On Wed, Nov 08, 2023 at 10:55:52PM -0800, Christoph Hellwig wrote: > On Wed, Nov 08, 2023 at 05:20:06PM +0100, Christian Brauner wrote: > > This would also allow tools that want to to detect when they're crossing > > into a new subvolume - be it on btrfs or bcachefs - and take appropriate > > measures deciding what they want to do just relying on statx() without > > any additional system calls. > > How? If they want to only rely on Posix and not just he historical > unix/linux behavior they need to compare st_dev for the inode and it's > parent to see if it the Posix concept of a mount point (not to be > confused with the Linux concept of a mountpoint apparently) because > that allows the file system to use a new inode number namespace. That doesn't work anymore. Both overlayfs and btrfs make this impossible or at least inconsistent. Tools that want to rely on that definition can continue to do so and really just ignore any of the new features. But tools that want to know about this and adjust behavior can really benefit from this. Just marking an inode as a subvolume root is worth it without committing to any filesystem specifics now that we have two of them.
