1 files changed, 72 insertions, 8 deletions

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 5cc1c162044f..3816f217f119 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -719,7 +719,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
                 __enqueue_entity(cfs_rq, se);
 }
 
-static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
+static void __clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
         if (cfs_rq->last == se)
                 cfs_rq->last = NULL;
@@ -728,6 +728,12 @@ static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
                 cfs_rq->next = NULL;
 }
 
+static void clear_buddies(struct cfs_rq *cfs_rq, struct sched_entity *se)
+{
+        for_each_sched_entity(se)
+                __clear_buddies(cfs_rq_of(se), se);
+}
+
 static void
 dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
 {
@@ -768,8 +774,14 @@ check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
 
         ideal_runtime = sched_slice(cfs_rq, curr);
         delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
-        if (delta_exec > ideal_runtime)
+        if (delta_exec > ideal_runtime) {
                 resched_task(rq_of(cfs_rq)->curr);
+                /*
+                 * The current task ran long enough, ensure it doesn't get
+                 * re-elected due to buddy favours.
+                 */
+                clear_buddies(cfs_rq, curr);
+        }
 }
 
 static void
@@ -1302,16 +1314,63 @@ out:
 }
 #endif /* CONFIG_SMP */
 
-static unsigned long wakeup_gran(struct sched_entity *se)
+/*
+ * Adaptive granularity
+ *
+ * se->avg_wakeup gives the average time a task runs until it does a wakeup,
+ * with the limit of wakeup_gran -- when it never does a wakeup.
+ *
+ * So the smaller avg_wakeup is the faster we want this task to preempt,
+ * but we don't want to treat the preemptee unfairly and therefore allow it
+ * to run for at least the amount of time we'd like to run.
+ *
+ * NOTE: we use 2*avg_wakeup to increase the probability of actually doing one
+ *
+ * NOTE: we use *nr_running to scale with load, this nicely matches the
+ *       degrading latency on load.
+ */
+static unsigned long
+adaptive_gran(struct sched_entity *curr, struct sched_entity *se)
+{
+        u64 this_run = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
+        u64 expected_wakeup = 2*se->avg_wakeup * cfs_rq_of(se)->nr_running;
+        u64 gran = 0;
+
+        if (this_run < expected_wakeup)
+                gran = expected_wakeup - this_run;
+
+        return min_t(s64, gran, sysctl_sched_wakeup_granularity);
+}
+
+static unsigned long
+wakeup_gran(struct sched_entity *curr, struct sched_entity *se)
 {
         unsigned long gran = sysctl_sched_wakeup_granularity;
 
+        if (cfs_rq_of(curr)->curr && sched_feat(ADAPTIVE_GRAN))
+                gran = adaptive_gran(curr, se);
+
         /*
-         * More easily preempt - nice tasks, while not making it harder for
-         * + nice tasks.
+         * Since its curr running now, convert the gran from real-time
+         * to virtual-time in his units.
          */
-        if (!sched_feat(ASYM_GRAN) || se->load.weight > NICE_0_LOAD)
-                gran = calc_delta_fair(sysctl_sched_wakeup_granularity, se);
+        if (sched_feat(ASYM_GRAN)) {
+                /*
+                 * By using 'se' instead of 'curr' we penalize light tasks, so
+                 * they get preempted easier. That is, if 'se' < 'curr' then
+                 * the resulting gran will be larger, therefore penalizing the
+                 * lighter, if otoh 'se' > 'curr' then the resulting gran will
+                 * be smaller, again penalizing the lighter task.
+                 *
+                 * This is especially important for buddies when the leftmost
+                 * task is higher priority than the buddy.
+                 */
+                if (unlikely(se->load.weight != NICE_0_LOAD))
+                        gran = calc_delta_fair(gran, se);
+        } else {
+                if (unlikely(curr->load.weight != NICE_0_LOAD))
+                        gran = calc_delta_fair(gran, curr);
+        }
 
         return gran;
 }
@@ -1338,7 +1397,7 @@ wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se)
         if (vdiff <= 0)
                 return -1;
 
-        gran = wakeup_gran(curr);
+        gran = wakeup_gran(curr, se);
         if (vdiff > gran)
                 return 1;
 
@@ -1452,6 +1511,11 @@ static struct task_struct *pick_next_task_fair(struct rq *rq)
 
         do {
                 se = pick_next_entity(cfs_rq);
+                /*
+                 * If se was a buddy, clear it so that it will have to earn
+                 * the favour again.
+                 */
+                __clear_buddies(cfs_rq, se);
                 set_next_entity(cfs_rq, se);
                 cfs_rq = group_cfs_rq(se);
         } while (cfs_rq);