When p is current and it's not of dl class, then there are no other
dl taks in the rq. If we had had pushable tasks in some other rq,
they would have been pushed earlier. So, skip "p == rq->curr" case.
This helps to skip excess schedule() when class is changed.
The situation may be not rare. For example:
a small number of deadline tasks which have common rt_spinlocks
with fair tasks. Priority inheritance mechanism does this every
time it is acquiring a lock.
[This is confirmed by Juri Lelli and LKML was CC'ed, but
unfotunately I can't find direct link on lkml.org]
Signed-off-by: Kirill Tkhai <[email protected]>
CC: Juri Lelli <[email protected]>
CC: Peter Zijlstra <[email protected]>
CC: Ingo Molnar <[email protected]>
---
kernel/sched/deadline.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 0de2482..dd19d6d 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -1559,7 +1559,7 @@ static void switched_to_dl(struct rq *rq, struct task_struct *p)
if (unlikely(p->dl.dl_throttled))
return;
- if (p->on_rq || rq->curr != p) {
+ if (p->on_rq && rq->curr != p) {
#ifdef CONFIG_SMP
if (rq->dl.overloaded && push_dl_task(rq) && rq != task_rq(p))
/* Only reschedule if pushing failed */