown, not "POWER7+", equivalent to my POWER5+ box
>>> showing only "PowerPC,POWER5". Compare my commit, which documents this:
>>>
>>> http://git.qemu.org/?p=qemu.git;a=commit;h=793826cd460828975591f289de78672af4a47ef9
>>>
>>> So, adding a POWER7P family seems correct to me, just the fw_name seems
>>> wrong - or you'll need to investigate further why there are conflicting
>>> reports of how it is shown. Possibly based on revision or pHyp vs. SLOF?
>>
>>
>> Yes we have had this discussion. Paul said it should "POWER7+". The only
>> P7+ machine I have handy shows "+":
>>
>> address@hidden ~]$ ls -d /proc/device-tree/cpus/PowerPC*
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>> /proc/device-tree/cpus/PowerPC,address@hidden
>>
>> And this is a host, not a guest. I do not see any good reason to make dt
>> names different.
>>
>> And this does not really matter if there is "+" or not for anybody as far
>> as we concerned, ppc64_cpu works either way.
> 
> Right, it may not matter, but I expect you to reference the above commit
> id and explain why it should be POWER7+ after all. You failed to come up
> with that answer before that patch got applied, so we need to correct
> me/it now.
>
> I have checked with Dinar that under Linux using the Sapphire firmware
> "PowerPC,address@hidden" does indeed show up in /proc/device-tree/cpus. So
> that matches what this patch changes and what you report above.
> What could be different in Jacques' setup that he reported it different
> from us? He was checking from AIX, is that possibly using a different
> firmware, pHyp as for my POWER5+?

It must be pHyp, I do not see any other options.

> In any case let's please document this properly in the commit message
> and not just make contradictory statements about what things should be.

I have no idea how to document this. No specification tells what the naming
should be so anything I write there is just my assumption.

"This defines the cpu node name as PowerPC,POWER7+ to stay in sync with the
Sapphire host-side firmware"?


> Also, in qemu.git POWER7 does not have the VSX flag, only the
> instruction set VSX flag. The addition of this VSX flag for POWER7+ is
> not mentioned in the commit message. Does it depend on any of the
> lengthy VSX Stage X series on the list or something in ppc-next changing
> it for POWER7?

The PPC-related patches I post are always made against Alex Graf "ppc-next"
tree and his tree contains VSX fixes. Since my patch simply copies POWER7
family, I do not see much sense in mentioning all the CPU features it
enables for the new family.


> Either way, if you or Alex improve on the commit message then you can
> add my Reviewed-by, I verified that the VSX flag, desc and fw_name are
> the only differences.
> 
> Thanks,
> Andreas
> 


-- 
Alexey