lkml.org 
[lkml]   [2014]   [Aug]   [20]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Subject[PATCH v5 3/5] sched: Remove double_rq_lock() from __migrate_task()
From
Date

Avoid double_rq_lock() and use TASK_ON_RQ_MIGRATING for __migrate_task().
The advantage is (obviously) not holding two 'rq->lock's at the same time
and thereby increasing parallelism.

The important point to note is that because we acquire dst->lock
immediately after releasing src->lock the potential wait time of
task_rq_lock() callers on TASK_ON_RQ_MIGRATING is not longer than it would
have been in the double rq lock scenario.

Signed-off-by: Kirill Tkhai <ktkhai@parallels.com>
---
kernel/sched/core.c | 23 +++++++++++++++--------
1 file changed, 15 insertions(+), 8 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index cef1a13..a50393b 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4663,20 +4663,20 @@ EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
*/
static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
{
- struct rq *rq_dest, *rq_src;
+ struct rq *rq;
int ret = 0;

if (unlikely(!cpu_active(dest_cpu)))
return ret;

- rq_src = cpu_rq(src_cpu);
- rq_dest = cpu_rq(dest_cpu);
+ rq = cpu_rq(src_cpu);

raw_spin_lock(&p->pi_lock);
- double_rq_lock(rq_src, rq_dest);
+ raw_spin_lock(&rq->lock);
/* Already moved. */
if (task_cpu(p) != src_cpu)
goto done;
+
/* Affinity changed (again). */
if (!cpumask_test_cpu(dest_cpu, tsk_cpus_allowed(p)))
goto fail;
@@ -4686,15 +4686,22 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
* placed properly.
*/
if (task_on_rq_queued(p)) {
- dequeue_task(rq_src, p, 0);
+ dequeue_task(rq, p, 0);
+ p->on_rq = TASK_ON_RQ_MIGRATING;
set_task_cpu(p, dest_cpu);
- enqueue_task(rq_dest, p, 0);
- check_preempt_curr(rq_dest, p, 0);
+ raw_spin_unlock(&rq->lock);
+
+ rq = cpu_rq(dest_cpu);
+ raw_spin_lock(&rq->lock);
+ BUG_ON(task_rq(p) != rq);
+ p->on_rq = TASK_ON_RQ_QUEUED;
+ enqueue_task(rq, p, 0);
+ check_preempt_curr(rq, p, 0);
}
done:
ret = 1;
fail:
- double_rq_unlock(rq_src, rq_dest);
+ raw_spin_unlock(&rq->lock);
raw_spin_unlock(&p->pi_lock);
return ret;
}




\
 
 \ /
  Last update: 2014-08-20 12:21    [W:1.847 / U:0.008 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site