Hi all,

Patch 3 fixed the issue with handling tasks with prio set to 0 during
the execution of blktests.

Kindly review.

Best Regards,
Xavier

Xavier (2):
  RT SCHED: Optimize the enqueue and dequeue operations for rt_se
  RT test: Adding test cases for RT group scheduling

 MAINTAINERS                                   |   7 +
 kernel/sched/debug.c                          |  48 +++
 kernel/sched/rt.c                             | 287 +++++++++++++++---
 kernel/sched/sched.h                          |   1 +
 tools/testing/selftests/sched/Makefile        |   4 +-
 tools/testing/selftests/sched/deadloop.c      | 192 ++++++++++++
 .../selftests/sched/rt_group_sched_test.sh    | 119 ++++++++
 7 files changed, 618 insertions(+), 40 deletions(-)
 create mode 100644 tools/testing/selftests/sched/deadloop.c
 create mode 100755 tools/testing/selftests/sched/rt_group_sched_test.sh

-- 
2.45.2

On Tue, Jul 16, 2024 at 01:25:41PM +0800, Xavier wrote:
> Hi all,
> 
> Patch 3 fixed the issue with handling tasks with prio set to 0 during
> the execution of blktests.

*sigh*... are you actually using this horror show?

The plan was to scrap this code -- and replace it with something based
on deadline servers. Sadly not a lot of people are able to work on that
:/

Hi Peter,

Your meaning is, will the RT scheduling policy be removed? I see that the
current kernel already includes the deadline and EEVDF scheduling policies.
I noticed that the enqueue and dequeue operations for RT are quite
complicated, so I proposed this patch for optimization.



At 2024-07-16 16:59:26, "Peter Zijlstra" <peterz@infradead.org> wrote:
>On Tue, Jul 16, 2024 at 01:25:41PM +0800, Xavier wrote:
>> Hi all,
>> 
>> Patch 3 fixed the issue with handling tasks with prio set to 0 during
>> the execution of blktests.
>
>*sigh*... are you actually using this horror show?
>
>The plan was to scrap this code -- and replace it with something based
>on deadline servers. Sadly not a lot of people are able to work on that
>:/

On Tue, Jul 16, 2024 at 05:17:50PM +0800, Xavier wrote:
> 
> Hi Peter,
> 
> Your meaning is, will the RT scheduling policy be removed? I see that the
> current kernel already includes the deadline and EEVDF scheduling policies.
> I noticed that the enqueue and dequeue operations for RT are quite
> complicated, so I proposed this patch for optimization.

No, not the policy, but we would really like to completely redo the
group scheduling part for RT.

For UP it is trivial, but we're not completely sure how to best do it
for SMP yet.

Anyway, I had gotten the impression that very few people indeed were
using the RT group scheduling feature, so I was surprised so see your
patches.