The basic idea with this set is to allow the kernel client to allow the client to do more d_revalidate functionality under the rcu_read_lock. The idea is basically to do what we can there while not blocking, and then return -ECHILD for the rest. The only really questionable bit is the part where we need to take the d_lock in ceph_d_release. I think it should be largely uncontended, but it's still less than ideal. Maybe there is some way to handle that locklessly? Also, it would be nice to not have to drop out of rcuwalk mode in order to run the lease renewal. It seems like the kind of thing that would be ideal to queue to a workqueue, but I don't see a way to do that without introducing a potential deadlock. My teuthology run with it passed all but one test, which seems to have been an unrelated problem (though I could be wrong there). Jeff Layton (4): ceph: remove ceph_mdsc_lease_release ceph: clear d_fsinfo pointer under d_lock ceph: allow dentry_lease_is_valid to work under RCU walk ceph: handle LOOKUP_RCU in ceph_d_revalidate fs/ceph/dir.c | 67 +++++++++++++++++++++++++++++++++++----------------- fs/ceph/mds_client.c | 41 -------------------------------- fs/ceph/mds_client.h | 4 ---- 3 files changed, 45 insertions(+), 67 deletions(-) -- 2.5.5 -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
