Re: [PATCH 7/8] dax/kmem: add sysfs interface for runtime hotplug state control

[Gregory Price <gourry@gourry.net>]

On Thu, Jan 22, 2026 at 11:49:48PM +0100, David Hildenbrand (Red Hat) wrote:
> 
> I'm merely wondering why, in the new world, you would even want the offline
> state.
> 
> So what are the use cases for that?
> 

I don't have one, and in the 5-patch series I killed it.  You are right,
it makes no sense.

However:

> Why would user space possibly want that? [plugged-in offline blocks]
> 

I don't think anyone does.

This is baggage.

The CXL driver auto-creates dax_kmem w/ offline memory blocks

Changing this behavior breaks existing systems :[

> Can't ndctl just use the old (existing) interface if such an operation is
> requested, and the new one (you want to add) when we want to do something
> reasonable (actually use system ram? :) ).

I think we're in agreement, I think I'm doing a poor job of explaining
the interconnected issues.

summarizing the long email:

   cxl/region + dax/cxl.c + dax/bus.c auto-probe baggage for
   BIOS-configured regions prevents any userland policy from
   from being plumbed from cxl to dax.  There's no interposition step.

So yes - new interfaces would resolve this and the old interfaces
could be left for compat.

~Gregory