On Tue, May 28, 2024 at 11:17:58AM +0200, Christian Brauner wrote: > As I've said earlier, independent of the new handle type returning the > new mount id is useful and needed because it allows the caller to > reliably generate a mount fd for use with open_by_handle_at() via > statmount(). That won't be solved by a new handle type and is racy with > the old mount id. So I intend to accept a version of this patch. The whole point is that with the fsid in the handle we do not even need a mount fd for open_by_handle_at. If you insist on making this interface even more crappy than it is and create confusion please record my explicit: Nacked-by: Christoph Hellwig <hch@xxxxxx> in the commit :(
