sched: revert back to per-rq vruntime

Vatsa rightly points out that having the runqueue weight in the vruntime
calculations can cause unfairness in the face of task joins/leaves.

Suppose: dv = dt * rw / w

Then take 10 tasks t_n, each of similar weight. If the first will run 1
then its vruntime will increase by 10. Now, if the next 8 tasks leave after
having run their 1, then the last task will get a vruntime increase of 2
after having run 1.

Which will leave us with 2 tasks of equal weight and equal runtime, of which
one will not be scheduled for 8/2=4 units of time.

Ergo, we cannot do that and must use: dv = dt / w.

This means we cannot have a global vruntime based on effective priority, but
must instead go back to the vruntime per rq model we started out with.

This patch was lightly tested by doing starting while loops on each nice level
and observing their execution time, and a simple group scenario of 1:2:3 pinned
to a single cpu.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

1 files changed, 15 insertions, 17 deletions

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 0c4bcac54761..a0aa38b10fdd 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -336,7 +336,7 @@ int sched_nr_latency_handler(struct ctl_table *table, int write,
 #endif
 
 /*
- * delta *= w / rw
+ * delta *= P[w / rw]
  */
 static inline unsigned long
 calc_delta_weight(unsigned long delta, struct sched_entity *se)
@@ -350,15 +350,13 @@ calc_delta_weight(unsigned long delta, struct sched_entity *se)
 }
 
 /*
- * delta *= rw / w
+ * delta /= w
  */
 static inline unsigned long
 calc_delta_fair(unsigned long delta, struct sched_entity *se)
 {
-        for_each_sched_entity(se) {
-                delta = calc_delta_mine(delta,
-                                cfs_rq_of(se)->load.weight, &se->load);
-        }
+        if (unlikely(se->load.weight != NICE_0_LOAD))
+                delta = calc_delta_mine(delta, NICE_0_LOAD, &se->load);
 
         return delta;
 }
@@ -388,26 +386,26 @@ static u64 __sched_period(unsigned long nr_running)
  * We calculate the wall-time slice from the period by taking a part
  * proportional to the weight.
  *
- * s = p*w/rw
+ * s = p*P[w/rw]
  */
 static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-        return calc_delta_weight(__sched_period(cfs_rq->nr_running), se);
+        unsigned long nr_running = cfs_rq->nr_running;
+
+        if (unlikely(!se->on_rq))
+                nr_running++;
+
+        return calc_delta_weight(__sched_period(nr_running), se);
 }
 
 /*
  * We calculate the vruntime slice of a to be inserted task
  *
- * vs = s*rw/w = p
+ * vs = s/w
  */
-static u64 sched_vslice_add(struct cfs_rq *cfs_rq, struct sched_entity *se)
+static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-        unsigned long nr_running = cfs_rq->nr_running;
-
-        if (!se->on_rq)
-                nr_running++;
-
-        return __sched_period(nr_running);
+        return calc_delta_fair(sched_slice(cfs_rq, se), se);
 }
 
 /*
@@ -629,7 +627,7 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
          * stays open at the end.
          */
         if (initial && sched_feat(START_DEBIT))
-                vruntime += sched_vslice_add(cfs_rq, se);
+                vruntime += sched_vslice(cfs_rq, se);
 
         if (!initial) {
                 /* sleeps upto a single latency don't count. */