sched/rt: Try to restart rt period timer when rt runtime exceeded

When rt_runtime is modified from -1 to a valid control value, it may
cause the task to be throttled all the time. Operations like the following
will trigger the bug. E.g:

  1. echo -1 > /proc/sys/kernel/sched_rt_runtime_us
  2. Run a FIFO task named A that executes while(1)
  3. echo 950000 > /proc/sys/kernel/sched_rt_runtime_us

When rt_runtime is -1, The rt period timer will not be activated when task
A enqueued. And then the task will be throttled after setting rt_runtime to
950,000. The task will always be throttled because the rt period timer is
not activated.

Fixes: d0b27fa7 ("sched: rt-group: synchonised bandwidth period")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Li Hua <hucool.lihua@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20211203033618.11895-1-hucool.lihua@huawei.com

kernel/sched/rt.c

@@ -52,11 +52,8 @@ void init_rt_bandwidth(struct rt_bandwidth *rt_b, u64 period, u64 runtime)
 	rt_b->rt_period_timer.function = sched_rt_period_timer;
 }
 
-static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
+static inline void do_start_rt_bandwidth(struct rt_bandwidth *rt_b)
 {
-	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
-		return;
-
 	raw_spin_lock(&rt_b->rt_runtime_lock);
 	if (!rt_b->rt_period_active) {
 		rt_b->rt_period_active = 1;
@@ -75,6 +72,14 @@ static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
 	raw_spin_unlock(&rt_b->rt_runtime_lock);
 }
 
+static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
+{
+	if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
+		return;
+
+	do_start_rt_bandwidth(rt_b);
+}
+
 void init_rt_rq(struct rt_rq *rt_rq)
 {
 	struct rt_prio_array *array;
@@ -1031,13 +1036,17 @@ static void update_curr_rt(struct rq *rq)
 
 	for_each_sched_rt_entity(rt_se) {
 		struct rt_rq *rt_rq = rt_rq_of_se(rt_se);
+		int exceeded;
 
 		if (sched_rt_runtime(rt_rq) != RUNTIME_INF) {
 			raw_spin_lock(&rt_rq->rt_runtime_lock);
 			rt_rq->rt_time += delta_exec;
-			if (sched_rt_runtime_exceeded(rt_rq))
+			exceeded = sched_rt_runtime_exceeded(rt_rq);
+			if (exceeded)
 				resched_curr(rq);
 			raw_spin_unlock(&rt_rq->rt_runtime_lock);
+			if (exceeded)
+				do_start_rt_bandwidth(sched_rt_bandwidth(rt_rq));
 		}
 	}
 }
@@ -2911,8 +2920,12 @@ static int sched_rt_global_validate(void)
 
 static void sched_rt_do_global(void)
 {
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
 	def_rt_bandwidth.rt_runtime = global_rt_runtime();
 	def_rt_bandwidth.rt_period = ns_to_ktime(global_rt_period());
+	raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
 }
 
 int sched_rt_handler(struct ctl_table *table, int write, void *buffer,