Re: [PATCH sched_ext/for-7.2] sched_ext: Refactor scx_root_enable_workfn() enablement task migration

Re: [PATCH sched_ext/for-7.2] sched_ext: Refactor scx_root_enable_workfn() enablement task migration

[Tejun Heo]

Hello,

Thanks for the patch, but I don't think this makes the code better.
scx_root_enable_workfn() is a linear setup sequence; length comes
from the number of ordered steps, not from doing unrelated things.
Each extracted helper has exactly one caller, so the split is pure
code motion. The locking story and the error path both fragment
across three functions, and small things drift - the TID_TO_TASK
read moved from ops->flags to sch->ops.flags, which is equal today
but sch->ops.flags is mutated elsewhere (SCX_OPS_HAS_CPU_PREEMPT),
so the two sources aren't guaranteed to stay in sync.

Thanks.

--
tejun

Re: [PATCH sched_ext/for-7.2] sched_ext: Refactor scx_root_enable_workfn() enablement task migration

[Cheng-Yang Chou]

Hi Tejun,

On Tue, Apr 21, 2026 at 07:30:18AM -1000, Tejun Heo wrote:
> Thanks for the patch, but I don't think this makes the code better.
> scx_root_enable_workfn() is a linear setup sequence; length comes
> from the number of ordered steps, not from doing unrelated things.
> Each extracted helper has exactly one caller, so the split is pure
> code motion. The locking story and the error path both fragment
> across three functions, and small things drift - the TID_TO_TASK
> read moved from ops->flags to sch->ops.flags, which is equal today
> but sch->ops.flags is mutated elsewhere (SCX_OPS_HAS_CPU_PREEMPT),
> so the two sources aren't guaranteed to stay in sync.

I see your point about maintaining the linear sequence for better
auditability of the locking and error paths. You're also right about the
TID_TO_TASK drift - checking the immutable ops->flags is indeed the correct
approach.

Regarding the ordering of the task enablement sequence: do you think
it would be useful to be included in the kdoc?

Thanks for your time.

-- 
Cheers,
Cheng-Yang

Re: [PATCH sched_ext/for-7.2] sched_ext: Refactor scx_root_enable_workfn() enablement task migration

[Tejun Heo]

On Wed, Apr 22, 2026 at 09:53:24AM +0800, Cheng-Yang Chou wrote:
> Regarding the ordering of the task enablement sequence: do you think
> it would be useful to be included in the kdoc?

Who'd be the audience tho? This isn't a level of detail that'd be
interesting to people who want to use SCX. For kernel devs, I don't think a
doc which just restates the code sequence is all that useful when you can
just throw an AI agent at it and get the same and more in no time.

-- 
tejun