diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
index 6cbb6cb136e9..27c214317942 100644
--- a/include/linux/interrupt.h
+++ b/include/linux/interrupt.h
@@ -593,6 +593,11 @@ enum
 #define SOFTIRQ_HOTPLUG_SAFE_MASK (BIT(TIMER_SOFTIRQ) | BIT(IRQ_POLL_SOFTIRQ) |\
 				   BIT(HRTIMER_SOFTIRQ) | BIT(RCU_SOFTIRQ))
 
+/* Softirq's where the handling might be long: */
+#define LONG_SOFTIRQ_MASK (BIT(NET_TX_SOFTIRQ)    | \
+			   BIT(NET_RX_SOFTIRQ)    | \
+			   BIT(BLOCK_SOFTIRQ)     | \
+			   BIT(IRQ_POLL_SOFTIRQ))
 
 /* map softirq index to softirq name. update 'softirq_to_name' in
  * kernel/softirq.c when adding a new softirq.
@@ -629,6 +634,10 @@ extern void raise_softirq(unsigned int nr);
 
 DECLARE_PER_CPU(struct task_struct *, ksoftirqd);
 
+#ifdef CONFIG_RT_SOFTIRQ_AWARE_SCHED
+DECLARE_PER_CPU(u32, active_softirqs);
+#endif
+
 static inline struct task_struct *this_cpu_ksoftirqd(void)
 {
 	return this_cpu_read(ksoftirqd);
diff --git a/init/Kconfig b/init/Kconfig
index bdbcfda1d22e..ee5334acb06a 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1344,6 +1344,16 @@ config SCHED_AUTOGROUP
 	  desktop applications.  Task group autogeneration is currently based
 	  upon task session.
 
+config RT_SOFTIRQ_AWARE_SCHED
+	bool "Improve RT scheduling during long softirq execution"
+	depends on SMP && !PREEMPT_RT
+	default n
+	help
+	  Enable an optimization which tries to avoid placing RT tasks on CPUs
+	  occupied by nonpreemptible tasks, such as a long softirq or CPUs
+	  which may soon block preemptions, such as a CPU running a ksoftirq
+	  thread which handles slow softirqs.
+
 config RELAY
 	bool "Kernel->user space relay support (formerly relayfs)"
 	select IRQ_WORK
diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c
index 77ae531b37ff..d04dbc34cb22 100644
--- a/kernel/sched/rt.c
+++ b/kernel/sched/rt.c
@@ -1541,6 +1541,32 @@ static void yield_task_rt(struct rq *rq)
 #ifdef CONFIG_SMP
 static int find_lowest_rq(struct task_struct *task);
 
+#ifdef CONFIG_RT_SOFTIRQ_AWARE_SCHED
+/*
+ * Return whether the given cpu is currently non-preemptible
+ * while handling a potentially long softirq, or if the current
+ * task is likely to block preemptions soon because it is a
+ * ksoftirq thread that is handling softirqs.
+ */
+static bool cpu_busy_with_softirqs(int cpu)
+{
+	u32 softirqs = per_cpu(active_softirqs, cpu) |
+		       __cpu_softirq_pending(cpu);
+
+	return softirqs & LONG_SOFTIRQ_MASK;
+}
+#else
+static bool cpu_busy_with_softirqs(int cpu)
+{
+	return false;
+}
+#endif /* CONFIG_RT_SOFTIRQ_AWARE_SCHED */
+
+static bool rt_task_fits_cpu(struct task_struct *p, int cpu)
+{
+	return rt_task_fits_capacity(p, cpu) && !cpu_busy_with_softirqs(cpu);
+}
+
 static int
 select_task_rq_rt(struct task_struct *p, int cpu, int flags)
 {
@@ -1585,22 +1611,24 @@ select_task_rq_rt(struct task_struct *p, int cpu, int flags)
 	 * This test is optimistic, if we get it wrong the load-balancer
 	 * will have to sort it out.
 	 *
-	 * We take into account the capacity of the CPU to ensure it fits the
-	 * requirement of the task - which is only important on heterogeneous
-	 * systems like big.LITTLE.
+	 * We use rt_task_fits_cpu() to evaluate if the CPU is busy with
+	 * potentially long-running softirq work, as well as take into
+	 * account the capacity of the CPU to ensure it fits the
+	 * requirement of the task - which is only important on
+	 * heterogeneous systems like big.LITTLE.
 	 */
 	test = curr &&
 	       unlikely(rt_task(curr)) &&
 	       (curr->nr_cpus_allowed < 2 || curr->prio <= p->prio);
 
-	if (test || !rt_task_fits_capacity(p, cpu)) {
+	if (test || !rt_task_fits_cpu(p, cpu)) {
 		int target = find_lowest_rq(p);
 
 		/*
 		 * Bail out if we were forcing a migration to find a better
 		 * fitting CPU but our search failed.
 		 */
-		if (!test && target != -1 && !rt_task_fits_capacity(p, target))
+		if (!test && target != -1 && !rt_task_fits_cpu(p, target))
 			goto out_unlock;
 
 		/*
@@ -1840,14 +1868,17 @@ static int find_lowest_rq(struct task_struct *task)
 		return -1; /* No other targets possible */
 
 	/*
-	 * If we're on asym system ensure we consider the different capacities
-	 * of the CPUs when searching for the lowest_mask.
+	 * If we're using the softirq optimization or if we are
+	 * on asym system, ensure we consider the softirq processing
+	 * or different capacities of the CPUs when searching for the
+	 * lowest_mask.
 	 */
-	if (sched_asym_cpucap_active()) {
+	if (IS_ENABLED(CONFIG_RT_SOFTIRQ_AWARE_SCHED) ||
+	    sched_asym_cpucap_active()) {
 
 		ret = cpupri_find_fitness(&task_rq(task)->rd->cpupri,
 					  task, lowest_mask,
-					  rt_task_fits_capacity);
+					  rt_task_fits_cpu);
 	} else {
 
 		ret = cpupri_find(&task_rq(task)->rd->cpupri,
diff --git a/kernel/softirq.c b/kernel/softirq.c
index 8371b7c4b517..b69e316c1688 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -64,6 +64,21 @@ static struct softirq_action softirq_vec[NR_SOFTIRQS] __cacheline_aligned_in_smp
 DEFINE_PER_CPU(struct task_struct *, ksoftirqd);
 EXPORT_PER_CPU_SYMBOL_GPL(ksoftirqd);
 
+#ifdef CONFIG_RT_SOFTIRQ_AWARE_SCHED
+/*
+ * active_softirqs -- per cpu, a mask of softirqs that are being handled,
+ * with the expectation that approximate answers are acceptable and therefore
+ * no synchronization.
+ */
+DEFINE_PER_CPU(u32, active_softirqs);
+static inline void set_active_softirqs(u32 pending)
+{
+	__this_cpu_write(active_softirqs, pending);
+}
+#else /* CONFIG_RT_SOFTIRQ_AWARE_SCHED */
+static inline void set_active_softirqs(u32 pending) {};
+#endif /* CONFIG_RT_SOFTIRQ_AWARE_SCHED */
+
 const char * const softirq_to_name[NR_SOFTIRQS] = {
 	"HI", "TIMER", "NET_TX", "NET_RX", "BLOCK", "IRQ_POLL",
 	"TASKLET", "SCHED", "HRTIMER", "RCU"
@@ -537,6 +552,7 @@ static void handle_softirqs(bool ksirqd)
 restart:
 	/* Reset the pending bitmask before enabling irqs */
 	set_softirq_pending(0);
+	set_active_softirqs(pending);
 
 	local_irq_enable();
 
@@ -566,6 +582,7 @@ restart:
 		pending >>= softirq_bit;
 	}
 
+	set_active_softirqs(0);
 	if (!IS_ENABLED(CONFIG_PREEMPT_RT) && ksirqd)
 		rcu_softirq_qs();