[PATCH 7/8] fs/configfs: switch to get_tree_keyed()

[Christian Brauner]

On Sat, Jun 13, 2026 at 01:14:36PM +0200, Hannes Reinecke wrote:
> Switch to get_tree_keyed() to instatiate a new superblock whenever
> the mount namespace changed.
> 
> Signed-off-by: Hannes Reinecke <hare at kernel.org>
> ---

So I wonder what keying superblock creation to the mount namespace
really gets you. Imho keying the superblock by namespace only makes
sense if the whole filesystem is tied to a namespace. For example, the
mqueue filesystem is semantically tied to ipc namespaces so tying the
creation of a new mqueue superblock to the ipc namespace makes sense.

But configs isn't semantically tied to the mount namespace at all. It
encompasses a bunch of different subsystems that might themselves be
semantically tied to different namespaces. The specific example you care
about is nvme tied to network namespaces. But there might be other
things that tie to other namespaces.

So even if you tie it to the mount namespace you presumably still need
to filter what subsystems a specific configfs instance will display to
each accessing task based on the namespaces of the accessing task?