[PATCH] cpufreq: timer: Move tick-sched specific code outside of cpufreq governors

[PATCH] cpufreq: timer: Move tick-sched specific code outside of cpufreq governors

[Viresh Kumar]

Multiple cpufreq governers have defined similar get_cpu_idle_time_***()
routines. These routines must be moved to some common place, so that all
governors can use them.

So moving them to tick-sched.c, which seems to be a better place for keeping
these routines.

Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
---
 drivers/cpufreq/cpufreq_conservative.c | 34 ---------------------------------
 drivers/cpufreq/cpufreq_ondemand.c     | 34 ---------------------------------
 include/linux/tick.h                   |  3 +++
 kernel/time/tick-sched.c               | 35 ++++++++++++++++++++++++++++++++++
 4 files changed, 38 insertions(+), 68 deletions(-)

diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c
index a152af7..181abad 100644
--- a/drivers/cpufreq/cpufreq_conservative.c
+++ b/drivers/cpufreq/cpufreq_conservative.c
@@ -95,40 +95,6 @@ static struct dbs_tuners {
 	.freq_step = 5,
 };
 
-static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
-{
-	u64 idle_time;
-	u64 cur_wall_time;
-	u64 busy_time;
-
-	cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
-
-	busy_time  = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
-
-	idle_time = cur_wall_time - busy_time;
-	if (wall)
-		*wall = jiffies_to_usecs(cur_wall_time);
-
-	return jiffies_to_usecs(idle_time);
-}
-
-static inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
-{
-	u64 idle_time = get_cpu_idle_time_us(cpu, NULL);
-
-	if (idle_time == -1ULL)
-		return get_cpu_idle_time_jiffy(cpu, wall);
-	else
-		idle_time += get_cpu_iowait_time_us(cpu, wall);
-
-	return idle_time;
-}
-
 /* keep track of frequency transitions */
 static int
 dbs_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
index 396322f..d7f774b 100644
--- a/drivers/cpufreq/cpufreq_ondemand.c
+++ b/drivers/cpufreq/cpufreq_ondemand.c
@@ -119,40 +119,6 @@ static struct dbs_tuners {
 	.powersave_bias = 0,
 };
 
-static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
-{
-	u64 idle_time;
-	u64 cur_wall_time;
-	u64 busy_time;
-
-	cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
-
-	busy_time  = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
-	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
-
-	idle_time = cur_wall_time - busy_time;
-	if (wall)
-		*wall = jiffies_to_usecs(cur_wall_time);
-
-	return jiffies_to_usecs(idle_time);
-}
-
-static inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
-{
-	u64 idle_time = get_cpu_idle_time_us(cpu, NULL);
-
-	if (idle_time == -1ULL)
-		return get_cpu_idle_time_jiffy(cpu, wall);
-	else
-		idle_time += get_cpu_iowait_time_us(cpu, wall);
-
-	return idle_time;
-}
-
 static inline cputime64_t get_cpu_iowait_time(unsigned int cpu, cputime64_t *wall)
 {
 	u64 iowait_time = get_cpu_iowait_time_us(cpu, wall);
diff --git a/include/linux/tick.h b/include/linux/tick.h
index f37fceb..79ca824 100644
--- a/include/linux/tick.h
+++ b/include/linux/tick.h
@@ -6,6 +6,7 @@
 #ifndef _LINUX_TICK_H
 #define _LINUX_TICK_H
 
+#include <asm/cputime.h>
 #include <linux/clockchips.h>
 #include <linux/irqflags.h>
 
@@ -142,4 +143,6 @@ static inline u64 get_cpu_idle_time_us(int cpu, u64 *unused) { return -1; }
 static inline u64 get_cpu_iowait_time_us(int cpu, u64 *unused) { return -1; }
 # endif /* !NO_HZ */
 
+cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall);
+
 #endif
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index a402608..b458402 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -271,6 +271,41 @@ u64 get_cpu_iowait_time_us(int cpu, u64 *last_update_time)
 }
 EXPORT_SYMBOL_GPL(get_cpu_iowait_time_us);
 
+static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
+{
+	u64 idle_time;
+	u64 cur_wall_time;
+	u64 busy_time;
+
+	cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
+
+	busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
+	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
+	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
+	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
+	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
+	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
+
+	idle_time = cur_wall_time - busy_time;
+	if (wall)
+		*wall = jiffies_to_usecs(cur_wall_time);
+
+	return jiffies_to_usecs(idle_time);
+}
+
+cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
+{
+	u64 idle_time = get_cpu_idle_time_us(cpu, NULL);
+
+	if (idle_time == -1ULL)
+		return get_cpu_idle_time_jiffy(cpu, wall);
+	else
+		idle_time += get_cpu_iowait_time_us(cpu, wall);
+
+	return idle_time;
+}
+EXPORT_SYMBOL_GPL(get_cpu_idle_time);
+
 static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 					 ktime_t now, int cpu)
 {
-- 
1.7.12.rc2.18.g61b472e

Re: [PATCH] cpufreq: timer: Move tick-sched specific code outside of cpufreq governors

[Michal Hocko]

On Mon 15-10-12 13:41:20, Viresh Kumar wrote:
> Multiple cpufreq governers have defined similar get_cpu_idle_time_***()
> routines. These routines must be moved to some common place, so that all
> governors can use them.
> 
> So moving them to tick-sched.c, which seems to be a better place for keeping
> these routines.

I do agree that this code duplication should be removed but tick-sched.c
is not a right place IMO. Who, apart from governors, should use those
"common" functions?
Having a generic get_cpu_idle_time, which in fact includes iowait time
as well is definitely not good. It is confusing and it doesn't match
get_cpu_idle_time_us.
I would suggest moving the common functionality into drivers/cpufreq/
(e.g. cpufreq_common.c).

> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
> ---
>  drivers/cpufreq/cpufreq_conservative.c | 34 ---------------------------------
>  drivers/cpufreq/cpufreq_ondemand.c     | 34 ---------------------------------
>  include/linux/tick.h                   |  3 +++
>  kernel/time/tick-sched.c               | 35 ++++++++++++++++++++++++++++++++++
>  4 files changed, 38 insertions(+), 68 deletions(-)
> 
> diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c
> index a152af7..181abad 100644
> --- a/drivers/cpufreq/cpufreq_conservative.c
> +++ b/drivers/cpufreq/cpufreq_conservative.c
> @@ -95,40 +95,6 @@ static struct dbs_tuners {
>  	.freq_step = 5,
>  };
>  
> -static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
> -{
> -	u64 idle_time;
> -	u64 cur_wall_time;
> -	u64 busy_time;
> -
> -	cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
> -
> -	busy_time  = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
> -
> -	idle_time = cur_wall_time - busy_time;
> -	if (wall)
> -		*wall = jiffies_to_usecs(cur_wall_time);
> -
> -	return jiffies_to_usecs(idle_time);
> -}
> -
> -static inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
> -{
> -	u64 idle_time = get_cpu_idle_time_us(cpu, NULL);
> -
> -	if (idle_time == -1ULL)
> -		return get_cpu_idle_time_jiffy(cpu, wall);
> -	else
> -		idle_time += get_cpu_iowait_time_us(cpu, wall);
> -
> -	return idle_time;
> -}
> -
>  /* keep track of frequency transitions */
>  static int
>  dbs_cpufreq_notifier(struct notifier_block *nb, unsigned long val,
> diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
> index 396322f..d7f774b 100644
> --- a/drivers/cpufreq/cpufreq_ondemand.c
> +++ b/drivers/cpufreq/cpufreq_ondemand.c
> @@ -119,40 +119,6 @@ static struct dbs_tuners {
>  	.powersave_bias = 0,
>  };
>  
> -static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
> -{
> -	u64 idle_time;
> -	u64 cur_wall_time;
> -	u64 busy_time;
> -
> -	cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
> -
> -	busy_time  = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
> -	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
> -
> -	idle_time = cur_wall_time - busy_time;
> -	if (wall)
> -		*wall = jiffies_to_usecs(cur_wall_time);
> -
> -	return jiffies_to_usecs(idle_time);
> -}
> -
> -static inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
> -{
> -	u64 idle_time = get_cpu_idle_time_us(cpu, NULL);
> -
> -	if (idle_time == -1ULL)
> -		return get_cpu_idle_time_jiffy(cpu, wall);
> -	else
> -		idle_time += get_cpu_iowait_time_us(cpu, wall);
> -
> -	return idle_time;
> -}
> -
>  static inline cputime64_t get_cpu_iowait_time(unsigned int cpu, cputime64_t *wall)
>  {
>  	u64 iowait_time = get_cpu_iowait_time_us(cpu, wall);
> diff --git a/include/linux/tick.h b/include/linux/tick.h
> index f37fceb..79ca824 100644
> --- a/include/linux/tick.h
> +++ b/include/linux/tick.h
> @@ -6,6 +6,7 @@
>  #ifndef _LINUX_TICK_H
>  #define _LINUX_TICK_H
>  
> +#include <asm/cputime.h>
>  #include <linux/clockchips.h>
>  #include <linux/irqflags.h>
>  
> @@ -142,4 +143,6 @@ static inline u64 get_cpu_idle_time_us(int cpu, u64 *unused) { return -1; }
>  static inline u64 get_cpu_iowait_time_us(int cpu, u64 *unused) { return -1; }
>  # endif /* !NO_HZ */
>  
> +cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall);
> +
>  #endif
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index a402608..b458402 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -271,6 +271,41 @@ u64 get_cpu_iowait_time_us(int cpu, u64 *last_update_time)
>  }
>  EXPORT_SYMBOL_GPL(get_cpu_iowait_time_us);
>  
> +static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
> +{
> +	u64 idle_time;
> +	u64 cur_wall_time;
> +	u64 busy_time;
> +
> +	cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
> +
> +	busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
> +	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
> +	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
> +	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
> +	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
> +	busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
> +
> +	idle_time = cur_wall_time - busy_time;
> +	if (wall)
> +		*wall = jiffies_to_usecs(cur_wall_time);
> +
> +	return jiffies_to_usecs(idle_time);
> +}
> +
> +cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
> +{
> +	u64 idle_time = get_cpu_idle_time_us(cpu, NULL);
> +
> +	if (idle_time == -1ULL)
> +		return get_cpu_idle_time_jiffy(cpu, wall);
> +	else
> +		idle_time += get_cpu_iowait_time_us(cpu, wall);
> +
> +	return idle_time;
> +}
> +EXPORT_SYMBOL_GPL(get_cpu_idle_time);
> +
>  static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
>  					 ktime_t now, int cpu)
>  {
> -- 
> 1.7.12.rc2.18.g61b472e
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
Michal Hocko
SUSE Labs

Re: [PATCH] cpufreq: timer: Move tick-sched specific code outside of cpufreq governors

[Viresh Kumar]

On 15 October 2012 14:05, Michal Hocko <mhocko@suse.cz> wrote:
> On Mon 15-10-12 13:41:20, Viresh Kumar wrote:
>> Multiple cpufreq governers have defined similar get_cpu_idle_time_***()
>> routines. These routines must be moved to some common place, so that all
>> governors can use them.
>>
>> So moving them to tick-sched.c, which seems to be a better place for keeping
>> these routines.
>
> I do agree that this code duplication should be removed but tick-sched.c
> is not a right place IMO. Who, apart from governors, should use those
> "common" functions?

Nobody leaving cpufreq for now.

> Having a generic get_cpu_idle_time, which in fact includes iowait time
> as well is definitely not good. It is confusing and it doesn't match
> get_cpu_idle_time_us.

ok

> I would suggest moving the common functionality into drivers/cpufreq/
> (e.g. cpufreq_common.c).

Initially i did that only, but then thought these routines must be present in
more generic files if possible, available across frameworks.

Can we try renaming these to show there exact functionality and then put
them in generic files like, tick-sched.c?

--
viresh

Re: [PATCH] cpufreq: timer: Move tick-sched specific code outside of cpufreq governors

[Michal Hocko]

On Mon 15-10-12 14:11:53, Viresh Kumar wrote:
> On 15 October 2012 14:05, Michal Hocko <mhocko@suse.cz> wrote:
[...]
> > I would suggest moving the common functionality into drivers/cpufreq/
> > (e.g. cpufreq_common.c).
> 
> Initially i did that only, but then thought these routines must be present in
> more generic files if possible, available across frameworks.
> 
> Can we try renaming these to show there exact functionality and then put
> them in generic files like, tick-sched.c?

But we already do have generic generic functionality for nohz case
get_cpu_{idle,iowait}_time_us and kcpustat_cpu for !nohz. The only one
that cares for both is /proc/stat code and it has its own helpers
get_{idle,iowait}_time and as you can see it's demands are different
from what governors want.

-- 
Michal Hocko
SUSE Labs