Re: [PATCH RFC] kvm: x86: add halt_poll module parameter

[Rik van Riel <riel@redhat.com>]

On 02/05/2015 02:20 PM, Paolo Bonzini wrote:
> 
> 
> On 05/02/2015 19:55, Jan Kiszka wrote:
>>> This patch introduces a new module parameter for the KVM module; when it
>>> is present, KVM attempts a bit of polling on every HLT before scheduling
>>> itself out via kvm_vcpu_block.
>>
>> Wouldn't it be better to tune this on a per-VM basis? Think of mixed
>> workloads with some latency-sensitive and some standard VMs.
> 
> Yes, but:
> 
> 1) this turned out to be very cheap, so a per-host tunable is not too bad;
> 
> 2) it also affects only very few workloads (for example network
> workloads can already do polling in the guest) so it only affects few
> people;
> 
> 3) long term anyway we want it to auto tune, which is better than tuning
> it per-VM.

We may want to auto tune it per VM.

However, if we make auto tuning work well, I do not
think we want to expose a user visible tunable per
VM, and commit to keeping that kind of interface
around forever.