rcu: Make synchronize_srcu_expedited() fast if running readers

The synchronize_srcu_expedited() function is currently quick if there
are no active readers, but will delay a full jiffy if there are any.
If these readers leave their SRCU read-side critical sections quickly,
this is way too long to wait.  So this commit first waits ten microseconds,
and only then falls back to jiffy-at-a-time waiting.

Reported-by: Avi Kivity <avi@redhat.com>
Reported-by: Marcelo Tosatti <mtosatti@redhat.com>
Tested-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>

init/Kconfig

@@ -488,6 +488,21 @@ config RCU_BOOST_DELAY
 
 	  Accept the default if unsure.
 
+config SRCU_SYNCHRONIZE_DELAY
+	int "Microseconds to delay before waiting for readers"
+	range 0 20
+	default 10
+	help
+	  This option controls how long SRCU delays before entering its
+	  loop waiting on SRCU readers.  The purpose of this loop is
+	  to avoid the unconditional context-switch penalty that would
+	  otherwise be incurred if there was an active SRCU reader,
+	  in a manner similar to adaptive locking schemes.  This should
+	  be set to be a bit longer than the common-case SRCU read-side
+	  critical-section overhead.
+
+	  Accept the default if unsure.
+
 endmenu # "RCU Subsystem"
 
 config IKCONFIG

kernel/srcu.c

@@ -31,6 +31,7 @@
 #include <linux/rcupdate.h>
 #include <linux/sched.h>
 #include <linux/smp.h>
+#include <linux/delay.h>
 #include <linux/srcu.h>
 
 static int init_srcu_struct_fields(struct srcu_struct *sp)
@@ -203,9 +204,14 @@ static void __synchronize_srcu(struct srcu_struct *sp, void (*sync_func)(void))
 	 * all srcu_read_lock() calls using the old counters have completed.
 	 * Their corresponding critical sections might well be still
 	 * executing, but the srcu_read_lock() primitives themselves
-	 * will have finished executing.
+	 * will have finished executing.  We initially give readers
+	 * an arbitrarily chosen 10 microseconds to get out of their
+	 * SRCU read-side critical sections, then loop waiting 1/HZ
+	 * seconds per iteration.
 	 */
 
+	if (srcu_readers_active_idx(sp, idx))
+		udelay(CONFIG_SRCU_SYNCHRONIZE_DELAY);
 	while (srcu_readers_active_idx(sp, idx))
 		schedule_timeout_interruptible(1);