Re: [PATCH v5 07/10] KVM: arm64: Treat CTR_EL0 as a VM feature ID register

[Marc Zyngier <maz@kernel.org>]

On Tue, 10 Sep 2024 08:16:49 +0100,
Shameerali Kolothum Thodi <shameerali.kolothum.thodi@huawei.com> wrote:
> 
> And now to the elephant in the room, handling MIDR differences and associated errata
> management 😊
> 
> Marc, you mentioned about a  prototype solution you have a while back[0],
> has that been shared public?

Nah, and I didn't have any time to get back to it. But maybe that's a
KVM Forum hackathon project! To be honest, this has little to do with
KVM itself. It is mostly guest enlightenment and a bit of
infrastructure to communicate the constraints back to the guest.

> Also not sure ARM has any plans to make this a specification soon.

Probably not until we come up with a decent prototype that solves the
problem.

> I was thinking of handling this in userspace for now by ignoring the MIDR write
> error on Migration and keeping the host MIDR value for destination VM. This is
> for use cases where we know that the hosts doesn't have any MIDR based errata or
> errata difference doesn't matter and assuming the user knows what they are doing.

If you know of *any* host that doesn't have errata to be mitigated in
the guest, send it to me, and I'll find you one pretty quickly! :D

But yes, you can always play that game (which is what I do for
big-little by letting the guest migrating between CPU types). But it
is ultimately doomed.

> Please let me know, if there are any other suggestions or work in progress on this one.

I don't have much free time at the moment, but I'll do what I can.

Thanks,

	M.

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