Re: [RFC PATCH 1/3] mm/numa: add exclusive node pool and numa=standby boot parameter

[Mike Rapoport <rppt@kernel.org>]

On Thu, Jun 11, 2026 at 10:04:01AM -0400, Gregory Price wrote:
> On Thu, Jun 11, 2026 at 12:00:17PM +0300, Mike Rapoport wrote:
> > > 1) Can we do dynamic addition of nodes?
> > > 
> > >    Not Trivially
> > > 
> > >    Some services utilize num_possible_nodes() as a static value to
> > >    calculate the amount of resources to use at runtime (bpf, md/raid5).
> > > 
> > >    Example: futex_init uses num_possible_nodes() as part of its
> > >             hashsize calculation during __init.
> > 
> > AFAIU, we don't add the additional nodes for generic hotplug memory but
> > rather for exclusive use of by drivers/applications that are aware of these
> > nodes.
> 
> The intent is to use for "non-generic" hotplug (see the whole private
> node series [1]), which would eventually still use the hotplug mechanism
> just not for generic memory.
> 
> [1] https://lore.kernel.org/linux-mm/20260222084842.1824063-1-gourry@gourry.net/
> 
> > Wouldn't adding them to possible nodes actually skew the calculation of the
> > resources by the services utilizing num_possible_nodes()?
> > 
> > With the futex_init() example, won't be hashsize scaled down two much
> > because we've added these special nodes to the possible mask?
> >
> 
> The result is the same as BIOS reserving nodes with PXM entries that
> don't get used.  The CXL ACPI Tables do this for CXL Fixed Memory
> Windows that may never be hotplugged.

Well, even without CXL there could be nodes that may never be hotplugged, I
suppose that with CXL the difference between available and possible
nodes can get much larger.
 
> So really i think you're pointing out that futex_init() here probably
> shouldn't be using num_possible_nodes?

I'd rather say that num_possible_nodes() with and without CXL (or other
differentiated memory) has different semantics.
Maybe we need to add a new primitive for possible differentiated nodes and
keep num_possible_nodes() to mean "number of possible nodes with normal
memory".
 
> ~Gregory

-- 
Sincerely yours,
Mike.