Re: [PATCH v10 2/3] KVM: s390: guest support for topology function

[Pierre Morel <pmorel@linux.ibm.com>]

On 6/28/22 17:01, Janis Schoetterl-Glausch wrote:
> On 6/28/22 16:13, Pierre Morel wrote:
>>
>>
>> On 6/28/22 14:18, Janis Schoetterl-Glausch wrote:
>>> On 6/28/22 12:58, Pierre Morel wrote:
>>>>
>>>>
>>>> On 6/28/22 10:59, Janis Schoetterl-Glausch wrote:
>>>>> On 6/20/22 14:54, Pierre Morel wrote:
>>>>>> We report a topology change to the guest for any CPU hotplug.
>>>>>>
>>>>>> The reporting to the guest is done using the Multiprocessor
>>>>>> Topology-Change-Report (MTCR) bit of the utility entry in the guest's
>>>>>> SCA which will be cleared during the interpretation of PTF.
>>>>>>
>>>>>> On every vCPU creation we set the MCTR bit to let the guest know the
>>>>>> next time he uses the PTF with command 2 instruction that the
>>>>>> topology changed and that he should use the STSI(15.1.x) instruction
>>>>>> to get the topology details.
>>>>>>
>>>>>> STSI(15.1.x) gives information on the CPU configuration topology.
>>>>>> Let's accept the interception of STSI with the function code 15 and
>>>>>> let the userland part of the hypervisor handle it when userland
>>>>>> support the CPU Topology facility.
>>>>>>
>>>>>> Signed-off-by: Pierre Morel <pmorel@linux.ibm.com>
>>>>>> ---
>>>>>>     arch/s390/include/asm/kvm_host.h | 11 ++++++++---
>>>>>>     arch/s390/kvm/kvm-s390.c         | 27 ++++++++++++++++++++++++++-
>>>>>>     arch/s390/kvm/priv.c             | 15 +++++++++++----
>>>>>>     arch/s390/kvm/vsie.c             |  3 +++
>>>>>>     4 files changed, 48 insertions(+), 8 deletions(-)
>>>>>>
>>>>> [...]
>>>>>
>>>>>> diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
>>>>>> index 8fcb56141689..95b96019ca8e 100644
>>>>>> --- a/arch/s390/kvm/kvm-s390.c
>>>>>> +++ b/arch/s390/kvm/kvm-s390.c
>>>>>> @@ -1691,6 +1691,25 @@ static int kvm_s390_get_cpu_model(struct kvm *kvm, struct kvm_device_attr *attr)
>>>>>>         return ret;
>>>>>>     }
>>>>>>
>>>>>> +/**
>>>>>> + * kvm_s390_sca_set_mtcr
>>>>>
>>>>> I wonder if there is a better name, kvm_s390_report_topology_change maybe?
>>>>>
>>>>>> + * @kvm: guest KVM description
>>>>>> + *
>>>>>> + * Is only relevant if the topology facility is present,
>>>>>> + * the caller should check KVM facility 11
>>>>>> + *
>>>>>> + * Updates the Multiprocessor Topology-Change-Report to signal
>>>>>> + * the guest with a topology change.
>>>>>> + */
>>>>>> +static void kvm_s390_sca_set_mtcr(struct kvm *kvm)
>>>>>> +{
>>>>>
>>>>> Do we need a sca_lock read_section here? If we don't why not?
>>>>> Did not see one up the stack, but I might have overlooked something.
>>>>
>>>> Yes we do.
>>>> As I said about your well justified comment in a previous mail, ipte_lock is not the right thing to use here and I will replace with an inter locked update.
>>>
>>> Not sure I'm understanding you right, you're saying we need both? i.e.:
>>>
>>> struct bsca_block *sca;
>>>
>>> read_lock(&vcpu->kvm->arch.sca_lock);
>>> sca = kvm->arch.sca;
>>> atomic_or(SCA_UTILITY_MTCR, &sca->utility);
>>> read_unlock(&vcpu->kvm->arch.sca_lock);
>>>
>>> Obviously you would need to change the definition of the utility field and could not use a bit field like Janosch
>>> suggested, unless you want to use a cmpxchg loop.
>>> It's a bit ugly that utility is a two byte value.
>>> Maybe there is a nicer way to set that bit, OR (OI, OIY) seem appropriate, but I don't know if they have a nice
>>> abstraction in Linux or if you'd need inline asm.
>>
>> I was think to something like this because it is what is used most of the time when a bit is to be change concurrently with firmware.
> 
> Ah, ok you want to keep the bitfield.
> 
> [...]
>>
>> static void kvm_s390_sca_set_mtcr(struct kvm *kvm, int val)
> 
> If you use a bool val you can simply do new.mtcr = val; below.
>> {
>>          struct bsca_block *sca = kvm->arch.sca;
>>          union sca_utility new, old;
>>
>>          read_lock(&kvm->arch.sca_lock);
> 
> Don't forget to move the sca = kvm->arch.sca; under the lock here.
>>          do {
>>                  old = READ_ONCE(sca->utility);
>>                  new = old;
>>                  new.mtcr = val ? 1 : 0;
>>          } while (cmpxchg(&sca->utility.val, old.val, new.val) != old.val);
>>          read_lock(&kvm->arch.sca_lock);
>> }

right, thanks, and to unlock at the end :)

>>
>>>>
>>>>>
>>>>>> +    struct bsca_block *sca = kvm->arch.sca; /* SCA version doesn't matter */
>>>>>> +
>>>>>> +    ipte_lock(kvm);
>>>>>> +    sca->utility |= SCA_UTILITY_MTCR;
>>>>>> +    ipte_unlock(kvm);
>>>>>> +}
>>>>>> +
>>>>>
>>>>> [...]
>>>>>
>>>>
>>>
>>
> 

-- 
Pierre Morel
IBM Lab Boeblingen