Re: [PATCH] vfio: selftests: Build tests on aarch64

[Matt Evans]

Hi Alex,

On 19/03/2026 14:37, Alex Williamson wrote:
> On Thu, 19 Mar 2026 13:51:29 +0000
> Matt Evans <mattev@xxxxxxxx> wrote:
> 
>> Hi Alex,
>>
>> On 18/03/2026 22:55, Alex Williamson wrote:
>>> On Tue, 17 Mar 2026 16:25:18 -0700
>>> Ted Logan <tedlogan@xxxxxx> wrote:
>>>   
>>>> Also build vfio self-tests on aarch64 variants, in addition to arm64 and
>>>> x86_64.
>>>>
>>>> Reported-by: Matt Evans <mattev@xxxxxxxx>
>>>> Closes: https://lore.kernel.org/all/e51b4ff2-13c4-47d4-b781-3dcbd740d274@xxxxxxxx/  
>>>> ---
>>>> Signed-off-by: Ted Logan <tedlogan@xxxxxx>  
>>>
>>> Careful, this is a good way to get your Sign-off lost in the tooling.
>>>   
>>>> ---
>>>>  tools/testing/selftests/vfio/Makefile | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/tools/testing/selftests/vfio/Makefile b/tools/testing/selftests/vfio/Makefile
>>>> index 8e90e409e91d..98bc8d6271a4 100644
>>>> --- a/tools/testing/selftests/vfio/Makefile
>>>> +++ b/tools/testing/selftests/vfio/Makefile
>>>> @@ -1,4 +1,4 @@
>>>> -ARCH ?= $(shell uname -m)
>>>> +ARCH ?= $(shell uname -m | sed -e 's/aarch64.*/arm64/')  
>>>
>>> Are we intentionally enabling aarch64_be with this glob or is that just
>>> carry-over from similar in the mm selftest Makefile?
>>>
>>> Big endian ARM64 support is really not on my radar for vfio, but it
>>> wouldn't be our first foray into big endian systems.
>>>
>>> If it wasn't intentional, maybe just add "aarch64" to the below filter
>>> list directly.  Thanks,  
>>
>> I think that won't work for cross-builds passing in ARCH=arm64 "from
>> above".  We can just drop the `.*` part of the sed regexp, giving
>> ARCH=arm64 or arm64_be (the latter would fail to match in the filter below).
>   
>>>>  ifeq (,$(filter $(ARCH),arm64 x86_64))
> 
> Sorry, I don't fully follow.  If we expand the above "arm64 x86_64" to
> include "aarch64" rather than distill the latter into "arm64", how does
> that break the cross-builds?  Thanks,

Oh, it sounded like the suggestion was just (i.e. solely) adding
"aarch64" instead of the "arm64" as well.  If it includes arm64 (for
when ARCH= is already set) as well as aarch64 (from uname) then it's all
good.

-Matt

> 
> Alex
> 
>>>>  # Do nothing on unsupported architectures
>>>>
>>>> ---
>>>> base-commit: 96ca4caf9066f5ebd35b561a521af588a8eb0215
>>>> change-id: 20260317-vfio-selftests-aarch64-636abf0f7674
>>>>
>>>> Best regards,  
>>>   
>>
>>
>