Permalink
Browse files

sched: Ensure cpu_power periodic update

With a lot of small task, the softirq sched is nearly never called
when no_hz is enable. Te load_balance is mainly called with
the newly_idle mode which doesn't update the cpu_power.
Add a next_update field which ensure a maximum update period when
there is short activity
  • Loading branch information...
1 parent 9047857 commit ec501691ba997d4a5af6a3bf1741fa29b260bc57 @vingu-linaro vingu-linaro committed with Nov 14, 2011
Showing with 17 additions and 8 deletions.
  1. +1 −0 include/linux/sched.h
  2. +16 −8 kernel/sched_fair.c
View
1 include/linux/sched.h
@@ -902,6 +902,7 @@ struct sched_group_power {
* single CPU.
*/
unsigned int power, power_orig;
+ unsigned long next_update;
};
struct sched_group {
View
24 kernel/sched_fair.c
@@ -91,6 +91,8 @@ unsigned int __read_mostly sysctl_sched_shares_window = 10000000UL;
static const struct sched_class fair_sched_class;
+static unsigned long __read_mostly max_load_balance_interval = HZ/10;
+
/**************************************************************
* CFS operations on generic schedulable entities:
*/
@@ -2655,6 +2657,11 @@ static void update_group_power(struct sched_domain *sd, int cpu)
struct sched_domain *child = sd->child;
struct sched_group *group, *sdg = sd->groups;
unsigned long power;
+ unsigned long interval;
+
+ interval = msecs_to_jiffies(sd->balance_interval);
+ interval = clamp(interval, 1UL, max_load_balance_interval);
+ sdg->sgp->next_update = jiffies + interval;
if (!child) {
update_cpu_power(sd, cpu);
@@ -2762,12 +2769,15 @@ static inline void update_sg_lb_stats(struct sched_domain *sd,
* domains. In the newly idle case, we will allow all the cpu's
* to do the newly idle load balance.
*/
- if (idle != CPU_NEWLY_IDLE && local_group) {
- if (balance_cpu != this_cpu) {
- *balance = 0;
- return;
- }
- update_group_power(sd, this_cpu);
+ if (local_group) {
+ if (idle != CPU_NEWLY_IDLE) {
+ if (balance_cpu != this_cpu) {
+ *balance = 0;
+ return;
+ }
+ update_group_power(sd, this_cpu);
+ } else if (time_after_eq(jiffies, group->sgp->next_update))
+ update_group_power(sd, this_cpu);
}
/* Adjust by relative CPU power of the group */
@@ -3867,8 +3877,6 @@ void select_nohz_load_balancer(int stop_tick)
static DEFINE_SPINLOCK(balancing);
-static unsigned long __read_mostly max_load_balance_interval = HZ/10;
-
/*
* Scale the max load_balance interval with the number of CPUs in the system.
* This trades load-balance latency on larger machines for less cross talk.

0 comments on commit ec50169

Please sign in to comment.