Re: [PATCH 1/2] mm,memory_hotplug: Implement numa node notifier

[David Hildenbrand]

On 04.04.25 14:56, Oscar Salvador wrote:
> On Fri, Apr 04, 2025 at 12:09:39PM +0200, David Hildenbrand wrote:
> > Assuming we can remove the _normal stuff and we can do what we do in patch
> > #2 here already ... meaning we unconditionally store the nid in the MEM
> > notifier ...
> >
> > What about extending the existing memory notifier instead?
> >
> > That is, we add
> >
> > MEM_NODE_BECOMING_MEM_AWARE ... and trigger it using the same notifier
> > chain. We only have to make sure these new events will be properly filtered
> > out (IIRC, for most we do that already).
> >
> > Of course, the range will not apply to these events, but the nid would apply
> > to all.
>
> But that would defeat the purpose of having a diferent notifier for
> those only interested in node changes, which is not having to bother at
> all with unrelated notifications.
> > Yes, the handling would be simpler than it is now, but honestly I 
still see value
> in having them both decoupled from eacher other, I think it is cleaner
> and expresses in a more clear way the change of what the consumer is interested in
> get notified for.

No strong opinion, just a thought.

> Unless there is a strong objection, I would pursue that path, getting
> rid of the _normal stuff for slub along the way.

Yeah, getting rid of _normal and just have a "nid" will make things a 
lot cleaner.

--
Cheers,

David / dhildenb