Re: [PATCH v3 0/5] Defer throttle when task exits to user

[Valentin Schneider]

On 15/07/25 15:16, Aaron Lu wrote:
> There are consequences because of this new throttle model, e.g. for a
> cfs_rq that has 3 tasks attached, when 2 tasks are throttled on their
> return2user path, one task still running in kernel mode, this cfs_rq is
> in a partial throttled state:
> - Should its pelt clock be frozen?
> - Should this state be accounted into throttled_time?
>
> For pelt clock, I chose to keep the current behavior to freeze it on
> cfs_rq's throttle time. The assumption is that tasks running in kernel
> mode should not last too long, freezing the cfs_rq's pelt clock can keep
> its load and its corresponding sched_entity's weight. Hopefully, this can
> result in a stable situation for the remaining running tasks to quickly
> finish their jobs in kernel mode.

OK, I finally got to testing the PELT side of things :-)

I shoved a bunch of periodic tasks in a CPU cgroup with quite low limits
(1ms runtime, 10ms period); I looked at the _avg values using the
trace_pelt* tracepoints.

Overall there isn't much change to the averages themselves. There are more
updates since the tasks are genuinely dequeued/enqueued during a throttle
cycle, but that's expected.

I'll wait for your next version, but you can have:

Tested-by: Valentin Schneider <vschneid@xxxxxxxxxx>