Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1548795

[patch-rt] softirq: Move ksoftirqd_running() under !CONFIG_PREEMPT_RT_FULL

From Mike Galbraith <efault@gmx.de>
Newsgroups linux.kernel
Subject [patch-rt] softirq: Move ksoftirqd_running() under !CONFIG_PREEMPT_RT_FULL
Date 2016-12-31 09:30 +0100
Message-ID <sUnd8-va-7@gated-at.bofh.it> (permalink)
References <sRB2V-Pv-3@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw


Otherwise, ktimersoftd may not be awakened when it has work to do.

[   84.087571] NOHZ: local_softirq_pending 02
[   84.087593] NOHZ: local_softirq_pending 02
[   84.087598] NOHZ: local_softirq_pending 02
[   84.087904] NOHZ: local_softirq_pending 02
[   84.088526] NOHZ: local_softirq_pending 02
[   84.088899] NOHZ: local_softirq_pending 02
[   84.089463] NOHZ: local_softirq_pending 02
[  115.013470] NOHZ: local_softirq_pending 02
[  115.013601] NOHZ: local_softirq_pending 02
[  115.013709] NOHZ: local_softirq_pending 02

Signed-off-by: Mike Galbraith <efault@gmx.de>
---
 kernel/softirq.c |   10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -206,6 +206,7 @@ static void handle_softirq(unsigned int
 	}
 }
 
+#ifndef CONFIG_PREEMPT_RT_FULL
 /*
  * If ksoftirqd is scheduled, we do not want to process pending softirqs
  * right now. Let ksoftirqd handle this at its own rate, to get fairness.
@@ -217,7 +218,6 @@ static bool ksoftirqd_running(void)
 	return tsk && (tsk->state == TASK_RUNNING);
 }
 
-#ifndef CONFIG_PREEMPT_RT_FULL
 static inline int ksoftirqd_softirq_pending(void)
 {
 	return local_softirq_pending();
@@ -794,13 +794,10 @@ void irq_enter(void)
 
 static inline void invoke_softirq(void)
 {
-#ifdef CONFIG_PREEMPT_RT_FULL
-	unsigned long flags;
-#endif
-
+#ifndef CONFIG_PREEMPT_RT_FULL
 	if (ksoftirqd_running())
 		return;
-#ifndef CONFIG_PREEMPT_RT_FULL
+
 	if (!force_irqthreads) {
 #ifdef CONFIG_HAVE_IRQ_EXIT_ON_IRQ_STACK
 		/*
@@ -821,6 +818,7 @@ static inline void invoke_softirq(void)
 		wakeup_softirqd();
 	}
 #else /* PREEMPT_RT_FULL */
+	unsigned long flags;
 
 	local_irq_save(flags);
 	if (__this_cpu_read(ksoftirqd) &&

Back to linux.kernel | Previous | NextPrevious in thread | Find similar | Unroll thread


Thread

[ANNOUNCE] v4.9-rt1 Sebastian Andrzej Siewior <bigeasy@linutronix.de> - 2016-12-23 17:40 +0100
  [patch-rt] kvm: Convert pvclock_gtod_sync_lock to raw_spinlock_t Mike Galbraith <efault@gmx.de> - 2016-12-26 08:00 +0100
  [rfc patch-rt] posix_cpu_timers: Kill hotplug cpu notifier Mike Galbraith <efault@gmx.de> - 2016-12-26 08:10 +0100
  [patch-rt] softirq: Move ksoftirqd_running() under  !CONFIG_PREEMPT_RT_FULL Mike Galbraith <efault@gmx.de> - 2016-12-31 09:30 +0100

csiph-web