On 3 Oct 2024, at 11:14, Alexander Aring wrote:
> Since commit 5f81880d5204 ("sysfs, kobject: allow creating kobject
> belonging to arbitrary users") it seems that there could be cases for
> kobjects belonging to arbitrary users. This callback is set by default
> when using kset_create_and_add() to allow creating kobjects with
> different ownerships according to its parent.
>
> This patch will assign the default callback now for nfs kobjects for
> cases when the parent has different ownership than the default one.
>
> Signed-off-by: Alexander Aring <aahringo@xxxxxxxxxx>
> ---
> fs/nfs/sysfs.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/fs/nfs/sysfs.c b/fs/nfs/sysfs.c
> index a6584203b7ff..b5737464b892 100644
> --- a/fs/nfs/sysfs.c
> +++ b/fs/nfs/sysfs.c
> @@ -27,6 +27,7 @@ static const struct kobj_ns_type_operations *nfs_netns_object_child_ns_type(
> }
>
> static struct kobj_type nfs_kset_type = {
> + .get_ownership = kset_get_ownership,
> .release = kset_release,
> .sysfs_ops = &kobj_sysfs_ops,
> .child_ns_type = nfs_netns_object_child_ns_type,
> --
> 2.43.0
Hi Alex, if I understand this correctly, this patch just punts the ownership
callback up to fs_kobj, which, because it has no .get_ownership is just
going to be the same result: root.
Does this patch add value?
Ben
