Re: [PATCH 00/10] KVM: arm64: Adopt scoped resource management (guard) for EL1 and EL2

[Fuad Tabba]

Hi Marc,

On Tue, 17 Mar 2026 at 08:20, Marc Zyngier <maz@xxxxxxxxxx> wrote:
>
> On Mon, 16 Mar 2026 17:35:21 +0000,
> Fuad Tabba <tabba@xxxxxxxxxx> wrote:
> >
> > Hi everyone,
> >
> > I stumbled upon a lock leak while reviewing Sebastien's recent GIC Hardening
> > series [1] on an early error path. It reminded me that we've run into this exact
> > kind of bug pretty often while working on pKVM. Since we're going to be
> > hopefully upstream even more pKVM code soon, I'd like to try and fix this
> > pattern once and for all.
>
> I'm of the opposite opinion. I'd rather convert things as code gets
> modified, because this otherwise makes backporting patches a real pain
> (we have been through that exercise with the irq stack, and it wasn't
> an enjoyable experience).
>
> So while I'm not opposed to this in general, I'd rather see it as a
> prefix for new features, instead of a standalone series.

I'm happy to do whatever is easier for you. My thinking was that,
since the coming pKVM series (assuming Will's series goes in soon) is
going to be pretty long, it would be easier to have this done earlier.

>
> [...]
>
> > There are definitely other parts of the KVM codebase that could benefit from
> > this (especially the vGIC), but I'm stopping here for now to see what everyone
> > thinks of the approach before touching anything else.
>
> The vgic is specially difficult to convert because of some of the
> constructs that take a lock in a function and release it in another,
> defeating the scope-based locking.
>
> I'll have a look at the series anyway.

Thanks! In case you think this is a good idea, let me know and I'll
respin without those "Change-ids".

Cheers,
/fuad

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