Description of problem: Since kernel interface for thinp has been extended (and is not backward compatible) libdm and lvm needs to be updated to reflect these changes. New supported thin-pool flags in constructor: ignore_discard no_discard_passdown - currently with kernel 3.4-rc7 it has incompatible interface change, when device without discard support is being used. New supported thin flags in constructor: external origin dev Version-Release number of selected component (if applicable): 2.02.95 How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
We have now proposal patch: https://www.redhat.com/archives/dm-devel/2012-May/msg00161.html Which would make things backward compatible. But still we need to decide, what level of support lvm should provide here. 1. We could recalculate discard support for data device each time we manipulate with tables - and provide a proper option to the target constructor. Since I'm unsure in which moment of the tree activation we are able to query device for discard support it may require to activate data device before thin-pool device in separate udev transaction and obtain information about discard support on this device. 2. We could ignore it - user will get error message in system log - but since there is low chance many user will be changing their discs with discard support. In my case - I get to these problem since I use tmpfs as backend filesystem. Once this patch: http://www.spinics.net/lists/linux-fsdevel/msg55155.html goes upstream, it will eliminate use-case of discardless discs on my box. But in this case if some device in the stack has really flipped its discard support condition, user will need to reactivate device to get the change propagate.
Upstreamed with some more following commits: https://www.redhat.com/archives/lvm-devel/2012-August/msg00028.html
The fix is included in lvm2-2.02.98-1.fc19.