[PATCH i-g-t] tests/intel/xe_pxp: Terminate PXP on exit

[Daniele Ceraolo Spurio]

On 4/24/2025 4:24 AM, Kamil Konieczny wrote:
> Hi Daniele,
> On 2025-04-18 at 09:53:42 -0700, Daniele Ceraolo Spurio wrote:
>>
>> On 4/18/2025 7:54 AM, Kamil Konieczny wrote:
>>> Hi Daniele,
>>> On 2025-04-17 at 16:57:19 -0700, Daniele Ceraolo Spurio wrote:
>>>> Most of out tests expect to start their execution from a clean GPU state.
>>>> However, the PXP tests currently leave a PXP session running when they
>>>> complete, so follow up tests will not have a clean starting state; given
>>>> that PXP can impact the timing of some operations (e.g. suspend/resume),
>>> What kind of problems could occur? Is it longer wakeup or refusing
>>> to enter sleep?
>> PXP can delay the suspend flow or, in some unlikely error cases, cause the
>> driver to abort the suspend attempt. AFAIU it can also delay idleness for
>> the media GT.
>>
> Thank you for explaining this, I spotted one nit, see below.
>
>>>> this could pollute the results of those tests. Therefore, let's make
>>>> sure that PXP is de-activated before we exit.
>>>>
>>>> Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
>>>> Cc: Alan Previn <alan.previn.teres.alexis at intel.com>
>>>> ---
>>>>    tests/intel/xe_pxp.c | 21 +++++++++++++++++----
>>>>    1 file changed, 17 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/tests/intel/xe_pxp.c b/tests/intel/xe_pxp.c
>>>> index 458d32322..3442396b2 100644
>>>> --- a/tests/intel/xe_pxp.c
>>>> +++ b/tests/intel/xe_pxp.c
>>>> @@ -1140,9 +1140,23 @@ static void termination_tests(int fd, bool pxp_supported, drmModeResPtr res,
>>>>    		}
>>>>    }
>>>> +int xe_fd = -1;
>>>> +
>>>> +static void exit_handler(int sig)
>>>> +{
>>>> +	/*
>>>> +	 * PXP can interact with some operations (e.g. suspend/resume), which
>>>> +	 * could impact the behavior of other tests if they unknowingly start
>>>> +	 * when PXP is already active. Therefore, let's make sure PXP is
>>>> +	 * de-activated before we exit.
>>>> +	 */
>>>> +	trigger_termination(xe_fd, PXP_TERMINATION_IRQ);
>>>> +	drm_close_driver(xe_fd);
>>>> +	usleep(50 * 1000); /* give time for the termination to be processed */
>>>> +}
>>>> +
>>>>    igt_main
>>>>    {
>>>> -	int xe_fd = -1;
>>>>    	bool pxp_supported = true;
>>>>    	drmModeResPtr res;
>>>> @@ -1151,6 +1165,7 @@ igt_main
>>>>    		igt_require(xe_has_engine_class(xe_fd, DRM_XE_ENGINE_CLASS_RENDER));
>>>>    		pxp_supported = is_pxp_hw_supported(xe_fd);
>>>>    		res = drmModeGetResources(xe_fd);
>>>> +		igt_install_exit_handler(exit_handler);
> Now I read it 2nd time and imho this should be done for pxp supported, so
>
>     		if (pxp_supported)
> 			igt_install_exit_handler(exit_handler);
>
>>>>    	}
>>>>    	igt_subtest_group {
>>>> @@ -1213,8 +1228,6 @@ igt_main
>>>>    		}
>>>>    	}
>>>> -	igt_fixture {
>>>> +	igt_fixture
>>>>    		drmModeFreeResources(res);
>>>> -		close(xe_fd);
> If we go with above change, fixture here should be:
> 	igt_fixture {
> 		drmModeFreeResources(res);
> 		if (!pxp_supported)
> 			drm_close_driver(xe_fd);
> 	}

Sure, I'll go ahead and to the changes.

Daniele

>
> Regards,
> Kamil
>
>>> Looks like you are fixing a bug here, it should be
>>> drm_close_driver(xe_fd);
>>>
>>> LGTM,
>>> Reviewed-by: Kamil Konieczny <kamil.konieczny at linux.intel.com>
>> Thanks!
>>
>> Daniele
>>
>>>
>>>> -	}
>>>>    }
>>>> -- 
>>>> 2.43.0
>>>>