Re: [PATCH] x86/boot/sev: Support memory acceptance in the EFI stub under SVSM

[Ard Biesheuvel]

On Sun, 4 May 2025 at 09:38, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
>
> * Ard Biesheuvel <ardb@xxxxxxxxxx> wrote:
>
> > On Thu, 1 May 2025 at 20:05, Tom Lendacky <thomas.lendacky@xxxxxxx> wrote:
> > >
> > > On 4/28/25 12:43, Ard Biesheuvel wrote:
> > > > From: Ard Biesheuvel <ardb@xxxxxxxxxx>
> > > >
> > > > Commit
> > > >
> > > >   d54d610243a4 ("x86/boot/sev: Avoid shared GHCB page for early memory acceptance")
> > > >
> > > > provided a fix for SEV-SNP memory acceptance from the EFI stub when
> > > > running at VMPL #0. However, that fix was insufficient for SVSM SEV-SNP
> > > > guests running at VMPL >0, as those rely on a SVSM calling area, which
> > > > is a shared buffer whose address is programmed into a SEV-SNP MSR, and
> > > > the SEV init code that sets up this calling area executes much later
> > > > during the boot.
> > > >
> > > > Given that booting via the EFI stub at VMPL >0 implies that the firmware
> > > > has configured this calling area already, reuse it for performing memory
> > > > acceptance in the EFI stub.
> > >
> > > This looks to be working for SNP guest boot and kexec. SNP guest boot with
> > > an SVSM is also working, but kexec isn't. But the kexec failure of an SVSM
> > > SNP guest is unrelated to this patch, I'll send a fix for that separately.
> > >
> >
> > Thanks for confirming.
> >
> > Ingo, Boris, can we get this queued as a fix, please, and merge it
> > back into x86/boot as was done before?
>
> Just to clarify, memory acceptance trough the EFI stub from VMPL >0
> SEV-SNP guests was broken last summer via fcd042e86422, and it hasn't
> worked since then?
>

It never worked correctly at all for SEV-SNP, since it was enabled in
d54d610243a4.

We never noticed because it appears that the SEV-SNP firmware rarely
exposes EFI_UNACCEPTED regions in chunks that are not 2M aligned, and
the EFI stub only accepts the misaligned bits so it can populate the
unaccepted_memory table accurately, which keeps track of unaccepted
memory with 2M granularity.