Re: [RFC PATCH v2 13/19] tests: do not run test-hmp on all machines for ARM KVM-only

[Fabiano Rosas]

Claudio Fontana <cfontana@suse.de> writes:

> On 1/10/23 15:02, Peter Maydell wrote:
>> On Tue, 10 Jan 2023 at 13:36, Fabiano Rosas <farosas@suse.de> wrote:
>>>
>>> Peter Maydell <peter.maydell@linaro.org> writes:
>>>
>>>> On Tue, 10 Jan 2023 at 13:00, Fabiano Rosas <farosas@suse.de> wrote:
>>>>>
>>>>> Thomas Huth <thuth@redhat.com> writes:
>>>>>
>>>>>> On 09/01/2023 23.42, Fabiano Rosas wrote:
>>>>>>> From: Claudio Fontana <cfontana@suse.de>
>>>>>>>
>>>>>>> on ARM we currently list and build all machines, even when
>>>>>>> building KVM-only, without TCG.
>>>>>>>
>>>>>>> Until we fix this (and we only list and build machines that are
>>>>>>> compatible with KVM), only test specifically using the "virt"
>>>>>>> machine in this case.
>>>>>>
>>>>>> Why don't you fix it immediately? ...
>>>>>
>>>>> My idea was to have in this series the minimum to unbreak the
>>>>> --disable-tcg build and later bring the rest of the changes
>>>>> incrementally.
>>>>
>>>> This test is basically checking "all the machines should
>>>> work". That's an important invariant to maintain, so
>>>> I don't think we should just skip it for Arm targets.
>>>
>>> But what does "all machines" mean in a no-TCG build? The original intent
>>> of the patch was that only 'virt' should be present and therefore the
>>> only one tested.
>> 
>> It means "every machine the user can create". If the
>> machine can't run then either we shouldn't compile it
>> in, or else we should be compiling in enough extra stuff
>> to allow it to work.
>> 
>> -- PMM
>
> Hi,
>
> once upon a time there was a series by Philippe to accomplish that (KConfig) 
> right?

Is this it?
https://www.mail-archive.com/qemu-devel@nongnu.org/msg777719.html

I have now moved the cpus into the tcg directory:

# ./qemu-system-aarch64 -cpu ? 
Available CPUs:
  a64fx
  cortex-a35
  cortex-a53
  cortex-a55
  cortex-a57
  cortex-a72
  cortex-a76
  host
  max
  neoverse-n1

What remains is changing the Kconfig to not bring all the
machines. Nowadays for KVM is the 'virt' machine the only one we use? I
did some tests yesterday by keeping only CONFIG_ARM_VIRT and was left
with this:

# ./qemu-system-aarch64 -machine ? 
Supported machines are:
none                 empty machine
virt-2.10            QEMU 2.10 ARM Virtual Machine
virt-2.11            QEMU 2.11 ARM Virtual Machine
virt-2.12            QEMU 2.12 ARM Virtual Machine
virt-2.6             QEMU 2.6 ARM Virtual Machine
virt-2.7             QEMU 2.7 ARM Virtual Machine
virt-2.8             QEMU 2.8 ARM Virtual Machine
virt-2.9             QEMU 2.9 ARM Virtual Machine
virt-3.0             QEMU 3.0 ARM Virtual Machine
virt-3.1             QEMU 3.1 ARM Virtual Machine
virt-4.0             QEMU 4.0 ARM Virtual Machine
virt-4.1             QEMU 4.1 ARM Virtual Machine
virt-4.2             QEMU 4.2 ARM Virtual Machine
virt-5.0             QEMU 5.0 ARM Virtual Machine
virt-5.1             QEMU 5.1 ARM Virtual Machine
virt-5.2             QEMU 5.2 ARM Virtual Machine
virt-6.0             QEMU 6.0 ARM Virtual Machine
virt-6.1             QEMU 6.1 ARM Virtual Machine
virt-6.2             QEMU 6.2 ARM Virtual Machine
virt-7.0             QEMU 7.0 ARM Virtual Machine
virt-7.1             QEMU 7.1 ARM Virtual Machine
virt-7.2             QEMU 7.2 ARM Virtual Machine
virt                 QEMU 8.0 ARM Virtual Machine (alias of virt-8.0)
virt-8.0             QEMU 8.0 ARM Virtual Machine
x-remote             Experimental remote machine

But qtest dislikes that even more. I need to figure out why.

The approach of that series seems interesting, moving CONFIG_FOO=y from
default.mak into Kconfig as 'default y if TCG'. Maybe that will yield
better results, I'll try to follow that route.