[v2 PATCH 4/4] timers: logic to enable timer migration.

[Arun R Bharadwaj <arun@linux.vnet.ibm.com>]

* Arun R Bharadwaj <arun@linux.vnet.ibm.com> [2009-03-04 17:42:49]:

This patch migrates all non pinned timers and hrtimers to the current
idle load balancer, from all the idle CPUs. Timers firing on busy CPUs
are not migrated.


Signed-off-by: Arun R Bharadwaj <arun@linux.vnet.ibm.com>
---
 include/linux/sched.h |    1 +
 kernel/hrtimer.c      |   12 +++++++++++-
 kernel/sched.c        |    5 +++++
 kernel/timer.c        |   13 ++++++++++++-
 4 files changed, 29 insertions(+), 2 deletions(-)

Index: linux.trees.git/kernel/timer.c
===================================================================
--- linux.trees.git.orig/kernel/timer.c
+++ linux.trees.git/kernel/timer.c
@@ -38,6 +38,7 @@
 #include <linux/tick.h>
 #include <linux/kallsyms.h>
 #include <linux/timer.h>
+#include <linux/sched.h>
 
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
@@ -628,7 +629,7 @@ __mod_timer(struct timer_list *timer, un
 {
 	struct tvec_base *base, *new_base;
 	unsigned long flags;
-	int ret;
+	int ret, current_cpu, preferred_cpu;
 
 	ret = 0;
 
@@ -649,6 +650,16 @@ __mod_timer(struct timer_list *timer, un
 
 	new_base = __get_cpu_var(tvec_bases);
 
+	current_cpu = smp_processor_id();
+	preferred_cpu = get_nohz_load_balancer();
+	if (enable_timer_migration && !tbase_get_pinned(timer->base) &&
+			idle_cpu(current_cpu) && preferred_cpu != -1) {
+		new_base = per_cpu(tvec_bases, preferred_cpu);
+		timer_set_base(timer, new_base);
+		timer->expires = expires;
+		internal_add_timer(new_base, timer);
+		goto out_unlock;
+	}
 	if (base != new_base) {
 		/*
 		 * We are trying to schedule the timer on the local CPU.
Index: linux.trees.git/kernel/hrtimer.c
===================================================================
--- linux.trees.git.orig/kernel/hrtimer.c
+++ linux.trees.git/kernel/hrtimer.c
@@ -43,6 +43,8 @@
 #include <linux/seq_file.h>
 #include <linux/err.h>
 #include <linux/debugobjects.h>
+#include <linux/sched.h>
+#include <linux/timer.h>
 
 #include <asm/uaccess.h>
 
@@ -198,8 +200,16 @@ int pinned)
 {
 	struct hrtimer_clock_base *new_base;
 	struct hrtimer_cpu_base *new_cpu_base;
+	int current_cpu, preferred_cpu;
+
+	current_cpu = smp_processor_id();
+	preferred_cpu = get_nohz_load_balancer();
+	if (enable_timer_migration && !pinned && preferred_cpu != -1 &&
+			idle_cpu(current_cpu))
+		new_cpu_base = &per_cpu(hrtimer_bases, preferred_cpu);
+	else
+		new_cpu_base = &__get_cpu_var(hrtimer_bases);
 
-	new_cpu_base = &__get_cpu_var(hrtimer_bases);
 	new_base = &new_cpu_base->clock_base[base->index];
 
 	if (base != new_base) {
Index: linux.trees.git/include/linux/sched.h
===================================================================
--- linux.trees.git.orig/include/linux/sched.h
+++ linux.trees.git/include/linux/sched.h
@@ -265,6 +265,7 @@ static inline int select_nohz_load_balan
 }
 #endif
 
+extern int get_nohz_load_balancer(void);
 /*
  * Only dump TASK_* tasks. (0 for all tasks)
  */
Index: linux.trees.git/kernel/sched.c
===================================================================
--- linux.trees.git.orig/kernel/sched.c
+++ linux.trees.git/kernel/sched.c
@@ -4009,6 +4009,11 @@ static struct {
 	.load_balancer = ATOMIC_INIT(-1),
 };
 
+inline int get_nohz_load_balancer(void)
+{
+	return atomic_read(&nohz.load_balancer);
+}
+
 /*
  * This routine will try to nominate the ilb (idle load balancing)
  * owner among the cpus whose ticks are stopped. ilb owner will do the idle