[PATCH] sched: unthrottle rt_rq in migrate_tasks()

[PATCH] sched: unthrottle rt_rq in migrate_tasks()

[Peter Boonstoppel]

migrate_tasks() uses _pick_next_task_rt() to get tasks from the
real-time runqueue to be migrated. When rt_rq is throttled
_pick_next_task_rt() won't return anything, in which case
migrate_tasks() can't move all threads over and gets stuck in an
infinite loop.

Instead unthrottle rt_rq before migrating tasks.

Signed-off-by: Peter Boonstoppel <pboonstoppel@nvidia.com>
---
 kernel/sched/core.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 9236a48..b37e0b3 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -5159,6 +5159,12 @@ static void migrate_tasks(unsigned int dead_cpu)
 	/* Ensure any throttled groups are reachable by pick_next_task */
 	unthrottle_offline_cfs_rqs(rq);
 
+	/*
+	 * Ensure rt_rq is not throttled so its threads can be migrated using
+	 * pick_next_task_rt
+	 */
+	rq->rt.rt_throttled = 0;
+
 	for ( ; ; ) {
 		/*
 		 * There's this thread running, bail when that's the only
-- 
1.7.4.1

Re: [PATCH] sched: unthrottle rt_rq in migrate_tasks()

[Peter Zijlstra]

On Fri, 2012-05-18 at 11:56 -0700, Peter Boonstoppel wrote:
> migrate_tasks() uses _pick_next_task_rt() to get tasks from the
> real-time runqueue to be migrated. When rt_rq is throttled
> _pick_next_task_rt() won't return anything, in which case
> migrate_tasks() can't move all threads over and gets stuck in an
> infinite loop.
> 
> Instead unthrottle rt_rq before migrating tasks.
> 
> Signed-off-by: Peter Boonstoppel <pboonstoppel@nvidia.com>
> ---
>  kernel/sched/core.c |    6 ++++++
>  1 files changed, 6 insertions(+), 0 deletions(-)
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 9236a48..b37e0b3 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -5159,6 +5159,12 @@ static void migrate_tasks(unsigned int dead_cpu)
>  	/* Ensure any throttled groups are reachable by pick_next_task */
>  	unthrottle_offline_cfs_rqs(rq);
>  
> +	/*
> +	 * Ensure rt_rq is not throttled so its threads can be migrated using
> +	 * pick_next_task_rt
> +	 */
> +	rq->rt.rt_throttled = 0;
> +
>  	for ( ; ; ) {
>  		/*
>  		 * There's this thread running, bail when that's the only

OK, so there's two things:

 - there's a set_rq_offline() call in CPU_DYING right before calling
migrate_tasks(). This would end up calling rq_offline_rt() which should
disable all the runtime muck.

 - If the above doesn't unthrottle rqs (looks like it doesn't, should
it?) you still only unthrottle the root and 'forgot' to deal with
cgroups.

So I think I'm going to ask you to make rq_offline_rt() dtrt, and while
you're at it, move unthrottle_offline_cfs_rqs() to rq_offline_fair().

[PATCH 1/1] sched: unthrottle rt runqueues in __disable_runtime()

[Peter Boonstoppel]

migrate_tasks() uses _pick_next_task_rt() to get tasks from the
real-time runqueues to be migrated. When rt_rq is throttled
_pick_next_task_rt() won't return anything, in which case
migrate_tasks() can't move all threads over and gets stuck in an
infinite loop.

Instead unthrottle rt runqueues before migrating tasks.

Additionally: move unthrottle_offline_cfs_rqs() to rq_offline_fair()

Signed-off-by: Peter Boonstoppel <pboonstoppel@nvidia.com>
---
 kernel/sched/core.c  |    3 ---
 kernel/sched/fair.c  |    7 +++++--
 kernel/sched/rt.c    |    1 +
 kernel/sched/sched.h |    1 -
 4 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 2df035a..2e7ecff 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -5221,9 +5221,6 @@ static void migrate_tasks(unsigned int dead_cpu)
 	 */
 	rq->stop = NULL;
 
-	/* Ensure any throttled groups are reachable by pick_next_task */
-	unthrottle_offline_cfs_rqs(rq);
-
 	for ( ; ; ) {
 		/*
 		 * There's this thread running, bail when that's the only
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 3704ad3..dc8341b 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2057,7 +2057,7 @@ static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
 	hrtimer_cancel(&cfs_b->slack_timer);
 }
 
-void unthrottle_offline_cfs_rqs(struct rq *rq)
+static void unthrottle_offline_cfs_rqs(struct rq *rq)
 {
 	struct cfs_rq *cfs_rq;
 
@@ -2111,7 +2111,7 @@ static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)
 	return NULL;
 }
 static inline void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b) {}
-void unthrottle_offline_cfs_rqs(struct rq *rq) {}
+static inline void unthrottle_offline_cfs_rqs(struct rq *rq) {}
 
 #endif /* CONFIG_CFS_BANDWIDTH */
 
@@ -5086,6 +5086,9 @@ static void rq_online_fair(struct rq *rq)
 static void rq_offline_fair(struct rq *rq)
 {
 	update_sysctl();
+
+	/* Ensure any throttled groups are reachable by pick_next_task */
+	unthrottle_offline_cfs_rqs(rq);
 }
 
 #endif /* CONFIG_SMP */
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 573e1ca..b9a94fb 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -691,6 +691,7 @@ balanced:
 		 * runtime - in which case borrowing doesn't make sense.
 		 */
 		rt_rq->rt_runtime = RUNTIME_INF;
+		rt_rq->rt_throttled = 0;
 		raw_spin_unlock(&rt_rq->rt_runtime_lock);
 		raw_spin_unlock(&rt_b->rt_runtime_lock);
 	}
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 4134d37..5d9aabe 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -1167,7 +1167,6 @@ extern void print_rt_stats(struct seq_file *m, int cpu);
 
 extern void init_cfs_rq(struct cfs_rq *cfs_rq);
 extern void init_rt_rq(struct rt_rq *rt_rq, struct rq *rq);
-extern void unthrottle_offline_cfs_rqs(struct rq *rq);
 
 extern void account_cfs_bandwidth_used(int enabled, int was_enabled);
 
-- 
1.7.0.4