On 2018-03-30 12:38, Adam Borowski wrote:
On Fri, Mar 30, 2018 at 10:42:10AM +0100, Pete wrote:
I've just notice work going on to make rmdir be able to delete
subvolumes. Is there an intent to allow ls -l to display directories as
subvolumes?
That's entirely up to coreutils guys.
Expanding on Adam's 100% accurate answer:
The primary issue here is that determining what is a subvolume requires
some non-trivial work. You either need to call a specific ioctl (and
make sure you call it in the right location) which last I checked
requires root privileges to use, or do some significant guess-work that
happens to fall apart once you throw explicitly mounted subvolumes into
the mix.
You can take a look at the `get_subvol_list` function in [1] for an
example of what that guess-work looks like in Python. That particular
function relies on some specific internals of BTRFS that really aren't
technically part of the ABI (namely that subvolumes all have an inode
number of 256), but it falls apart if the subvolumes are explicitly
mounted (because if you don't exclude explicit mounts, you may
accidentally recurse into another BTRFS volume).
There's also a secondary issue in that the concept of a subvolume isn't
really universal. BTRFS and ZFS are the only filesystems supported on
Linux that have such a concept, but ZFS datasets are a very different
implementation of the concept that doesn't really make sense to mark
with `ls`. Given the level of specificity, it just doesn't make sense
to mark subvolumes as such in `ls`, just like it doesn't make much sense
to treat NILFS2 checkpoints any differently from read-only directories.
[1]
https://raw.githubusercontent.com/Ferroin/btrfs-subv-backup/master/btrfs-subv-backup.py
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
