Re: [i-g-t] tests/core_hotunplug: Open render capable DRM device

["Randhawa, Jagmeet" <jagmeet.randhawa@intel.com>]

On 3/4/2026 8:31 AM, Janusz Krzysztofik wrote:
> Hi,
>
> On Wednesday, 4 March 2026 14:56:01 CET Kamil Konieczny wrote:
>> Hi Jagmeet,
>> On 2026-03-04 at 06:53:32 +0800, Jagmeet Randhawa wrote:
>>> The test may select simpledrm when opening DRIVER_ANY,
>>> which does not expose a render node. This causes the
>>> later render open to fail.
>>>
>>> Open a render capable device initially using
>>> __drm_open_driver_render(DRIVER_ANY) to avoid
>>> selecting simpledrm.
>> This is strange, as this test assumes card open and then tries
>> uses render for verification so I am not sure if you could
>> start with render. Adding Janusz to Cc.
> My design didn't expect a missing render node, and that was a mistake.
> In real world, devices with only master or only render node can exist.
> If we want to keep the test hardware agnostic, as it was designed, then
> it should probe both node types for their existence on start, store that
> information, and then use it while running to skip those processing paths
> that are not applicable.
>
> Thanks,
> Janusz


Hello Janusz,

Thank you for the clarification, I understand your point that in general 
a device

may expose only a card or only a render node. In this particular test 
file however

all subtests require a render node for their checks, so a render-capable 
device is effectively

a prerequisite for the test to run. Selecting a render-capable device 
upfront guarantees the render

node required by test while still providing corresponding card node used 
by test logic, and

also avoids selecting card-only drivers such as simpledrm.

>
>> Could you also add links to issues on GitLab, if there are any?
>>
>> Other idea to solve this would be to add simpledrm to known
>> drivers and a blocked ones, like VGEM.
>>
>> Regards,
>> Kamil
>>
>> Cc: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com>
>> Cc: "Zbigniew Kempczyński" <zbigniew.kempczynski@intel.com>
>>
>>> Signed-off-by: Jagmeet Randhawa <jagmeet.randhawa@intel.com>
>>> ---
>>>   tests/core_hotunplug.c | 4 ++--
>>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/tests/core_hotunplug.c b/tests/core_hotunplug.c
>>> index 7c9dae1bf..3a612fce2 100644
>>> --- a/tests/core_hotunplug.c
>>> +++ b/tests/core_hotunplug.c
>>> @@ -696,8 +696,8 @@ int igt_main()
>>>   		int fd_drm;
>>>   		char driver[32] = {};
>>>   
>>> -		fd_drm = __drm_open_driver(DRIVER_ANY);
>>> -		igt_skip_on_f(fd_drm < 0, "No known DRM device found\n");
>>> +		fd_drm = __drm_open_driver_render(DRIVER_ANY);
>>> +		igt_skip_on_f(fd_drm < 0, "No known DRM render device found\n");
>>>   
>>>   		priv.chipset = drm_get_chipset(fd_drm);
>>>   		/* Make sure it opens the same driver */
>
>
>