All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] clocksource: don't run watchdog forever
@ 2021-03-02  2:54 Feng Tang
  2021-03-02  9:16 ` Peter Zijlstra
  0 siblings, 1 reply; 9+ messages in thread
From: Feng Tang @ 2021-03-02  2:54 UTC (permalink / raw)
  To: John Stultz, Thomas Gleixner, Stephen Boyd, linux-kernel
  Cc: Qais Yousef, Peter Zijlstra, andi.kleen, Feng Tang

clocksource watchdog runs every 500ms, which creates some OS noise.
As the clocksource wreckage (especially for those that has per-cpu
reading hook) usually happens shortly after CPU is brought up or
after system resumes from sleep state, so add a time limit for
clocksource watchdog to only run for a period of time, and make
sure it run at least twice for each CPU.

Regarding performance data, there is no improvement data with the
micro-benchmarks we have like hackbench/netperf/fio/will-it-scale
etc. But it obviously reduces periodic timer interrupts, and may
help in following cases:
* When some CPUs are isolated to only run scientific or high
  performance computing tasks on a NOHZ_FULL kernel, where there
  is almost no interrupts, this could make it more quiet
* On a cluster which runs a lot of systems in parallel with
  barriers there are always enough systems which run the watchdog
  and make everyone else wait

Signed-off-by: Feng Tang <feng.tang@intel.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
---
 include/linux/clocksource.h |  7 +++++++
 kernel/cpu.c                |  3 +++
 kernel/time/clocksource.c   | 31 +++++++++++++++++++++++++++++--
 3 files changed, 39 insertions(+), 2 deletions(-)

diff --git a/include/linux/clocksource.h b/include/linux/clocksource.h
index 86d143d..cf428a2 100644
--- a/include/linux/clocksource.h
+++ b/include/linux/clocksource.h
@@ -252,6 +252,13 @@ static inline void __clocksource_update_freq_khz(struct clocksource *cs, u32 khz
 	__clocksource_update_freq_scale(cs, 1000, khz);
 }
 
+#ifdef CONFIG_CLOCKSOURCE_WATCHDOG
+extern void clocksource_kick_watchdog(void);
+#else
+static inline void clocksource_kick_watchdog(void) { }
+#endif
+
+
 #ifdef CONFIG_ARCH_CLOCKSOURCE_INIT
 extern void clocksource_arch_init(struct clocksource *cs);
 #else
diff --git a/kernel/cpu.c b/kernel/cpu.c
index 1b6302e..fdf3c69 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -32,6 +32,7 @@
 #include <linux/relay.h>
 #include <linux/slab.h>
 #include <linux/percpu-rwsem.h>
+#include <linux/clocksource.h>
 
 #include <trace/events/power.h>
 #define CREATE_TRACE_POINTS
@@ -1286,6 +1287,8 @@ static int cpu_up(unsigned int cpu, enum cpuhp_state target)
 	}
 
 	err = _cpu_up(cpu, 0, target);
+
+	clocksource_kick_watchdog();
 out:
 	cpu_maps_update_done();
 	return err;
diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c
index cce484a..aba985a 100644
--- a/kernel/time/clocksource.c
+++ b/kernel/time/clocksource.c
@@ -104,6 +104,7 @@ static DECLARE_WORK(watchdog_work, clocksource_watchdog_work);
 static DEFINE_SPINLOCK(watchdog_lock);
 static int watchdog_running;
 static atomic_t watchdog_reset_pending;
+static unsigned long watchdog_stop_time;	/* in jiffies */
 
 static inline void clocksource_watchdog_lock(unsigned long *flags)
 {
@@ -295,10 +296,16 @@ static void clocksource_watchdog(struct timer_list *unused)
 		next_cpu = cpumask_first(cpu_online_mask);
 
 	/*
-	 * Arm timer if not already pending: could race with concurrent
-	 * pair clocksource_stop_watchdog() clocksource_start_watchdog().
+	 * Arm timer if not already pending or pass the check time window:
+	 * could race with concurrent pair clocksource_stop_watchdog()
+	 * clocksource_start_watchdog().
 	 */
 	if (!timer_pending(&watchdog_timer)) {
+		if (time_after(jiffies, watchdog_stop_time)) {
+			atomic_inc(&watchdog_reset_pending);
+			watchdog_running = 0;
+			goto out;
+		}
 		watchdog_timer.expires += WATCHDOG_INTERVAL;
 		add_timer_on(&watchdog_timer, next_cpu);
 	}
@@ -308,6 +315,16 @@ static void clocksource_watchdog(struct timer_list *unused)
 
 static inline void clocksource_start_watchdog(void)
 {
+	unsigned long check_ticks;
+
+	/*
+	 * As all CPUs will be looped to run the timer, make sure each
+	 * CPU can run the timer twice, and the check run for at least
+	 * 10 minutes.
+	 */
+	check_ticks = max_t(unsigned long, num_possible_cpus(), 600) * HZ;
+	watchdog_stop_time = jiffies + check_ticks;
+
 	if (watchdog_running || !watchdog || list_empty(&watchdog_list))
 		return;
 	timer_setup(&watchdog_timer, clocksource_watchdog, 0);
@@ -324,6 +341,15 @@ static inline void clocksource_stop_watchdog(void)
 	watchdog_running = 0;
 }
 
+void clocksource_kick_watchdog(void)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&watchdog_lock, flags);
+	clocksource_start_watchdog();
+	spin_unlock_irqrestore(&watchdog_lock, flags);
+}
+
 static inline void clocksource_reset_watchdog(void)
 {
 	struct clocksource *cs;
@@ -618,6 +644,7 @@ void clocksource_resume(void)
 			cs->resume(cs);
 
 	clocksource_resume_watchdog();
+	clocksource_kick_watchdog();
 }
 
 /**
-- 
2.7.4


^ permalink raw reply related	[flat|nested] 9+ messages in thread
end of thread, other threads:[~2021-03-25 11:43 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2021-03-02  2:54 [PATCH] clocksource: don't run watchdog forever Feng Tang
2021-03-02  9:16 ` Peter Zijlstra
2021-03-02 12:06   ` Feng Tang
2021-03-03 15:50     ` Thomas Gleixner
2021-03-04  7:43       ` Feng Tang
2021-03-04 14:15         ` Thomas Gleixner
2021-03-05  2:30           ` Feng Tang
2021-03-25  8:34       ` Feng Tang
2021-03-25 11:40         ` Thomas Gleixner

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.