[PATCH 0/2] Fixes for MI_REPORT_PERF_COUNT

[Umesh Nerlige Ramappa]

On Tue, Dec 17, 2024 at 09:13:12AM -0800, Souza, Jose wrote:
>On Tue, 2024-12-17 at 17:05 +0000, Souza, Jose wrote:
>> On Mon, 2024-12-16 at 16:58 -0800, Umesh Nerlige Ramappa wrote:
>> > OA programming sequence for query mode or MI_REPORT_PERF_COUNT requires
>> > modifying some HW registers in the same hw context as the user exec
>> > queue. User passes the exec_queue to the OA interface and OA
>> > implementation submits an MI_LOAD_REGISTER_IMM to this queue to modify
>> > the registers.
>> >
>> > The OA implementation submits a batch mapped in GGTT to the user exec
>> > queue and hence, some plumbing is added into relevant code to enable
>> > that (as per suggestions from Matthew Brost).
>> >
>> > v2: review rework
>>
>> This series fixes GPU hang happening when close OA stream.
>>
>> Tested-by: José Roberto de Souza <jose.souza at intel.com>
>
>Ah can you add a Fixes tag to 8135f1c09dd2eecee7cb637f7ec9a29e57300eb8 ?
>Would be nice to get this patch backported to older kernel releases.

Sure, will do. Thanks for testing it out.

Umesh
>
>>
>> >
>> > Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>
>> >
>> > Umesh Nerlige Ramappa (2):
>> >   xe: Allow a GGTT mapped batch to be submitted to user exec queue
>> >   xe/oa: Use MI_LOAD_REGISTER_IMMEDIATE to enable OAR/OAC
>> >
>> >  drivers/gpu/drm/xe/xe_oa.c              | 71 ++++++++++++++++++-------
>> >  drivers/gpu/drm/xe/xe_ring_ops.c        |  5 +-
>> >  drivers/gpu/drm/xe/xe_sched_job_types.h |  2 +
>> >  3 files changed, 57 insertions(+), 21 deletions(-)
>> >
>>
>