Re: QEMU 32-bit vs. 64-bit binaries

[Peter Maydell]

On Tue, 10 May 2022 at 10:01, Thomas Huth <thuth@redhat.com> wrote:
>
> On 10/05/2022 10.54, Markus Armbruster wrote:
> > Thomas Huth <thuth@redhat.com> writes:
> >
> > [...]
> >
> >> I once suggested in the past already that we should maybe get rid of
> >> the 32-bit variants in case the 64-bit variant is a full superset, so
> >> we can save compile- and test times (which is quite a bit for QEMU),
> >> but I've been told that the 32-bit variants are mostly still required
> >> for supporting KVM on 32-bit host machines.
> >
> > Do we still care for 32-bit host machines?
>
> As long as the Linux kernel still supports 32-bit KVM virtualization, I
> think we have to keep the userspace around for that, too.
>
> But I wonder why we're keeping qemu-system-arm around? 32-bit KVM support
> for ARM has been removed with Linux kernel 5.7 as far as I know, so I think
> we could likely drop the qemu-system-arm nowadays, too? Peter, Richard,
> what's your opinion on this?

Two main reasons, I think:
 * command-line compatibility (ie there are lots of
   command lines out there using that binary name)
 * nobody has yet cared enough to come up with a plan for what
   we want to do differently for these 32-bit architectures,
   so the default is "keep doing what we always have"

In particular, I don't want to get rid of qemu-system-arm as the
*only* 32-bit target binary we drop. Either we stick with what
we have or we have a larger plan for sorting this out consistently
across target architectures.

-- PMM