Re: [RFC 15/16] sched/fair: Account kthread runtime debt for CFS bandwidth

[Tejun Heo]

Hello,

On Fri, Jan 14, 2022 at 10:40:06AM +0100, Peter Zijlstra wrote:
> You could do a special case sched_move_task(), that takes a css argument
> instead of using the current task_css. Then for cgroups it looks like
> nothing changes, but the scheduler will DTRT and act like it is in the
> target cgroup. Then at the end, simply move it back to task_css.
> 
> This obviously doesn't work for SoftIRQ accounting, but that is
> 'special' anyway. Softirq stuff is not otherwise under scheduler
> control and has preemption disabled.

So, if this particular use case doesn't fit the backcharge model (I'm not
sure yet). I'd much prefer it to maintain dynamic per-cgroup helper threads
than move tasks around dynamically. Nothing else is using migration this way
and we don't even need migration for seeding cgroups w/ CLONE_INTO_CGROUP.
In the future, this should allow further optimizations and likely
simplifications. It'd suck to have an odd exception usage.

Thanks.

-- 
tejun