On Thu, Nov 05, 2026 at 01:56:51PM +0000, Peter Maydell wrote:
> On Thu, 5 Nov 2026 at 13:48, Daniel P. Berrangé <[email protected]> wrote:
> >
> > On Thu, Nov 05, 2026 at 01:45:00PM +0000, Peter Maydell wrote:
> > > On Thu, 5 Nov 2026 at 12:42, Daniel P. Berrangé <[email protected]> 
> > > wrote:
> > > > So that essentially matches the historical semantics where one
> > > > CPU class magically changes its behaviour based on what target
> > > > is running.  So in that sense, it works, and is no worse than
> > > > the semantics we've had for 'max' models forever.
> > > >
> > > > Splitting into 2 distinct types is worthwhile if we want to
> > > > make it possible for the aarch64 emulator to support a runtime
> > > > choice of both 32-bit and 64-bit "max" models.
> > >
> > > What would the difference between that and configuring the 64-bit
> > > capable "max" to not have AArch64 via "-cpu max-v8:aarch64=off"
> > > (which should work today) ?
> >
> > Is that really the same result ?
> >
> > It seems like aarch64_max_v8_tcg_initfn does alot more CPU modelling
> > than aarch32_max_v8_tcg_initfn, than just a single flag ?
> 
> Conceptually it gets you to the same place -- a 32-bit only CPU
> with all the 32-bit features turned on that we support.
> 
> aarch64_max_v8_tcg_initfn() sets up all the 64-bit-only ID registers,
> which is most of what its code is doing. But the handling of
> "aarch64=off" then clears them all out via arm_clear_aarch64_idregs(),
> and even if it didn't, any 64-bit-only features are by definition
> unreachable from 32-bit code.
> 
> It's possible it might not get you 100% absolutely an identical thing
> from the guest's point of view, but any deviations should be minor
> and I think we might consider them bugs.
> 
> There are some minor UI-level things relating to CPU properties,
> e.g. "qemu-system-arm -cpu max,sve=on" gives a property not found
> error, but "qemu-system-aarch64 -cpu max-v8,aarch64=off,sve=on"
> will run (with no visible difference to the guest since SVE
> is AArch64 only). It probably ought to error out complaining
> that you've asked for two things that conflict, but this is
> "minor bug" I think.

Ok, in that case I'd agree aarch64 wouldn't need to expose a
distinct 32-bit CPU model for max


With regards,
Daniel
-- 
|: https://berrange.com       ~~        https://hachyderm.io/@berrange :|
|: https://libvirt.org          ~~          https://entangle-photo.org :|
|: https://pixelfed.art/berrange   ~~    https://fstop138.berrange.com :|


Reply via email to