v10:
- collect r-bs
- comment/renice the function monitor_get_clock(), add some commit
  message [Stefan, Markus]

v9:
- two more patches to implement Markus's idea to init monitor earlier
  (which are patch 5 & 6)
- touch up patch 7 to init the fdset lock in monitor_init_globals()

v8:
- some wording changes according to previous comments [Markus]
- return -ENOENT too in stubs/fdset.c:monitor_fdset_get_fd() [Stefan]
- refactor the fdset functions a bit, drop "ret" where proper [Markus]
- one more patch to fix monitor_lock comment [Markus]
- regular rebase and torturing

Stefan reported this problem that in the future we might start to have
more threads operating on the same Monitor object.  This seris try to
add fundamental support for it.

Please review.  Thanks,

Peter Xu (7):
  monitor: rename out_lock to mon_lock
  monitor: protect mon->fds with mon_lock
  monitor: more comments on lock-free elements
  monitor: fix comment for monitor_lock
  monitor: remove event_clock_type
  monitor: move init global earlier
  monitor: add lock to protect mon_fdsets

 monitor.c     | 168 +++++++++++++++++++++++++++++++++++---------------
 stubs/fdset.c |   2 +-
 util/osdep.c  |   3 +-
 vl.c          |   7 +--
 4 files changed, 121 insertions(+), 59 deletions(-)

-- 
2.17.1