[PATCH] KVM: arm64: Consider NUMA affinity when allocating per-CPU stack_page

[Marc Zyngier]

On Thu, 18 Apr 2024 07:53:58 +0100,
"Li,Rongqing" <lirongqing at baidu.com> wrote:
> > Li RongQing <lirongqing at baidu.com> wrote:
> > >
> > > per-CPU stack_page are dominantly accessed from their own local CPUs,
> > > so allocate them node-local to improve performance.
> > 
> > Do you have any performance data to back this up?
> > 
> > Given that this is only used in the non-VHE case, and that by doing so you have
> > left quite a lot of performance on the floor already, I'm even more surprised to
> > see the performance argument.
> > 
> 
> Sorry, I have not setup to test it.
> 
> > Furthermore, you don't address the allocation of per-CPU data, which has a
> > much larger potential impact, given how the HYP code is structured.
> > 
> 
> I will add it in V2

Please test your patch before posting it. I'm not interested in
looking at patches that have not been tested by their author.

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.