On Mon, Jul 11, 2016 at 04:07:25PM +0200, Benedikt Morbach wrote:

Hi David,

On Fri, Jun 17, 2016 at 7:18 PM, David Sterba [off-list ref] wrote:

quoted

thanks for the fix. The mail thread holds enough information to
reproduce the issue, I'll get to that on Monday. I'm not sure about the
use of strstr to match the subvolume name so I want to write the tests
first and craft some paths to see how it works.

any update on this? I've been running this patch for a few weeks now
with my hourly backups.
I could also take a stab at writing a test case but I think I'd need
some pointers for that. ;-)

Sorry, I got distracted. The things with paths are a bit complicated,
there's chroot support, explicit mountpoint (-m), so stripping the
root_path would not work in all cases.

Apparently the full_clone_path is wrong there as it's not relative to
root_path and thus openat(mnt_fd, ...) will fail, so instead of
stripping the path, we would need to provide the clone path relative to
root_path directly.

I took a look at the send/receive patches that Filipe sent earlier and
afaict this isn't fixed in those.

No, this is not related to kernel.