Re: [PATCH 2/3] rbtree: update augmented data on the way down in rb_add_augmented_cached()

[Peter Zijlstra]

On Mon, Sep 28, 2026 at 04:05:13PM +0200, Peter Zijlstra wrote:
> On Mon, Sep 28, 2026 at 03:37:33PM +0200, Peter Zijlstra wrote:
> > On Mon, Sep 28, 2026 at 08:26:10PM +0800, Yiwei Lin wrote:
> > 
> > > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> > > index 7455a83a6a990..60db4624b9897 100644
> > > --- a/kernel/sched/fair.c
> > > +++ b/kernel/sched/fair.c
> > > @@ -1066,9 +1066,20 @@ static inline bool min_vruntime_update(struct sched_entity *se, bool exit)
> > >  	       se->max_slice == old_max_slice;
> > >  }
> > >  
> > > +/*
> > > + * Fold @new's subtree data into @se, for each @se on @new's insertion path.
> > > + */
> > > +static inline void
> > > +min_vruntime_merge(struct sched_entity *se, struct sched_entity *new)
> > > +{
> > > +	__min_vruntime_update(se, &new->run_node);
> > > +	__min_slice_update(se, &new->run_node);
> > > +	__max_slice_update(se, &new->run_node);
> > > +}
> > >  
> > 
> > So min_vruntime_update() can be written in terms of this helper like:
> > 
> > static inline bool min_vruntime_update(struct sched_entity *se, bool exit)
> > {
> > 	u64 old_min_vruntime = se->min_vruntime;
> > 	u64 old_min_slice = se->min_slice;
> > 	u64 old_max_slice = se->max_slice;
> > 	struct rb_node *node = &se->run_node;
> > 
> > 	se->min_vruntime = se->vruntime;
> > 	se->min_slice = se->slice;
> > 	se->max_slice = se->slice;
> > 
> > 	min_vruntime_merge(se, node->rb_right);
> > 	min_vruntime_merge(se, node->rb_left);
> > 
> > 	return se->min_vruntime == old_min_vruntime &&
> > 	       se->min_slice == old_min_slice &&
> > 	       se->max_slice == old_max_slice;
> > }
> > 
> > And that is *very* close to being generalizable, obviating the need for
> > RBCOMPUTE. Does your LLM see a way to make that happen?

Also, since you're poking at things here, can you please include a patch
to rename RB_DECLARE_CALLBACKS_MAX()'s RBCOMPUTE, it is totally
different from the earlier RBCOMPUTE. Them sharing a name is highly
confusing.

Perhaps rename to something like RBVALUE?