Re: [PATCH] KVM: s390: introduce module parameter kvm.use_gisa

[Michael Mueller <mimu@linux.ibm.com>]

On 27.02.20 14:24, Christian Borntraeger wrote:
> 
> 
> On 27.02.20 13:43, Michael Mueller wrote:
>>
>>
>> On 27.02.20 13:27, Christian Borntraeger wrote:
>>>
>>>
>>> On 27.02.20 10:10, Michael Mueller wrote:
>>>> The boolean module parameter "kvm.use_gisa" controls if newly
>>>> created guests will use the GISA facility if provided by the
>>>> host system. The default is yes.
>>>>
>>>>     # cat /sys/module/kvm/parameters/use_gisa
>>>>     Y
>>>>
>>>> The parameter can be changed on the fly.
>>>>
>>>>     # echo N > /sys/module/kvm/parameters/use_gisa
>>>>
>>>> Already running guests are not affected by this change.
>>>>
>>>> The kvm s390 debug feature shows if a guest is running with GISA.
>>>>
>>>>     # grep gisa /sys/kernel/debug/s390dbf/kvm-$pid/sprintf
>>>>     00 01582725059:843303 3 - 08 00000000e119bc01  gisa 0x00000000c9ac2642 initialized
>>>>     00 01582725059:903840 3 - 11 000000004391ee22  00[0000000000000000-0000000000000000]: AIV gisa format-1 enabled for cpu 000
>>>>     ...
>>>>     00 01582725059:916847 3 - 08 0000000094fff572  gisa 0x00000000c9ac2642 cleared
>>>>
>>>> In general, that value should not be changed as the GISA facility
>>>> enhances interruption delivery performance.
>>>>
>>>> A reason to switch the GISA facility off might be a performance
>>>> comparison run or debugging.
>>>>
>>>> Signed-off-by: Michael Mueller <mimu@linux.ibm.com>
>>>
>>> Looks good to me. Regarding the other comments, I think allowing for dynamic changes
>>> and keeping use_gisa vs disable_gisa makes sense. So I would think that the patch
>>> as is makes sense.
>>>
>>> The only question is: shall we set use_gisa to 0 when the machine does not support
>>> it (e.g. VSIE?) and then also forbid setting it to 1? Could be overkill.
>>
>> Then I would rename the parameter to "try_to_use_gisa" instead. (a joke ;) )
>>
>> In that case we exit gisa_init() because of the missing AIV facility.
>>
>> void kvm_s390_gisa_init(struct kvm *kvm)
>> {
>>          struct kvm_s390_gisa_interrupt *gi = &kvm->arch.gisa_int;
>>
>> -->    if (!css_general_characteristics.aiv)
>>                  return;
>>          gi->origin = &kvm->arch.sie_page2->gisa;
>>          gi->alert.mask = 0;
>>      ...
>> }
>>
> 
> I know. My point was more: "can we expose this". But this is probably overkill.

I agree with Connie here, that would make the whole thing
just more error-prone. That way the messages are at least
consistent.

>