sched/deadline: Reduce rq lock contention by eliminating locking of non-feasible...
authorWanpeng Li <wanpeng.li@linux.intel.com>
Wed, 13 May 2015 06:01:03 +0000 (14:01 +0800)
committerIngo Molnar <mingo@kernel.org>
Fri, 19 Jun 2015 08:06:46 +0000 (10:06 +0200)
This patch adds a check that prevents futile attempts to move DL tasks
to a CPU with active tasks of equal or earlier deadline. The same
behavior as commit 80e3d87b2c55 ("sched/rt: Reduce rq lock contention
by eliminating locking of non-feasible target") for rt class.

Signed-off-by: Wanpeng Li <wanpeng.li@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1431496867-4194-3-git-send-email-wanpeng.li@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/sched/deadline.c

index 1c4bc31eb0f5de0f35194236a732c7c169d62581..98f787143db4058f46e5d12b09d44a276b18a292 100644 (file)
@@ -1012,7 +1012,9 @@ select_task_rq_dl(struct task_struct *p, int cpu, int sd_flag, int flags)
            (p->nr_cpus_allowed > 1)) {
                int target = find_later_rq(p);
 
-               if (target != -1)
+               if (target != -1 &&
+                               dl_time_before(p->dl.deadline,
+                                       cpu_rq(target)->dl.earliest_dl.curr))
                        cpu = target;
        }
        rcu_read_unlock();
@@ -1359,6 +1361,17 @@ static struct rq *find_lock_later_rq(struct task_struct *task, struct rq *rq)
 
                later_rq = cpu_rq(cpu);
 
+               if (!dl_time_before(task->dl.deadline,
+                                       later_rq->dl.earliest_dl.curr)) {
+                       /*
+                        * Target rq has tasks of equal or earlier deadline,
+                        * retrying does not release any lock and is unlikely
+                        * to yield a different result.
+                        */
+                       later_rq = NULL;
+                       break;
+               }
+
                /* Retry if something changed. */
                if (double_lock_balance(rq, later_rq)) {
                        if (unlikely(task_rq(task) != rq ||