[PATCH] irq_poll: Use raise_softirq_irqoff() in cpu_dead notifier

From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To: linux-kernel@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@linutronix.de>
Subject: [PATCH] irq_poll: Use raise_softirq_irqoff() in cpu_dead notifier
Date: Thu, 2 Apr 2020 21:50:28 +0200	[thread overview]
Message-ID: <20200402195028.skpufyn4c6ktgai3@linutronix.de> (raw)

__raise_softirq_irqoff() adds a pending sofirq mask and this is it. The
softirq won't be invoked unless something checks for a pending softirq
which should happen once the next interrupt fires. The CPU might go
idle in the meantime.

Use raise_softirq_irqoff() to wake ksoftirqd() if the list has work
items pending.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---

Maybe preempt_check_resched() should be added to act on a possible
need_resched flag. Otherwise it will be delayed to the next
cond_resched() / spin_unlock().

 lib/irq_poll.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/lib/irq_poll.c b/lib/irq_poll.c
index 2f17b488d58e1..484dbd2aa09fb 100644
--- a/lib/irq_poll.c
+++ b/lib/irq_poll.c
@@ -194,7 +194,8 @@ static int irq_poll_cpu_dead(unsigned int cpu)
 	local_irq_disable();
 	list_splice_init(&per_cpu(blk_cpu_iopoll, cpu),
 			 this_cpu_ptr(&blk_cpu_iopoll));
-	__raise_softirq_irqoff(IRQ_POLL_SOFTIRQ);
+	if (!list_empty(this_cpu_ptr(&blk_cpu_iopoll)))
+		raise_softirq_irqoff(IRQ_POLL_SOFTIRQ);
 	local_irq_enable();
 
 	return 0;
-- 
2.26.0

