On Fri, Feb 16, 2007 at 05:31:34PM -0800, Steve Langasek wrote: > I'm told that dmcrypt+XFS has never worked in the upstream kernel or in > Debian, so this is essentially an unsupported configuration. But you've > filed this bug as critical with the justification that it "causes serious > data loss". Did you lose data as a result of this bug? Could you explain > the process by which that happened? It's my impression that this > combination is so unreliable that it will oops before you really have a > chance to try to use it for storing data, so you can't really lose any data > if you can't put it there in the first place.
Yes, I did lose data (though fortunately only test data). But you are
correct in that this happened quite early, before I really had a
chance to start to use it seriously.
I copied data to an XFS filesystem in a crypted partition, and I think
I got a few gigabytes copied before the oops. After that I googled and
figured out the rest, I didn't test if there really was some data loss
(didn't ever mount the partition again).
> Based on the status as a known-buggy and unsupported config I think this bug
> should be downgraded to non-RC status for etch, but I'd like to be sure
> first that I understand the impact of any real-world risk of data loss.
No objection to that. Though it of course would have been be very
helpful if something somewhere (eg. the kernel at mount time) notified
me that the config is unsupported.
Sami
signature.asc
Description: Digital signature
