Commit f549da848eca595abca14ebc5e1bf00fd72aa53d

Authored by Suresh Siddha
Committed by Ingo Molnar
1 parent f8700df7c4

sched: skip updating rq's next_balance under null SD

Was playing with sched_smt_power_savings/sched_mc_power_savings and
found out that while the scheduler domains are reconstructed when sysfs
settings change, rebalance_domains() can get triggered with null domain
on other cpus, which is setting next_balance to jiffies + 60*HZ.
Resulting in no idle/busy balancing for 60 seconds.

Fix this.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Showing 1 changed file with 12 additions and 2 deletions Side-by-side Diff

... ... @@ -3043,6 +3043,7 @@
3043 3043 struct sched_domain *sd;
3044 3044 /* Earliest time when we have to do rebalance again */
3045 3045 unsigned long next_balance = jiffies + 60*HZ;
  3046 + int update_next_balance = 0;
3046 3047  
3047 3048 for_each_domain(cpu, sd) {
3048 3049 if (!(sd->flags & SD_LOAD_BALANCE))
3049 3050  
... ... @@ -3079,8 +3080,10 @@
3079 3080 if (sd->flags & SD_SERIALIZE)
3080 3081 spin_unlock(&balancing);
3081 3082 out:
3082   - if (time_after(next_balance, sd->last_balance + interval))
  3083 + if (time_after(next_balance, sd->last_balance + interval)) {
3083 3084 next_balance = sd->last_balance + interval;
  3085 + update_next_balance = 1;
  3086 + }
3084 3087  
3085 3088 /*
3086 3089 * Stop the load balance at this level. There is another
... ... @@ -3090,7 +3093,14 @@
3090 3093 if (!balance)
3091 3094 break;
3092 3095 }
3093   - rq->next_balance = next_balance;
  3096 +
  3097 + /*
  3098 + * next_balance will be updated only when there is a need.
  3099 + * When the cpu is attached to null domain for ex, it will not be
  3100 + * updated.
  3101 + */
  3102 + if (likely(update_next_balance))
  3103 + rq->next_balance = next_balance;
3094 3104 }
3095 3105  
3096 3106 /*