[PATCH] x86/sev: Fix position dependent variable references in startup code

[Ard Biesheuvel]

On Tue, 6 Feb 2024 at 14:00, Borislav Petkov <bp at alien8.de> wrote:
>
> On Tue, Feb 06, 2024 at 11:19:06AM +0000, Ard Biesheuvel wrote:
> > Looks good to me - thanks.
>
> Thanks.
>
> Now, next question: I'm presuming we want this in stable?
>

Yes.

> If so, which one?
>

Ideally, any kernel that could reasonably be expected to work
correctly in a SEV guest should get this. And IMHO, the state of SEV
host support for any of its flavors is irrelevant there. But let's see
how painful it gets working our way back. I think Kevin mentioned
v5.10 but I haven't had a look yet how feasible that is.

> And if there are conflicts during backporting over there, would you guys
> address them and test the backports?
>

Yes - I cannot speak for Kevin outright but I'm sure that between him
and me, we will be able to allocate the time and effort to make sure
that those conflicts are resolved and the resolutions tested
sufficiently. (With Kevin's help, I have been able to test these
changes myself on our internal SEV infrastructure)

> I'm presuming there's only a subset of stable kernels which you care
> about. If so, we probably don't want to backport it all the way back to
> 5.19... but mark it for a later kernel only...
>
> Thoughts?
>

One slight complication is that the change currently relies on your
patch to remove CONFIG_AMD_MEM_ENCRYPT_ACTIVE_BY_DEFAULT, so if we
don't want to backport that, I will have to tweak the it. Personally,
I think it would be fine but others may feel differently.