Re: [PATCH 01/18] move do_timer() from kernel/timer.c into kernel/time/timekeeping.c

[Yong Zhang <yong.zhang0@gmail.com>]

On Tue, Jan 25, 2011 at 05:51:59PM +0100, Peter Zijlstra wrote:
> On Tue, 2011-01-25 at 11:34 +0100, Peter Zijlstra wrote:
> 
> > Something like the (completely untested) below would do I guess:
> 
> > @@ -8172,6 +8180,8 @@ void __init sched_init(void)
> >  	init_idle(current, smp_processor_id());
> >  
> >  	calc_load_update = jiffies + LOAD_FREQ;
> > +	global_load_timer.slack = 0;
> > +	mod_timer(&global_load_timer, calc_load_update + 10);
> >  
> >  	/*
> >  	 * During early bootup we pretend to be a normal task:
> 
> OK, so calling mod_timer() before init_timers() is _not_ a good idea ;-)

So I change your patch a little, and it works well on my PC ;)

0) move calc_global_times things to one function start_calc_global_timer();
1) call start_calc_global_timer() in sched_init_smp() instead of
   sched_init(); thus we change the boot-time load calculation(later
   than before), but I don't think it's matter;
2) remove timer_before() check in calc_global_load(), if that check
   returns true, we will lose global-load-timer forever.

Thanks,
Yong

---
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Subject: [PATCH] sched: Move the calc_global_load() call into the scheduler

Remove the calc_global_load() call from the timekeeping code and make
it local to the scheduler.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
---
 include/linux/sched.h |    2 --
 kernel/sched.c        |   34 +++++++++++++++++++++++-----------
 kernel/timer.c        |    1 -
 3 files changed, 23 insertions(+), 14 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index d747f94..f224dcc 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -142,8 +142,6 @@ extern unsigned long nr_iowait_cpu(int cpu);
 extern unsigned long this_cpu_load(void);
 
 
-extern void calc_global_load(unsigned long ticks);
-
 extern unsigned long get_parent_ip(unsigned long addr);
 
 struct seq_file;
diff --git a/kernel/sched.c b/kernel/sched.c
index 18d38e4..89f6725 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -3159,7 +3159,7 @@ calc_load_n(unsigned long load, unsigned long exp,
  * Once we've updated the global active value, we need to apply the exponential
  * weights adjusted to the number of cycles missed.
  */
-static void calc_global_nohz(unsigned long ticks)
+static void calc_global_nohz(void)
 {
 	long delta, active, n;
 
@@ -3179,8 +3179,9 @@ static void calc_global_nohz(unsigned long ticks)
 	/*
 	 * If we were idle for multiple load cycles, apply them.
 	 */
-	if (ticks >= LOAD_FREQ) {
-		n = ticks / LOAD_FREQ;
+	delta = jiffies - calc_load_update - 10;
+	if (delta >= LOAD_FREQ) {
+		n = delta / LOAD_FREQ;
 
 		active = atomic_long_read(&calc_load_tasks);
 		active = active > 0 ? active * FIXED_1 : 0;
@@ -3213,7 +3214,7 @@ static inline long calc_load_fold_idle(void)
 	return 0;
 }
 
-static void calc_global_nohz(unsigned long ticks)
+static void calc_global_nohz(void)
 {
 }
 #endif
@@ -3233,18 +3234,27 @@ void get_avenrun(unsigned long *loads, unsigned long offset, int shift)
 	loads[2] = (avenrun[2] + offset) << shift;
 }
 
+static void calc_global_load(unsigned long __data);
+
+static struct timer_list global_load_timer =
+	TIMER_DEFERRED_INITIALIZER(calc_global_load, 0, 0);
+
+static __init void start_calc_global_timer()
+{
+	calc_load_update = jiffies + LOAD_FREQ;
+	set_timer_slack(&global_load_timer, 0);
+	mod_timer(&global_load_timer, calc_load_update + 10);
+}
+
 /*
  * calc_load - update the avenrun load estimates 10 ticks after the
  * CPUs have updated calc_load_tasks.
  */
-void calc_global_load(unsigned long ticks)
+static void calc_global_load(unsigned long __data)
 {
 	long active;
 
-	calc_global_nohz(ticks);
-
-	if (time_before(jiffies, calc_load_update + 10))
-		return;
+	calc_global_nohz();
 
 	active = atomic_long_read(&calc_load_tasks);
 	active = active > 0 ? active * FIXED_1 : 0;
@@ -3254,6 +3264,7 @@ void calc_global_load(unsigned long ticks)
 	avenrun[2] = calc_load(avenrun[2], EXP_15, active);
 
 	calc_load_update += LOAD_FREQ;
+	mod_timer(&global_load_timer, calc_load_update + 10);
 }
 
 /*
@@ -7741,6 +7752,8 @@ void __init sched_init_smp(void)
 {
 	cpumask_var_t non_isolated_cpus;
 
+	start_calc_global_timer();
+
 	alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
 	alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
 
@@ -7777,6 +7790,7 @@ void __init sched_init_smp(void)
 #else
 void __init sched_init_smp(void)
 {
+	start_calc_global_timer();
 	sched_init_granularity();
 }
 #endif /* CONFIG_SMP */
@@ -8044,8 +8058,6 @@ void __init sched_init(void)
 	 */
 	init_idle(current, smp_processor_id());
 
-	calc_load_update = jiffies + LOAD_FREQ;
-
 	/*
 	 * During early bootup we pretend to be a normal task:
 	 */
diff --git a/kernel/timer.c b/kernel/timer.c
index 43ca993..afdc13b 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -1303,7 +1303,6 @@ void do_timer(unsigned long ticks)
 {
 	jiffies_64 += ticks;
 	update_wall_time();
-	calc_global_load(ticks);
 }
 
 #ifdef __ARCH_WANT_SYS_ALARM
-- 
1.7.0.4