Re: [PATCH] target/s390x/kvm: Enable adapter interruption suppression again

[Thomas Huth <thuth@redhat.com>]

On 16/01/2020 21.26, Cornelia Huck wrote:
> On Thu, 16 Jan 2020 15:19:13 -0500
> Matthew Rosato <mjrosato@linux.ibm.com> wrote:
> 
>> On 1/16/20 7:20 AM, Thomas Huth wrote:
>>> The AIS feature has been disabled late in the v2.10 development
>>> cycle since there were some issues with migration (see commit
>>> 3f2d07b3b01ea61126b - "s390x/ais: for 2.10 stable: disable ais
>>> facility"). We originally wanted to enable it again for newer
>>> machine types, but apparently we forgot to do this so far. Let's
>>> do it for the new s390-ccw-virtio-5.0 machine now.
>>>
>>> Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1756946
>>> Signed-off-by: Thomas Huth <thuth@redhat.com>
>>> ---
>>>   hw/s390x/s390-virtio-ccw.c         |  4 ++++
>>>   include/hw/s390x/s390-virtio-ccw.h |  4 ++++
>>>   target/s390x/kvm.c                 | 11 ++++++++---
>>>   3 files changed, 16 insertions(+), 3 deletions(-)
> 
> (...)
> 
>>> diff --git a/target/s390x/kvm.c b/target/s390x/kvm.c
>>> index 15260aeb9a..4c1c8c0208 100644
>>> --- a/target/s390x/kvm.c
>>> +++ b/target/s390x/kvm.c
>>> @@ -329,6 +329,8 @@ static void ccw_machine_class_foreach(ObjectClass *oc, void *opaque)
>>>   
>>>   int kvm_arch_init(MachineState *ms, KVMState *s)
>>>   {
>>> +    S390CcwMachineClass *smc = S390_CCW_MACHINE_OBJ_GET_CLASS(ms);
>>> +  
>>
>> I still can't run a proper test due to unavailable hw but in the 
>> meantime I tried to virsh define a libvirt guest pointed at qemu (master 
>> + this patch).  Regardless of machine type (s390-ccw-virtio-5.0 or 
>> s390-ccw-virtio-4.2) I get:
>>
>> virsh define guest.xml
>> error: Failed to define domain from /path/to/guest.xml
>> error: invalid argument: could not find capabilities for arch=s390x 
>> domaintype=kvm
>>
>> Similarly:
>>
>> virsh domcapabilities
>> error: failed to get emulator capabilities
>> error: invalid argument: unable to find any emulator to serve 's390x' 
>> architecture
>>
>> Rolling back to qemu master, the define and domcapabilities work (with 
>> no ais of course).
>>
>> So: there is some incompatibility between the way libvirt invokes qemu 
>> to detect capabilities and this code.  The above line seems to be the 
>> root problem - if I take your patch and remove 'smc' then libvirt works 
>> as expected and I can see ais in the domcapabilities.
>>
>> Looking at those wrappers David mentioned...  I suspect you need this 
>> for the 'none' machine case.  I tried a quick hack with the following:
>>
>> bool ais_allowed(void)
>> {
>>      /* for "none" machine this results in true */
>>      return get_machine_class()->kvm_ais_allowed;
>> }
>>
>> and
>>
>> if (ais_allowed() &&
>>      kvm_check_extension(s, KVM_CAP_S390_AIS_MIGRATION)) {
>>      kvm_vm_enable_cap(s, KVM_CAP_S390_AIS, 0);
>> }
>>
>> This works and doesn't break libvirt compatibility detection.
> 
> Oh, "none" machine fun again... I think you're on the right track, and
> we really need a wrapper.

D'oh, so this is the real reason for the wrappers ... ok, I'll respin my
patch accordingly.

Thanks a lot for the testing, Matthew!

 Thomas