[Intel-xe] ✗ CI.BAT: failure for Add OA functionality to Xe (rev5)

[Dixit, Ashutosh]

On Thu, 07 Sep 2023 22:10:17 -0700, Patchwork wrote:
>
> Patch Details
>
>  Series:  Add OA functionality to Xe (rev5)
>  URL:  https://patchwork.freedesktop.org/series/121084/
>  State:  failure
>  Details:  https://intel-gfx-ci.01.org/tree/intel-xe/xe-pw-121084v5/index.html
>
> CI Bug Log - changes from xe-362-758f01ec2a6774fda1e8f8075d31f42b81783caf_BAT -> xe-pw-121084v5_BAT
>
> Summary
>
> FAILURE
>
> Serious unknown changes coming with xe-pw-121084v5_BAT absolutely need to be
> verified manually.
>
> If you think the reported changes have nothing to do with the changes
> introduced in xe-pw-121084v5_BAT, please notify your bug team (lgci.bug.filing at intel.com) to allow them
> to document this new failure mode, which will reduce false positives in CI.
>
> Participating hosts (4 -> 4)
>
> No changes in participating hosts
>
> Possible new issues
>
> Here are the unknown changes that may have been introduced in xe-pw-121084v5_BAT:
>
> IGT changes
>
> Possible regressions
>
> * igt at kms_addfb_basic@invalid-set-prop:
>
>  * bat-adlp-7: PASS -> INCOMPLETE
>
> * igt at xe_exec_balancer@twice-cm-virtual-rebind:
>
>  * bat-adlp-7: PASS -> FAIL +43 other tests fail
>
> * igt at xe_exec_balancer@twice-virtual-userptr-invalidate:
>
>  * bat-dg2-oem2: PASS -> FAIL +38 other tests fail
>
> * igt at xe_exec_compute_mode@twice-userptr-invalidate:
>
>  * bat-atsm-2: PASS -> FAIL +38 other tests fail
>
> * igt at xe_exec_threads@threads-mixed-basic:
>
>  * bat-dg2-oem2: PASS -> TIMEOUT +3 other tests timeout
>
>  * bat-atsm-2: PASS -> TIMEOUT +2 other tests timeout
>
> * igt at xe_exec_threads@threads-mixed-shared-vm-userptr-invalidate-race:
>
>  * bat-adlp-7: PASS -> TIMEOUT +5 other tests timeout
>
>  * bat-dg2-oem2: PASS -> INCOMPLETE
>
> * igt at xe_exec_threads@threads-mixed-userptr-invalidate:
>
>  * bat-atsm-2: PASS -> INCOMPLETE

These failures are due to the fact that the kernel patch has added an
additional field to 'struct drm_xe_engine_class_instance':

	drm/xe/uapi: Introduce OA (observability architecture) uapi

	@@ -748,6 +756,7 @@ struct drm_xe_engine_class_instance {

	        __u16 engine_instance;
	        __u16 gt_id;
	+       __u16 oa_unit_id;
	 };

The tests pass with my IGT version but there's no way to fix these in CI
since XE CI does not have a "Test-with". Let's see, we may add some padding
to the struct first and merge that and then replace that padding with the
field above.

This field will not be needed if we go with 'struct drm_xe_query_oa_info'
(which is introduced in xe_drm.h) but not implemented yet.

Thanks.
--
Ashutosh