Re: [RFC PATCH 1/1] Dirty quota-based throttling of vcpus

[Shivam Kumar <shivam.kumar1@nutanix.com>]

On 22/11/22 9:30 am, Shivam Kumar wrote:
> 
> 
> On 21/11/22 5:05 pm, Philippe Mathieu-Daudé wrote:
>> Hi,
>>
>> On 20/11/22 23:54, Shivam Kumar wrote:
>>> +
>>> +void dirty_quota_migration_start(void)
>>> +{
>>> +    if (!kvm_state->dirty_quota_supported) {
>>
>> You are accessing an accelerator-specific variable in an 
>> accelerator-agnostic file, this doesn't sound correct.
>>
>> You might introduce some hooks in AccelClass and implement them in
>> accel/kvm/. See for example gdbstub_supported_sstep_flags() and
>> kvm_gdbstub_sstep_flags().
>>
> Ack.
> 
> Thanks,
> Shivam

Hi Philippe,

I had received a suggestion on the kernel-side patchset to make dirty 
quota a more generic feature and not limit its use to live migration. 
Incorporating this ask might lead to a significant change in the dirty 
quota interface. So, I haven't been able to post the next version of the 
QEMU patchset. I intend to post it once the new proposition looks good 
to the KVM reviewers.

Thanks,
Shivam