Heh ... following the discussion on the IRC, I actually prefer a solution that will push tests that are not filesystm-agnostic to the filesystem-specific .checkpoint method. Then also put it in the generic .checkpoint method to catch the "default" case. I think it's cleaner and does not introduce needless noise in the kernel. FWIW, not only for the d_unlinked() test, but also for the locks issue; even though locks are supposed to be VFS generic and not depend on specific filesystem, apparently they do. (Either that or fix find_locks_with_owner() in a generic way). Oren. Dan Smith wrote: > OL> Comments ? > > Works for me (although the issue remains that find_locks_with_owner() > thinks there are locks present on a socket). > _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers