Re: [PATCH] sched_ext: Document task ownership state machine

[Andrea Righi]

On Wed, Mar 04, 2026 at 06:23:57AM -1000, Tejun Heo wrote:
> Hello,
> 
> On Wed, Mar 04, 2026 at 04:33:43PM +0100, Andrea Righi wrote:
> > + * Task Ownership State Machine (sched_ext_entity->ops_state)
> >   *
> > + * The sched_ext core uses this state machine to track task ownership
> > + * between the SCX core and the BPF scheduler. This allows the BPF
> > + * scheduler to dispatch tasks without strict ordering requirements, while
> > + * the SCX core safely rejects invalid dispatches.
> >   *
> > + * State Transitions
> >   *
> > + *       .------------> NONE (owned by SCX core)
> > + *       |               |           ^
> > + *       |       enqueue |           | direct dispatch
> > + *       |               v           |
> > + *       |           QUEUEING -------'
> > + *       |               |
> > + *       |       enqueue |
> > + *       |     completes |
> > + *       |               v
> > + *       |            QUEUED (owned by BPF scheduler)
> > + *       |               |           ^
> > + *       |      dispatch |           | affinity changed
> > + *       |               |           | (dispatch aborts)
> > + *       |               v           |
> > + *       |          DISPATCHING -----'
> > + *       |               |
> > + *       |      dispatch |
> > + *       |     completes |
> > + *       `---------------'
> 
> I don't think we transition tasks from DISPATCHING to QUEUED on affinity
> changes. Not yet anyway, right?

Ah, you're right! That transition was part of my invalidation patch, so I
had it in head, but it's not actually in the kernel (yet). :)

I'll remove it.

Thanks,
-Andrea