On 16/12/2008 Christian Jaeger wrote: > Christian Jaeger wrote: > >> and if the missing recursion of get_lvm_deps() is really > >> the reason, why does it only fail on some kernels for you? > >> > > > > As I did say in my previous mails, I don't know. > > > > And I don't know whether it's got anything to do with kernels. Or the > > order in which something is initialized at boot, or the phase of moon, > > or maybe whether I've got snapshots running. > > > > Hey, I've tested now with and without snapshots and it's clearly the > culprit. > > See the attached observation log. > > You can see (1) that while there is a snapshot on the root volume, on my > setup, the current cryptsetup from Debian is broken; it is *not* broken > if I remove the snapshot. (2) I see confirmed that my patch works for me > in both cases. > > So, I hope now you've got at least something to reproduce the case.
Ok, I spend some hours on debugging the issue and finally was able to reproduce it. Yes, it's correct that creating a lv snapshot seems to add another level of mapping. the device main-root no longer directly depends on the physical volume, but instead depends on main-root-real, and that one finally depends on the physical volume. I did some tests with a recursive get_lvm_deps(), and it seems to work as expected. Thanks for your great debugging work, Christian. I wouldn't have been able to track down this bug that soon without your invaluable help. Same goes to Ben Hutchings and Yves-Alexis Perez. You rock! I've just prepared cryptsetup 1.0.6-7 with this bug and several others fixed. Would you mind testing the packages before I actually upload them to unstable and ask for inclusion in lenny? Many changes since 1.0.6-6 are documentation improvements, but there are also some code fixes in initscripts, initramfs scripts and keyscripts. you can find the packages at http://people.debian.org/~mejo/cryptsetup/ greetings, jonas
signature.asc
Description: Digital signature
