Re: [PATCH v2 1/2] hw/i386/isapc.c: remove support for -cpu host and -cpu max

[Igor Mammedov <imammedo@redhat.com>]

On Fri, 26 Sep 2025 16:22:31 +0100
Mark Cave-Ayland <mark.caveayland@nutanix.com> wrote:

> On 26/09/2025 16:12, Daniel P. Berrangé wrote:
> 
> > On Fri, Sep 26, 2025 at 04:01:23PM +0100, Mark Cave-Ayland wrote:  
> >> On 26/09/2025 14:55, Daniel P. Berrangé wrote:
> >>  
> >>> On Fri, Sep 26, 2025 at 02:49:00PM +0100, Mark Cave-Ayland wrote:  
> >>>> On 26/09/2025 13:39, Daniel P. Berrangé wrote:
> >>>>  
> >>>>> On Fri, Sep 26, 2025 at 01:36:25PM +0200, Markus Armbruster wrote:  
> >>>>>> Mark Cave-Ayland <mark.caveayland@nutanix.com> writes:
> >>>>>>  
> >>>>>>> Following recent discussions on the mailing list, it has been decided
> >>>>>>> that instead of mapping -cpu host and -cpu max to a suitable 32-bit x86 CPU,
> >>>>>>> it is preferable to disallow them and use the existing valid_cpu_types
> >>>>>>> validation logic so that an error is returned to the user instead.
> >>>>>>>
> >>>>>>> Signed-off-by: Mark Cave-Ayland <mark.caveayland@nutanix.com>
> >>>>>>> ---
> >>>>>>>     hw/i386/isapc.c | 27 ---------------------------
> >>>>>>>     1 file changed, 27 deletions(-)
> >>>>>>>
> >>>>>>> diff --git a/hw/i386/isapc.c b/hw/i386/isapc.c
> >>>>>>> index 44f4a44672..6c35a397df 100644
> >>>>>>> --- a/hw/i386/isapc.c
> >>>>>>> +++ b/hw/i386/isapc.c
> >>>>>>> @@ -41,31 +41,6 @@ static void pc_init_isa(MachineState *machine)
> >>>>>>>         DriveInfo *hd[MAX_IDE_BUS * MAX_IDE_DEVS];
> >>>>>>>         int i;
> >>>>>>> -    /*
> >>>>>>> -     * There is a small chance that someone unintentionally passes "-cpu max"
> >>>>>>> -     * for the isapc machine, which will provide a much more modern 32-bit
> >>>>>>> -     * CPU than would be expected for an ISA-era PC. If the "max" cpu type has
> >>>>>>> -     * been specified, choose the "best" 32-bit cpu possible which we consider
> >>>>>>> -     * be the pentium3 (deliberately choosing an Intel CPU given that the
> >>>>>>> -     * default 486 CPU for the isapc machine is also an Intel CPU).
> >>>>>>> -     */
> >>>>>>> -    if (!strcmp(machine->cpu_type, X86_CPU_TYPE_NAME("max"))) {
> >>>>>>> -        machine->cpu_type = X86_CPU_TYPE_NAME("pentium3");
> >>>>>>> -        warn_report("-cpu max is invalid for isapc machine, using pentium3");
> >>>>>>> -    }
> >>>>>>> -
> >>>>>>> -    /*
> >>>>>>> -     * Similarly if someone unintentionally passes "-cpu host" for the isapc
> >>>>>>> -     * machine then display a warning and also switch to the "best" 32-bit
> >>>>>>> -     * cpu possible which we consider to be the pentium3. This is because any
> >>>>>>> -     * host CPU will already be modern than this, but it also ensures any
> >>>>>>> -     * newer CPU flags/features are filtered out for older guests.
> >>>>>>> -     */
> >>>>>>> -    if (!strcmp(machine->cpu_type, X86_CPU_TYPE_NAME("host"))) {
> >>>>>>> -        machine->cpu_type = X86_CPU_TYPE_NAME("pentium3");
> >>>>>>> -        warn_report("-cpu host is invalid for isapc machine, using pentium3");
> >>>>>>> -    }
> >>>>>>> -
> >>>>>>>         if (machine->ram_size > 3.5 * GiB) {
> >>>>>>>             error_report("Too much memory for this machine: %" PRId64 " MiB, "
> >>>>>>>                          "maximum 3584 MiB", machine->ram_size / MiB);
> >>>>>>> @@ -162,8 +137,6 @@ static void isapc_machine_options(MachineClass *m)
> >>>>>>>             X86_CPU_TYPE_NAME("pentium2"),
> >>>>>>>             X86_CPU_TYPE_NAME("pentium3"),
> >>>>>>>             X86_CPU_TYPE_NAME("qemu32"),
> >>>>>>> -        X86_CPU_TYPE_NAME("max"),
> >>>>>>> -        X86_CPU_TYPE_NAME("host"),
> >>>>>>>             NULL
> >>>>>>>         };
> >>>>>>>         PCMachineClass *pcmc = PC_MACHINE_CLASS(m);  
> >>>>>>
> >>>>>> This reverts the "smart" part of recent
> >>>>>>
> >>>>>> commit e1e2909f8e74051a34a044940f90d4650b6e784a
> >>>>>> Author: Mark Cave-Ayland <mark.caveayland@nutanix.com>
> >>>>>> Date:   Thu Aug 28 12:09:44 2025 +0100
> >>>>>>
> >>>>>>        hw/i386/pc_piix.c: restrict isapc machine to 32-bit CPUs
> >>>>>>        The isapc machine represents a legacy ISA PC with a 486 CPU. Whilst it is
> >>>>>>        possible to specify any CPU via -cpu on the command line, it makes no
> >>>>>>        sense to allow modern 64-bit CPUs to be used.
> >>>>>>        Restrict the isapc machine to the available 32-bit CPUs, taking care to
> >>>>>>        handle the case where if a user inadvertently uses either -cpu max or
> >>>>>>        -cpu host then the "best" 32-bit CPU is used (in this case the pentium3).  
> >>>>>
> >>>>> What is written here made sense from the POV of use of isapc with
> >>>>> qemu-system-x86_64, but in qemu-system-i686, both 'max' and 'host'
> >>>>> where already 32-bit CPUs IIUC. Both this original patch and
> >>>>> the new patch block them from being used in qemu-system-i686
> >>>>> which feels wrong given the justification above.  
> >>>>
> >>>> I tried stepping through with -cpu host/-cpu max on qemu-system-i386 and
> >>>> it's a bit confusing: I think we end up with some kind of custom AMD vendor
> >>>> CPU but with LM disabled. I can't easily see a way to understand what
> >>>> features are currently enabled?
> >>>>
> >>>> I must admit I'm struggling to see the usefulness of -cpu host/-cpu max for
> >>>> isapc given that older OSs can be quite picky when it comes to hardware.  
> >>>
> >>> But x86 CPU vendors go to ridiculous levels of complexity to retain
> >>> historical back compat over many decades. If anything, I'd be surprised
> >>> about the opposite - an OS that didn't work with -cpu max.  
> >>
> >> It's not the CPU vendor I'd be worried about, but the OS vendor who may for
> >> example execute CPUID and become confused if it returns an AMD vendor ID
> >> instead of an Intel vendor ID.  

ha, that's how KVM works if I recall it right, one doesn't need host/max for it,
flip to host's vendor would happen with _any_ cpumodel.
That's leads to some degree of fun on guest OS side to deal with.

> > 
> > IIRC/IIUC, AMD sold i486 CPUs with CPUID present in the ISA era, so I
> > would have thought anything checking vendor ID should expect to see
> > more than just Intel ?  
> 
> Unfortunately I don't remember the specifics, but I have a vague memory 
> that someone has mentioned this to me in the past.
> >> What do you think is the best way forward? I'm still not convinced of the
> >> utility of -cpu host/-cpu max for isapc, so what if instead of mapping them
> >> to the pentium3 CPU we follow the standard deprecation path and emit a
> >> warning on startup, and then remove them from valid_cpu_types in 2 releases
> >> time? The advantage to this approach is that if people are actually using
> >> -cpu host/-cpu max with the isapc machine then they would at least file an
> >> issue in Gitlab and make us aware of it.  
> > 
> > If we want to deprecate it the formal route, that's fine.  
> 
> Works for me. Igor, any objections?

given host/max are accelerator specific models with all possible features for it,
it' would be fair to make sure that those models in qemu-system-i686 case expose
only 32-bit CPU's features
(do we wanna go to that extent for i386? I'm not sure)

However let's not confuse it with capabilities of specific machine type,
it shouldn't have any influence on cpu models. It looks like this patch does
the right thing /limiting permitted cpus to back then reality/.

If we strictly follow rules,
then deprecation is the way to go, so it wouldn't look as regression.

(aka QEMU used to start fine with host/max, and we warn in advance that it's
going out in favor of .../because of ...
shout out if you have a use case.)
 
> 
> 
> ATB,
> 
> Mark.
>