Re: Issue: With an OverlayFS that has Btrfs as the upper layer, removal of directories silently fails

From: Miklos Szeredi
Date: Mon Oct 12 2015 - 11:38:00 EST


On Fri, Aug 21, 2015 at 9:23 PM, Neal Gompa <ngompa@xxxxxxxxxxxxxxxxx> wrote:
> Hello,
>
> For those who have received this message twice already, I sincerely apologize.
>
> Yesterday I filed an issue in the CentOS bug tracker (#9297[1]) and
> the Red Hat Bugzilla (#1255512[2]) about OverlayFS with Btrfs as the
> upper layer.
>
> The issue was discovered by a colleague of mine, and we verified the
> issue exists in the EL7 kernel and mainline kernels (tested on Arch
> Linux). I additionally verified the problem exists in Fedora kernels.
>
> Rather than quoting the issue, I'll just note that the issue is
> described quite well in the filed bugs noted earlier. In the linked
> bugs, there's a simple Bash script that will reliably reproduce the
> problem. While the script uses tmpfs at the lower layer, we originally
> discovered the problem with ext4 as the lower layer and verified that
> it doesn't matter which lower layer filesystem it is, the problem
> exists. Additionally, the problem doesn't exist when using ext4 as the
> upper layer.
>
> I was advised by Josh Boyer to email you guys (and the mailing lists,
> to be sure to get everyone a chance to look over it) about the
> problem, since it affects mainline.
>
> [1]: https://bugs.centos.org/view.php?id=9297
> [2]: https://bugzilla.redhat.com/show_bug.cgi?id=1255512

AFAICS btrfs doesn't support RENAME_EXCHANGE and RENAME_WHITEOUT.
These flags are needed for any filesystem that wants to be a fully
functional overlayfs upper layer.

Thanks,
Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/