Re: [Qemu-devel] [PULL 57/60] target/ppc: add HV support for POWER9

[Cédric Le Goater]

On 02/07/2019 02:14, David Gibson wrote:
> On Mon, Jul 01, 2019 at 11:45:23AM +0200, Philippe Mathieu-Daudé wrote:
>> On 7/1/19 7:04 AM, David Gibson wrote:
>>> On Fri, Jun 28, 2019 at 03:20:32PM +0200, Philippe Mathieu-Daudé wrote:
>>>> Hi,
>>>>
>>>> On 3/12/19 8:58 PM, Cédric Le Goater wrote:
>>>>> On 3/12/19 8:30 PM, Cleber Rosa wrote:
>>>>>>> From: "Cleber Rosa" <address@hidden>
>>>>>>> Sent: Tuesday, March 12, 2019 3:14:09 PM
>>>>>>> Subject: Re: [Qemu-devel] [PULL 57/60] target/ppc: add HV support for 
>>>>>>> POWER9
>>>>>>>
>>>>>>> On Tue, Mar 12, 2019 at 07:34:04PM +0100, Cédric Le Goater wrote:
>>>>>>>> On 3/12/19 4:01 PM, Cleber Rosa wrote:
>>>>>>>>> On Sun, Mar 10, 2019 at 07:27:00PM +1100, David Gibson wrote:
>>>>>>>>>> From: Cédric Le Goater <address@hidden>
>>>>>>>>>>
>>>>>>>>>> We now have enough support to boot a PowerNV machine with a POWER9
>>>>>>>>>> processor. Allow HV mode on POWER9.
>>>>>>>>>>
>>>>>>>>>> Signed-off-by: Cédric Le Goater <address@hidden>
>>>>>>>>>> Message-Id: <address@hidden>
>>>>>>>>>> Signed-off-by: David Gibson <address@hidden>
>>>>>>>>>> ---
>>>>>>>>>>  target/ppc/translate_init.inc.c | 3 ++-
>>>>>>>>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>>>>>>>>
>>>>>>>>>> diff --git a/target/ppc/translate_init.inc.c
>>>>>>>>>> b/target/ppc/translate_init.inc.c
>>>>>>>>>> index af70a3b78c..0bd555eb19 100644
>>>>>>>>>> --- a/target/ppc/translate_init.inc.c
>>>>>>>>>> +++ b/target/ppc/translate_init.inc.c
>>>>>>>>>> @@ -8895,7 +8895,7 @@ POWERPC_FAMILY(POWER9)(ObjectClass *oc, void
>>>>>>>>>> *data)
>>>>>>>>>>                         PPC_CACHE | PPC_CACHE_ICBI | PPC_CACHE_DCBZ |
>>>>>>>>>>                         PPC_MEM_SYNC | PPC_MEM_EIEIO |
>>>>>>>>>>                         PPC_MEM_TLBSYNC |
>>>>>>>>>> -                       PPC_64B | PPC_64BX | PPC_ALTIVEC |
>>>>>>>>>> +                       PPC_64B | PPC_64H | PPC_64BX | PPC_ALTIVEC |
>>>>>>>>>>                         PPC_SEGMENT_64B | PPC_SLBI |
>>>>>>>>>>                         PPC_POPCNTB | PPC_POPCNTWD |
>>>>>>>>>>                         PPC_CILDST;
>>>>>>>>>> @@ -8907,6 +8907,7 @@ POWERPC_FAMILY(POWER9)(ObjectClass *oc, void
>>>>>>>>>> *data)
>>>>>>>>>>                          PPC2_ISA205 | PPC2_ISA207S | 
>>>>>>>>>> PPC2_FP_CVT_S64 |
>>>>>>>>>>                          PPC2_TM | PPC2_ISA300 | PPC2_PRCNTL;
>>>>>>>>>>      pcc->msr_mask = (1ull << MSR_SF) |
>>>>>>>>>> +                    (1ull << MSR_SHV) |
>>>>>>>>>>                      (1ull << MSR_TM) |
>>>>>>>>>>                      (1ull << MSR_VR) |
>>>>>>>>>>                      (1ull << MSR_VSX) |
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> This change prevents a Fedora 29 kernel[1] from booting... is this
>>>>>>>>> intended or a known limitation of the Fedora 29 kernel?
>>>>>>>>
>>>>>>>> The default CPU is still power8_v2.0. This is curious.
>>>>>>>>
>>>>>>>
>>>>>>> Are you sure?  I'm getting:
>>>>>>>
>>>>>>>  $ git rev-parse HEAD
>>>>>>>  cfc3fef6b4e493bf1a7ee16790ad584e20dfbbd1
>>>>>>>  $ ./ppc64-softmmu/qemu-system-ppc64 -qmp unix:/tmp/qmp-sock,server
>>>>>>>  $ ./scripts/qmp/qom-get -s /tmp/qmp-sock 
>>>>>>> /machine/unattached/device[0].type
>>>>>>>  power9_v2.0-spapr-cpu-core
>>>>>
>>>>> That's a pseries machine, not a powernv machine. pseries should use P9
>>>>> processor by default but the patch above should not impact f29 on 
>>>>> pseries. 
>>>>> If it does, then we have a bug. 
>>>>>  
>>>>>> Looks like the overall default is "power9_v2.0", and then on pseries-3.1 
>>>>>> and
>>>>>> lower, it's "power8_v2.0", as per 34a6b015a98.
>>>>>
>>>>> I was looking at pnv_machine_class_init() which sets the default CPU :
>>>>>
>>>>>     mc->default_cpu_type = POWERPC_CPU_TYPE_NAME("power8_v2.0");
>>>>
>>>> I found this thread while trying auto-bisection for LP#1834613:
>>>> https://bugs.launchpad.net/bugs/1834613
>>>>
>>>> When trying the options suggested by Laurent here:
>>>> https://lists.gnu.org/archive/html/qemu-devel/2019-06/msg06209.html
>>>>
>>>> this one hangs:
>>>>
>>>> $ qemu-system-ppc64 \
>>>>  -kernel vmlinuz-vanilla \
>>>>  -nographic -append "console=hvc0" \
>>>>  -M cap-cfpc=broken,cap-sbbc=broken,cap-ibs=broken
>>>>
>>>> but this one works:
>>>>
>>>> $ qemu-system-ppc64 \
>>>>  -kernel vmlinuz-vanilla \
>>>>  -nographic -append "console=hvc0" \
>>>>  -M pseries-3.1
>>>
>>> Sorry, I missed most of this thread while on holidays.  What's the
>>> actual bug here?
>>
>> I don't think there is a bug, this seems the result of adding a new feature.
>> The commit message is not obvious that old kernels won't work on the
>> default machine type, and we have to add the extra '-M pseries-3.1'
>> command line option to run such images.
> 
> Hm, except apparently turning off the Spectre options doesn't change
> that, so I wonder what change in the newer machine type is causing the
> breakage.
> 

I don't see such an issue on David's ppc-for-4.1 branch.

C.