Hi Nick,
Any progress of fixing the d_parent behavior that's breaking ceph and
sillyrename? Just a reminder: ceph currently oopses on d_kill, and
sillyrename entries don't go away. Your message from 3 weeks ago
suggested the fix was relatively simple and imminent, but it's getting
late in the cycle and I'm getting nervous :).
We do have an ugly hack[1] that will fix the ceph problem (by making a
copy of d_parent in the d_fsdata struct), but I'd obviously rather not
push that.
FWIW the d_parent clearing in d_kill was introduced by 949854d0.
Thanks!
sage
[1] http://ceph.newdream.net/git/?p=ceph-client.git;a=commitdiff;h=9c3db357aa3de6fd5e3ef2865e20848251aa05a4