lkml.org 
[lkml]   [2022]   [Sep]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v8 003/103] KVM: Refactor CPU compatibility check on module initialization
On Thu, Sep 01, 2022, Marc Zyngier wrote:
> Sean,
>
> On Thu, 11 Aug 2022 18:39:53 +0100,
> Sean Christopherson <seanjc@google.com> wrote:
> >
> > +Will (for arm crud)
>
> When it comes to KVM/arm64, I'd appreciate if you could Cc me.

Sorry, will do.

> > arm64 is also quite evil and circumvents KVM's hardware enabling
> > logic to some extent. kvm_arch_init() => init_subsystems()
> > unconditionally enables hardware, and for pKVM _leaves_ hardware
> > enabled. And then hyp_init_cpu_pm_notifier() disables/enables
> > hardware across lower power enter+exit, except if pKVM is enabled.
> > The icing on the cake is "disabling" hardware doesn't even do
> > anything (AFAICT) if the kernel is running at EL2 (which I think is
> > nVHE + not-pKVM?).
>
> In the cases where disabling doesn't do anything (which are the exact
> opposite of the cases you describe), that's because there is
> absolutely *nothing* to do:

Yes, I know.

> - If VHE, the kernel is the bloody hypervisor: disable virtualisation,
> kill the kernel.
>
> - if pKVM, the kernel is basically a guest, and has no business
> touching anything at all.
>
> So much the 'evil' behaviour.

The colorful language is tongue-in-cheek.

I get the impression that you feel I am attacking ARM. That is very much not what
I intended. If anything, I'm attacking x86 for forcing its quirks on everyone else.

What am trying to point out here is that ARM and other architectures are not
well-served by KVM's current hardware enabling/disabling infrastructure. I am not
saying that ARM is broken and needs to be fixed, I am saying that KVM is broken and
needs to be fixed, and that ARM is a victim of KVM's x86-centric origins.

\
 
 \ /
  Last update: 2022-09-01 16:10    [W:0.143 / U:3.352 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site