[PATCH v5 4/9] drm: selftest: convert drm_format selftest to KUnit

Maíra Canal maira.canal at usp.br
Sun Jul 17 17:48:04 UTC 2022


On 7/14/22 21:03, Daniel Latypov wrote:
> On Thu, Jul 14, 2022 at 4:51 PM Guenter Roeck <linux at roeck-us.net> wrote:
>>
>> On Fri, Jul 08, 2022 at 05:30:47PM -0300, Maíra Canal wrote:
>>> Considering the current adoption of the KUnit framework, convert the
>>> DRM format selftest to the KUnit API.
>>>
>>> Tested-by: David Gow <davidgow at google.com>
>>> Acked-by: Daniel Latypov <dlatypov at google.com>
>>> Reviewed-by: Javier Martinez Canillas <javierm at redhat.com>
>>> Signed-off-by: Maíra Canal <maira.canal at usp.br>
>>
>> This patch results in:
>>
>> Building powerpc:allmodconfig ... failed
>> --------------
>> Error log:
>> drivers/gpu/drm/tests/drm_format_test.c: In function 'igt_check_drm_format_min_pitch':
>> drivers/gpu/drm/tests/drm_format_test.c:271:1: error: the frame size of 3712 bytes is larger than 2048 bytes
>>
>> presumably due to function nesting.
> 
> This can happen when there's a lot of KUNIT_EXPECT_* calls in a single function.
> See [1] for some related context.
> There were a number of patches that went into 5.18 ([2] and others) to
> try and mitigate this, but it's not always enough.
> 
> Ideally the compiler would see that the stack-local variables used in
> these macros don't need to stick around, but it doesn't always
> happen...

As a matter of fact, for GCC 12, the warning -Wframe-larger-than=
doesn't show up due to compiler improvement, but for GCC 11.3, it does.

As I have GCC 12 on my machine, I didn't even get the warning.

Anyway, I'll separate the test into multiple functions to avoid any
problems related to stack size.

Best Regards,
- Maíra Canal

> One workaround would be to split up the test case functions into smaller chunks.
> 
> [1] https://lore.kernel.org/linux-kselftest/20210929212713.1213476-1-brendanhiggins@google.com/
> [2] https://lore.kernel.org/linux-kselftest/20220113165931.451305-1-dlatypov@google.com/
> 
> Daniel


More information about the dri-devel mailing list