rcu: Reduce cache-miss initialization latencies for large systems

Commit #0209f649 (rcu: limit rcu_node leaf-level fanout) set an upper
limit of 16 on the leaf-level fanout for the rcu_node tree.  This was
needed to reduce lock contention that was induced by the synchronization
of scheduling-clock interrupts, which was in turn needed to improve
energy efficiency for moderate-sized lightly loaded servers.

However, reducing the leaf-level fanout means that there are more
leaf-level rcu_node structures in the tree, which in turn means that
RCU's grace-period initialization incurs more cache misses.  This is
not a problem on moderate-sized servers with only a few tens of CPUs,
but becomes a major source of real-time latency spikes on systems with
many hundreds of CPUs.  In addition, the workloads running on these large
systems tend to be CPU-bound, which eliminates the energy-efficiency
advantages of synchronizing scheduling-clock interrupts.  Therefore,
these systems need maximal values for the rcu_node leaf-level fanout.

This commit addresses this problem by introducing a new kernel parameter
named RCU_FANOUT_LEAF that directly controls the leaf-level fanout.
This parameter defaults to 16 to handle the common case of a moderate
sized lightly loaded servers, but may be set higher on larger systems.

Reported-by: Mike Galbraith <efault@gmx.de>
Reported-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

init/Kconfig

@@ -458,6 +458,33 @@ config RCU_FANOUT
 	  Select a specific number if testing RCU itself.
 	  Take the default if unsure.
 
+config RCU_FANOUT_LEAF
+	int "Tree-based hierarchical RCU leaf-level fanout value"
+	range 2 RCU_FANOUT if 64BIT
+	range 2 RCU_FANOUT if !64BIT
+	depends on TREE_RCU || TREE_PREEMPT_RCU
+	default 16
+	help
+	  This option controls the leaf-level fanout of hierarchical
+	  implementations of RCU, and allows trading off cache misses
+	  against lock contention.  Systems that synchronize their
+	  scheduling-clock interrupts for energy-efficiency reasons will
+	  want the default because the smaller leaf-level fanout keeps
+	  lock contention levels acceptably low.  Very large systems
+	  (hundreds or thousands of CPUs) will instead want to set this
+	  value to the maximum value possible in order to reduce the
+	  number of cache misses incurred during RCU's grace-period
+	  initialization.  These systems tend to run CPU-bound, and thus
+	  are not helped by synchronized interrupts, and thus tend to
+	  skew them, which reduces lock contention enough that large
+	  leaf-level fanouts work well.
+
+	  Select a specific number if testing RCU itself.
+
+	  Select the maximum permissible value for large systems.
+
+	  Take the default if unsure.
+
 config RCU_FANOUT_EXACT
 	bool "Disable tree-based hierarchical RCU auto-balancing"
 	depends on TREE_RCU || TREE_PREEMPT_RCU