On Thu 26-10-23 22:44:26, Christoph Hellwig wrote: > On Wed, Oct 25, 2023 at 07:04:45PM +0200, Jan Kara wrote: > > Well, this is the discussion how btrfs should be presenting its subvolumes > > to VFS / userspace, isn't it? > > Yes. Which we've pressured to resolve forever, but it's been ignored. > > > I never dived into that too closely but as > > far as I remember it was discussed to death without finding an acceptable > > (to all parties) solution? I guess having a different fsid per subvolume > > makes sense (and we can't change that given it is like that forever even if > > we wanted). Having different subvolumes share one superblock is more > > disputable but there were reasons for that as well. So I'm not sure how you > > imagine to resolve this... > > We need to solve this out kernel wide, and right now the kernel doesn't > support different dev_t / fsids inside a single file syste at all. > SuSE hacks around that badly for limited user interfaces with the > horrible get_inode_dev method they've added, but this has been rejected > upstream for good reason. What this series does is to add another > limited version of this through the backdoor. OK, I see. I agree adding ->get_fsid is just piling on top of the problems so I can see why you don't like it. Band aids are double-edged sword ;) Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
