kernel client
- setattr should check lease before believing something is a no-op
-- use list_for_each_safe for caps removal?
- - revisit cap removal locking, make sure it's okay....
- trim expired leases so we don't indefinitely hold dcache refs...
- carry wrbuffer/rdcache caps until data is flushed
- this should make the utimes bit kick in
-- make sure link/unlink results reflected by inode/dentry cache (let fill_trace do it? invalidate? do actual update?)
- procfs/debugfs
- adjust granular debug levels too
- should we be using debugfs?
- a dir for each client instance (client###)?
- hooks to get mds, osd, monmap epoch #s
-- msgr layer
- - standby state, to keep connect_seq around after a disconnect
-- audit locking
-- audit use of kmalloc vs spinlocks
- convert rest of ceph_fs.h to le32/le64 notation, cleaning up userspace appropriately
- vfs
- can we use dentry_path(), if it gets merged into mainline?
-- mds client
- - test ceph_fill_trace when files/directories are moved around by another client
- osd client
- readpages (readahead)
- async (caching) mode