uclient: fix null deref in unlink() on null dn
mon: fix offload_targets
uclient: Put some guards up around the Inode->cap map accesses
msg: handle MMDSLoadTargets
mon: Handle MMDSLoadTarget messages
mds/message: MMDSLoadTargets should use a set<int32_t> not a map
mds: Pull my_target checking out into a function targets_safe
mds: Don't do rebalancing until the target mds' are in the MDSMap.
This will break rebalancing until the monitor code to
add targets is in place.
messages: Add a completely useless and unused MMDSOffloadComplete mes…
mon: Put in placeholder code to handle MDS target messages
mds: Break apart rebalance function into prepwork and offload
messages: Add an MMDSLoadTargets message for mds->monitor info.
Message: add constants for MDS offload messages, and clean up the ali…
Check for invalid layout input in uclient and MDS, replace with sane …
libceph: Actually implement ceph_get_local_osd so that libhadoop works!
kclient: add missing include for in_pton
msgr: use getnameinfo to format socket addresses
use sockaddr_storage; and some ipv6 support groundwork
kclient: don't choke on missing snaprealm
kclient: no system.* xattrs, yet
kclient: kill generic_*xattr calls until we implement xattr_handler
kclient: clean up module startup
Merge branch 'unstable' of ceph.newdream.net:/git/ceph into unstable
kclient: fix ra_pages setup after bdi initiation
kclient: single thread wq for invalidate, truncate
Keep multiple threads for the writeback.
kclient: use KMEM_CACHE macro where possible
We can't for the inode cache bc we have a init_once method.
kclient: mark initialization functions __init
kclient: fix maintainers record
kclient: backport sync_page_range change
kclient: add to MAINTAINERS
kclient: sync_page_range gone in 2.6.32; use vfs_fsync_range
This is what generic_write_sync does, but we also want to do the
sync if we are NEARFULL.
kclient: current_is_pdflush() gone in 2.6.32