[RFC PATCH RESEND 1/4] sched/rt: Modify check_preempt_equal_prio() for multiple tasks queued at the same priority

[Xunlei Pang]

From: Xunlei Pang <pang.xunlei@xxxxxxxxxx>

In check_preempt_equal_prio(), when p is queued, there may be other
tasks already queued at the same priority in the "run queue", so we
should peek the most front one to do the preemption, not the p.

This patch modified it and moved the preemption job to a new function
named check_preempt_equal_prio_common() to make the logic clearer.

Signed-off-by: Xunlei Pang <pang.xunlei@xxxxxxxxxx>
---
 kernel/sched/rt.c | 70 ++++++++++++++++++++++++++++++++++++++++++-------------
 1 file changed, 54 insertions(+), 16 deletions(-)

diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 575da76..0c0f4df 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1366,33 +1366,66 @@ out:
 	return cpu;
 }
 
-static void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
+static struct task_struct *peek_next_task_rt(struct rq *rq);
+
+static void check_preempt_equal_prio_common(struct rq *rq)
 {
+	struct task_struct *curr = rq->curr;
+	struct task_struct *next;
+
+	/* Current can't be migrated, useless to reschedule */
+	if (curr->nr_cpus_allowed == 1 ||
+	    !cpupri_find(&rq->rd->cpupri, curr, NULL))
+		return;
+
 	/*
-	 * Current can't be migrated, useless to reschedule,
-	 * let's hope p can move out.
+	 * Can we find any task with the same priority as
+	 * curr? To accomplish this, firstly requeue curr
+	 * to the tail, then peek next, finally put curr
+	 * back to the head if a different task was peeked.
 	 */
-	if (rq->curr->nr_cpus_allowed == 1 ||
-	    !cpupri_find(&rq->rd->cpupri, rq->curr, NULL))
+	requeue_task_rt(rq, curr, 0);
+	next = peek_next_task_rt(rq);
+	if (next == curr)
+		return;
+
+	requeue_task_rt(rq, curr, 1);
+
+	if (next->prio != curr->prio)
 		return;
 
 	/*
-	 * p is migratable, so let's not schedule it and
-	 * see if it is pushed or pulled somewhere else.
+	 * Got the right "next" queued with the same priority
+	 * as current. If next is migratable, don't schedule
+	 * it as it will be pushed or pulled somewhere else.
 	 */
-	if (p->nr_cpus_allowed != 1
-	    && cpupri_find(&rq->rd->cpupri, p, NULL))
+	if (next->nr_cpus_allowed != 1 &&
+	    cpupri_find(&rq->rd->cpupri, next, NULL))
 		return;
 
 	/*
 	 * There appears to be other cpus that can accept
-	 * current and none to run 'p', so lets reschedule
-	 * to try and push current away:
+	 * current and none to run next, so lets reschedule
+	 * to try and push current away.
 	 */
-	requeue_task_rt(rq, p, 1);
+	requeue_task_rt(rq, next, 1);
 	resched_curr(rq);
 }
 
+static inline
+void check_preempt_equal_prio(struct rq *rq, struct task_struct *p)
+{
+	/*
+	 * p is migratable, so let's not schedule it and
+	 * see if it is pushed or pulled somewhere else.
+	 */
+	if (p->nr_cpus_allowed != 1 &&
+	    cpupri_find(&rq->rd->cpupri, p, NULL))
+		return;
+
+	check_preempt_equal_prio_common(rq);
+}
+
 #endif /* CONFIG_SMP */
 
 /*
@@ -1440,10 +1473,9 @@ static struct sched_rt_entity *pick_next_rt_entity(struct rq *rq,
 	return next;
 }
 
-static struct task_struct *_pick_next_task_rt(struct rq *rq)
+static struct task_struct *peek_next_task_rt(struct rq *rq)
 {
 	struct sched_rt_entity *rt_se;
-	struct task_struct *p;
 	struct rt_rq *rt_rq  = &rq->rt;
 
 	do {
@@ -1452,9 +1484,15 @@ static struct task_struct *_pick_next_task_rt(struct rq *rq)
 		rt_rq = group_rt_rq(rt_se);
 	} while (rt_rq);
 
-	p = rt_task_of(rt_se);
-	p->se.exec_start = rq_clock_task(rq);
+	return rt_task_of(rt_se);
+}
 
+static inline struct task_struct *_pick_next_task_rt(struct rq *rq)
+{
+	struct task_struct *p;
+
+	p = peek_next_task_rt(rq);
+	p->se.exec_start = rq_clock_task(rq);
 	return p;
 }
 
-- 
1.9.1


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/