[PATCH] sched/nohz: add debugfs control over sched_tick_max_deferment

[PATCH] sched/nohz: add debugfs control over sched_tick_max_deferment

[Kevin Hilman]

From: Kevin Hilman <khilman@linaro.org>

Allow debugfs override of sched_tick_max_deferment in order to ease
finding/fixing the remaining issues with full nohz.

The value to be written is in jiffies, and -1 means the max deferment
is disabled (scheduler_tick_max_deferment() returns KTIME_MAX.)

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Kevin Hilman <khilman@linaro.org>
---
 kernel/sched/core.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index bc1638b33449..dee044a5d447 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2514,6 +2514,8 @@ void scheduler_tick(void)
 }
 
 #ifdef CONFIG_NO_HZ_FULL
+static u32 sched_tick_max_deferment = HZ;
+
 /**
  * scheduler_tick_max_deferment
  *
@@ -2532,13 +2534,25 @@ u64 scheduler_tick_max_deferment(void)
 	struct rq *rq = this_rq();
 	unsigned long next, now = ACCESS_ONCE(jiffies);
 
-	next = rq->last_sched_tick + HZ;
+	if (sched_tick_max_deferment == -1)
+		return KTIME_MAX;
+
+	next = rq->last_sched_tick + sched_tick_max_deferment;
 
 	if (time_before_eq(next, now))
 		return 0;
 
 	return jiffies_to_nsecs(next - now);
 }
+
+static __init int sched_nohz_full_init_debug(void)
+{
+	debugfs_create_u32("sched_tick_max_deferment", 0644, NULL,
+			   &sched_tick_max_deferment);
+
+	return 0;
+}
+late_initcall(sched_nohz_full_init_debug);
 #endif
 
 notrace unsigned long get_parent_ip(unsigned long addr)
-- 
2.1.0

[PATCH] sched/nohz: add debugfs control over sched_tick_max_deferment

[Kevin Hilman]

Hi Frederic,

On Fri, Sep 26, 2014 at 12:45 PM, Kevin Hilman <khilman@kernel.org> wrote:
> From: Kevin Hilman <khilman@linaro.org>
>
> Allow debugfs override of sched_tick_max_deferment in order to ease
> finding/fixing the remaining issues with full nohz.
>
> The value to be written is in jiffies, and -1 means the max deferment
> is disabled (scheduler_tick_max_deferment() returns KTIME_MAX.)
>
> Cc: Frederic Weisbecker <fweisbec@gmail.com>
> Signed-off-by: Kevin Hilman <khilman@linaro.org>

This is just an attempt to revive and hopefully finalize the
discussion started a while back[1].

Kevin

[1] https://lkml.org/lkml/2014/1/14/914

[PATCH] sched/nohz: add debugfs control over sched_tick_max_deferment

[Frederic Weisbecker]

On Fri, Sep 26, 2014 at 12:45:32PM -0700, Kevin Hilman wrote:
> From: Kevin Hilman <khilman@linaro.org>
> 
> Allow debugfs override of sched_tick_max_deferment in order to ease
> finding/fixing the remaining issues with full nohz.
> 
> The value to be written is in jiffies, and -1 means the max deferment
> is disabled (scheduler_tick_max_deferment() returns KTIME_MAX.)
> 
> Cc: Frederic Weisbecker <fweisbec@gmail.com>
> Signed-off-by: Kevin Hilman <khilman@linaro.org>

So, I'm worried that it becomes a hack that everybody uses to shutdown
the tick completely then nobody will come and fix the issue that prevents
from doing it properly.

I seriously doubt this will be used for development purpose to help
fixing the real problem. Quite the opposite. If developers want to do
testing, they can as well comment out the call to scheduler_max_tick_deferment().


> ---
>  kernel/sched/core.c | 16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index bc1638b33449..dee044a5d447 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -2514,6 +2514,8 @@ void scheduler_tick(void)
>  }
>  
>  #ifdef CONFIG_NO_HZ_FULL
> +static u32 sched_tick_max_deferment = HZ;
> +
>  /**
>   * scheduler_tick_max_deferment
>   *
> @@ -2532,13 +2534,25 @@ u64 scheduler_tick_max_deferment(void)
>  	struct rq *rq = this_rq();
>  	unsigned long next, now = ACCESS_ONCE(jiffies);
>  
> -	next = rq->last_sched_tick + HZ;
> +	if (sched_tick_max_deferment == -1)
> +		return KTIME_MAX;
> +
> +	next = rq->last_sched_tick + sched_tick_max_deferment;
>  
>  	if (time_before_eq(next, now))
>  		return 0;
>  
>  	return jiffies_to_nsecs(next - now);
>  }
> +
> +static __init int sched_nohz_full_init_debug(void)
> +{
> +	debugfs_create_u32("sched_tick_max_deferment", 0644, NULL,
> +			   &sched_tick_max_deferment);
> +
> +	return 0;
> +}
> +late_initcall(sched_nohz_full_init_debug);
>  #endif
>  
>  notrace unsigned long get_parent_ip(unsigned long addr)
> -- 
> 2.1.0
> 

[PATCH] sched/nohz: add debugfs control over sched_tick_max_deferment

[Kevin Hilman]

Frederic Weisbecker <fweisbec@gmail.com> writes:

> On Fri, Sep 26, 2014 at 12:45:32PM -0700, Kevin Hilman wrote:
>> From: Kevin Hilman <khilman@linaro.org>
>> 
>> Allow debugfs override of sched_tick_max_deferment in order to ease
>> finding/fixing the remaining issues with full nohz.
>> 
>> The value to be written is in jiffies, and -1 means the max deferment
>> is disabled (scheduler_tick_max_deferment() returns KTIME_MAX.)
>> 
>> Cc: Frederic Weisbecker <fweisbec@gmail.com>
>> Signed-off-by: Kevin Hilman <khilman@linaro.org>
>
> So, I'm worried that it becomes a hack that everybody uses to shutdown
> the tick completely then nobody will come and fix the issue that prevents
> from doing it properly.

OK.

> I seriously doubt this will be used for development purpose to help
> fixing the real problem. 

Well, that's what I was using it for, until I was distracted by other
tasks.

> Quite the opposite. If developers want to do
> testing, they can as well comment out the call to scheduler_max_tick_deferment().

OK, fair enough.

Thanks,

Kevin