[PATCH tip/core/rcu 0/3] Parallelize expedited grace periods for v4.18

[Paul E. McKenney]

Hello!

This series parallelizes grace-period initialization for expedited
grace periods and adds a couple of fixes:

1.	Parallelize expedited grace-period initialization in order to
	reduce the growth of latency with increasing numbers of CPUs.

2.	Updates comments on which locks are to be held to match the
	code, courtesy of Boqun Feng.

3.	Protect all calls to sync_rcu_preempt_exp_done() with the
	corresponding rcu_node structure's ->lock held, courtesy of
	Boqun Feng.

							Thanx, Paul

------------------------------------------------------------------------

 rcu.h      |    1 
 tree.c     |    3 
 tree.h     |   10 ++
 tree_exp.h |  222 ++++++++++++++++++++++++++++++++++++-------------------------
 4 files changed, 148 insertions(+), 88 deletions(-)