Re: [PATCH v14 08/11] qapi/s390/cpu topology: change-topology monitor command

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

On 1/18/23 16:17, Kevin Wolf wrote:
> Am 18.01.2023 um 11:53 hat Thomas Huth geschrieben:
>> On 17/01/2023 14.31, Nina Schoetterl-Glausch wrote:
>>> On Tue, 2023-01-17 at 08:30 +0100, Thomas Huth wrote:
>>>> On 16/01/2023 22.09, Nina Schoetterl-Glausch wrote:
>>>>> On Thu, 2023-01-05 at 15:53 +0100, Pierre Morel wrote:
>>>>>> The modification of the CPU attributes are done through a monitor
>>>>>> commands.
>>>>>>
>>>>>> It allows to move the core inside the topology tree to optimise
>>>>>> the cache usage in the case the host's hypervizor previously
>>>>>> moved the CPU.
>>>>>>
>>>>>> The same command allows to modifiy the CPU attributes modifiers
>>>>>> like polarization entitlement and the dedicated attribute to notify
>>>>>> the guest if the host admin modified scheduling or dedication of a vCPU.
>>>>>>
>>>>>> With this knowledge the guest has the possibility to optimize the
>>>>>> usage of the vCPUs.
>>>>>>
>>>>>> Signed-off-by: Pierre Morel <pmorel@linux.ibm.com>
>>>>>> ---
>> ...
>>>>>> +    s390_topology.sockets[s390_socket_nb(id)]--;
>>>>>
>>>>> I suppose this function cannot run concurrently, so the same CPU doesn't get removed twice.
>>>>
>>>> QEMU has the so-called BQL - the Big Qemu Lock. Instructions handlers are
>>>> normally called with the lock taken, see qemu_mutex_lock_iothread() in
>>>> target/s390x/kvm/kvm.c.
>>>
>>> That is good to know, but is that the relevant lock here?
>>> We don't want to concurrent qmp commands. I looked at the code and it's pretty complicated.
>>
>> Not sure, but I believe that QMP commands are executed from the main
>> iothread, so I think this should be safe? ... CC:-ing some more people who
>> might know the correct answer.
> 
> In general yes, QMP commands are processed one after another in the main
> thread while holding the BQL. And I think this is the relevant case for
> you.
> 
> The exception is out-of-band commands, which I think run in the monitor
> thread while some other (normal) command could be processed. OOB
> commands are quite limited in what they are allowed to do, though, and
> there aren't many of them. They are mainly meant to fix situations where
> something (including other QMP commands) got stuck.
> 
> Kevin
> 

Thanks Kevin,

regards,
Pierre
-- 
Pierre Morel
IBM Lab Boeblingen