sched: smpnice - fix average load per run queue calculations
PATCH 1/1 sched: smpnice - fix average load per run queue calculations
The logical value to use for the average load per task on a run queue without any runnable tasks is the "default" load for a nice==0 task. Failure to do this may lead to anomalies in try_to_wake_up(), etc.
Signed-off-by Peter Williams <[email protected]>
Peter
Signature · 4 lines
Peter Williams [email protected] "Learning, n. The kind of ignorance distinguishing the studious." -- Ambrose Bierce
kernel/sched.c +1 -1
@@ -1058,7 +1058,7 @@ static inline unsigned long cpu_avg_loadrunqueue_t *rq =cpu_rq (cpu);unsigned long n = rq->nr_running; -return n ? rq->raw_weighted_load / n : rq->raw_weighted_load; +return n ? rq->raw_weighted_load / n :SCHED_LOAD_SCALE ; }/*
Quoted 3 lines · Peter Williams
The logical value to use for the average load per task on a run queue without any runnable tasks is the "default" load for a nice==0 task. Failure to do this may lead to anomalies in
try_to_wake_up(), etc.Signed-off-by Peter Williams <[email protected]>
I missed a bit. Here's a fixed patch.
Signed-off-by Peter Williams <[email protected]>
Sorry,
Peter
Signature · 4 lines
Peter Williams [email protected] "Learning, n. The kind of ignorance distinguishing the studious." -- Ambrose Bierce
kernel/sched.c +3 -1
@@ -1058,7 +1058,7 @@ static inline unsigned long cpu_avg_loadrunqueue_t *rq =cpu_rq (cpu);unsigned long n = rq->nr_running; -return n ? rq->raw_weighted_load / n : rq->raw_weighted_load; +return n ? rq->raw_weighted_load / n :SCHED_LOAD_SCALE ; }/*
@@ -2153,6 +2153,8 @@ find_busiest_group(struct sched_domain *min (busiest_load_per_task, max_load);if (this_nr_running) this_load_per_task /= this_nr_running; +else + this_load_per_task =SCHED_LOAD_SCALE ; pwr_now += this->cpu_power *min (this_load_per_task, this_load); pwr_now /=SCHED_LOAD_SCALE ;