2018-07-20 22:23:24

by Joel Fernandes

Subject: [PATCH v2] cpufreq: trace frequency limits change

From: Ruchi Kandoi <[email protected]>

systrace used for tracing for Android systems has carried a patch for
many years in the Android tree that traces when the cpufreq limits
change. With the help of this information, systrace can know when the
policy limits change and can visually display the data. Lets add
upstream support for the same.

Signed-off-by: Ruchi Kandoi <[email protected]>
Signed-off-by: Joel Fernandes (Google) <[email protected]>
---
v1->v2: Minor changes suggested by Viresh

Documentation/trace/events-power.rst | 1 +
drivers/cpufreq/cpufreq.c | 1 +
include/trace/events/power.h | 25 +++++++++++++++++++++++++
3 files changed, 27 insertions(+)

diff --git a/Documentation/trace/events-power.rst b/Documentation/trace/events-power.rst
index a77daca75e30..2ef318962e29 100644
--- a/Documentation/trace/events-power.rst
+++ b/Documentation/trace/events-power.rst
@@ -27,6 +27,7 @@ cpufreq.

cpu_idle "state=%lu cpu_id=%lu"
cpu_frequency "state=%lu cpu_id=%lu"
+ cpu_frequency_limits "min=%lu max=%lu cpu_id=%lu"

A suspend event is used to indicate the system going in and out of the
suspend mode:
diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index b0dfd3222013..4fd935df101e 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -2236,6 +2236,7 @@ static int cpufreq_set_policy(struct cpufreq_policy *policy,

policy->min = new_policy->min;
policy->max = new_policy->max;
+ trace_cpu_frequency_limits(policy->min, policy->max, policy->cpu);

policy->cached_target_freq = UINT_MAX;

diff --git a/include/trace/events/power.h b/include/trace/events/power.h
index 908977d69783..f5bec45108b6 100644
--- a/include/trace/events/power.h
+++ b/include/trace/events/power.h
@@ -148,6 +148,31 @@ DEFINE_EVENT(cpu, cpu_frequency,
TP_ARGS(frequency, cpu_id)
);

+TRACE_EVENT(cpu_frequency_limits,
+
+ TP_PROTO(unsigned int min_freq, unsigned int max_freq,
+ unsigned int cpu_id),
+
+ TP_ARGS(min_freq, max_freq, cpu_id),
+
+ TP_STRUCT__entry(
+ __field(u32, min_freq)
+ __field(u32, max_freq)
+ __field(u32, cpu_id)
+ ),
+
+ TP_fast_assign(
+ __entry->min_freq = min_freq;
+ __entry->max_freq = max_freq;
+ __entry->cpu_id = cpu_id;
+ ),
+
+ TP_printk("min=%lu max=%lu cpu_id=%lu",
+ (unsigned long)__entry->min_freq,
+ (unsigned long)__entry->max_freq,
+ (unsigned long)__entry->cpu_id)
+);
+
TRACE_EVENT(device_pm_callback_start,

TP_PROTO(struct device *dev, const char *pm_ops, int event),
--
2.18.0.233.g985f88cf7e-goog

2018-07-23 03:02:35

by Viresh Kumar

Subject: Re: [PATCH v2] cpufreq: trace frequency limits change

On 20-07-18, 15:21, Joel Fernandes wrote:
> From: Ruchi Kandoi <[email protected]>
>
> systrace used for tracing for Android systems has carried a patch for
> many years in the Android tree that traces when the cpufreq limits
> change. With the help of this information, systrace can know when the
> policy limits change and can visually display the data. Lets add
> upstream support for the same.
>
> Signed-off-by: Ruchi Kandoi <[email protected]>
> Signed-off-by: Joel Fernandes (Google) <[email protected]>
> ---
> v1->v2: Minor changes suggested by Viresh
>
> Documentation/trace/events-power.rst | 1 +
> drivers/cpufreq/cpufreq.c | 1 +
> include/trace/events/power.h | 25 +++++++++++++++++++++++++
> 3 files changed, 27 insertions(+)
>
> diff --git a/Documentation/trace/events-power.rst b/Documentation/trace/events-power.rst
> index a77daca75e30..2ef318962e29 100644
> --- a/Documentation/trace/events-power.rst
> +++ b/Documentation/trace/events-power.rst
> @@ -27,6 +27,7 @@ cpufreq.
>
> cpu_idle "state=%lu cpu_id=%lu"
> cpu_frequency "state=%lu cpu_id=%lu"
> + cpu_frequency_limits "min=%lu max=%lu cpu_id=%lu"
>
> A suspend event is used to indicate the system going in and out of the
> suspend mode:
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index b0dfd3222013..4fd935df101e 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -2236,6 +2236,7 @@ static int cpufreq_set_policy(struct cpufreq_policy *policy,
>
> policy->min = new_policy->min;
> policy->max = new_policy->max;
> + trace_cpu_frequency_limits(policy->min, policy->max, policy->cpu);
>
> policy->cached_target_freq = UINT_MAX;
>
> diff --git a/include/trace/events/power.h b/include/trace/events/power.h
> index 908977d69783..f5bec45108b6 100644
> --- a/include/trace/events/power.h
> +++ b/include/trace/events/power.h
> @@ -148,6 +148,31 @@ DEFINE_EVENT(cpu, cpu_frequency,
> TP_ARGS(frequency, cpu_id)
> );
>
> +TRACE_EVENT(cpu_frequency_limits,
> +
> + TP_PROTO(unsigned int min_freq, unsigned int max_freq,
> + unsigned int cpu_id),
> +
> + TP_ARGS(min_freq, max_freq, cpu_id),
> +
> + TP_STRUCT__entry(
> + __field(u32, min_freq)
> + __field(u32, max_freq)
> + __field(u32, cpu_id)
> + ),
> +
> + TP_fast_assign(
> + __entry->min_freq = min_freq;
> + __entry->max_freq = max_freq;
> + __entry->cpu_id = cpu_id;
> + ),
> +
> + TP_printk("min=%lu max=%lu cpu_id=%lu",
> + (unsigned long)__entry->min_freq,
> + (unsigned long)__entry->max_freq,
> + (unsigned long)__entry->cpu_id)
> +);
> +
> TRACE_EVENT(device_pm_callback_start,
>
> TP_PROTO(struct device *dev, const char *pm_ops, int event),

Acked-by: Viresh Kumar <[email protected]>

--
viresh

2018-07-24 13:51:11

by Steven Rostedt

Subject: Re: [PATCH v2] cpufreq: trace frequency limits change

On Fri, 20 Jul 2018 15:21:14 -0700
Joel Fernandes <[email protected]> wrote:

> From: Ruchi Kandoi <[email protected]>
>
> systrace used for tracing for Android systems has carried a patch for
> many years in the Android tree that traces when the cpufreq limits
> change. With the help of this information, systrace can know when the
> policy limits change and can visually display the data. Lets add
> upstream support for the same.
>
> Signed-off-by: Ruchi Kandoi <[email protected]>
> Signed-off-by: Joel Fernandes (Google) <[email protected]>
> ---
> v1->v2: Minor changes suggested by Viresh
>
> Documentation/trace/events-power.rst | 1 +
> drivers/cpufreq/cpufreq.c | 1 +
> include/trace/events/power.h | 25 +++++++++++++++++++++++++
> 3 files changed, 27 insertions(+)
>
> diff --git a/Documentation/trace/events-power.rst b/Documentation/trace/events-power.rst
> index a77daca75e30..2ef318962e29 100644
> --- a/Documentation/trace/events-power.rst
> +++ b/Documentation/trace/events-power.rst
> @@ -27,6 +27,7 @@ cpufreq.
>
> cpu_idle "state=%lu cpu_id=%lu"
> cpu_frequency "state=%lu cpu_id=%lu"
> + cpu_frequency_limits "min=%lu max=%lu cpu_id=%lu"
>
> A suspend event is used to indicate the system going in and out of the
> suspend mode:
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index b0dfd3222013..4fd935df101e 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -2236,6 +2236,7 @@ static int cpufreq_set_policy(struct cpufreq_policy *policy,
>
> policy->min = new_policy->min;
> policy->max = new_policy->max;
> + trace_cpu_frequency_limits(policy->min, policy->max, policy->cpu);

Wouldn't it make more sense just to pass in "policy"?

>
> policy->cached_target_freq = UINT_MAX;
>
> diff --git a/include/trace/events/power.h b/include/trace/events/power.h
> index 908977d69783..f5bec45108b6 100644
> --- a/include/trace/events/power.h
> +++ b/include/trace/events/power.h
> @@ -148,6 +148,31 @@ DEFINE_EVENT(cpu, cpu_frequency,
> TP_ARGS(frequency, cpu_id)
> );
>
> +TRACE_EVENT(cpu_frequency_limits,
> +
> + TP_PROTO(unsigned int min_freq, unsigned int max_freq,
> + unsigned int cpu_id),
> +
> + TP_ARGS(min_freq, max_freq, cpu_id),
> +
> + TP_STRUCT__entry(
> + __field(u32, min_freq)
> + __field(u32, max_freq)
> + __field(u32, cpu_id)
> + ),
> +
> + TP_fast_assign(
> + __entry->min_freq = min_freq;
> + __entry->max_freq = max_freq;
> + __entry->cpu_id = cpu_id;

Then have here:

__entry->min_freq = policy->min;
__entry->max_freq = policy->max;
__entry->puc_id = policy->cpu;

It would also make the footprint of the tracepoint in the code smaller
as it would pass fewer parameters to the trace event.

-- Steve

> + ),
> +
> + TP_printk("min=%lu max=%lu cpu_id=%lu",
> + (unsigned long)__entry->min_freq,
> + (unsigned long)__entry->max_freq,
> + (unsigned long)__entry->cpu_id)
> +);
> +
> TRACE_EVENT(device_pm_callback_start,
>
> TP_PROTO(struct device *dev, const char *pm_ops, int event),

2018-07-24 17:25:50

by Joel Fernandes

Subject: Re: [PATCH v2] cpufreq: trace frequency limits change

On Tue, Jul 24, 2018 at 09:49:32AM -0400, Steven Rostedt wrote:
> On Fri, 20 Jul 2018 15:21:14 -0700
> Joel Fernandes <[email protected]> wrote:
>
> > From: Ruchi Kandoi <[email protected]>
> >
> > systrace used for tracing for Android systems has carried a patch for
> > many years in the Android tree that traces when the cpufreq limits
> > change. With the help of this information, systrace can know when the
> > policy limits change and can visually display the data. Lets add
> > upstream support for the same.
> >
> > Signed-off-by: Ruchi Kandoi <[email protected]>
> > Signed-off-by: Joel Fernandes (Google) <[email protected]>
> > ---
> > v1->v2: Minor changes suggested by Viresh
> >
> > Documentation/trace/events-power.rst | 1 +
> > drivers/cpufreq/cpufreq.c | 1 +
> > include/trace/events/power.h | 25 +++++++++++++++++++++++++
> > 3 files changed, 27 insertions(+)
> >
> > diff --git a/Documentation/trace/events-power.rst b/Documentation/trace/events-power.rst
> > index a77daca75e30..2ef318962e29 100644
> > --- a/Documentation/trace/events-power.rst
> > +++ b/Documentation/trace/events-power.rst
> > @@ -27,6 +27,7 @@ cpufreq.
> >
> > cpu_idle "state=%lu cpu_id=%lu"
> > cpu_frequency "state=%lu cpu_id=%lu"
> > + cpu_frequency_limits "min=%lu max=%lu cpu_id=%lu"
> >
> > A suspend event is used to indicate the system going in and out of the
> > suspend mode:
> > diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> > index b0dfd3222013..4fd935df101e 100644
> > --- a/drivers/cpufreq/cpufreq.c
> > +++ b/drivers/cpufreq/cpufreq.c
> > @@ -2236,6 +2236,7 @@ static int cpufreq_set_policy(struct cpufreq_policy *policy,
> >
> > policy->min = new_policy->min;
> > policy->max = new_policy->max;
> > + trace_cpu_frequency_limits(policy->min, policy->max, policy->cpu);
>
> Wouldn't it make more sense just to pass in "policy"?

Yes I agree, good idea.

> >
> > policy->cached_target_freq = UINT_MAX;
> >
> > diff --git a/include/trace/events/power.h b/include/trace/events/power.h
> > index 908977d69783..f5bec45108b6 100644
> > --- a/include/trace/events/power.h
> > +++ b/include/trace/events/power.h
> > @@ -148,6 +148,31 @@ DEFINE_EVENT(cpu, cpu_frequency,
> > TP_ARGS(frequency, cpu_id)
> > );
> >
> > +TRACE_EVENT(cpu_frequency_limits,
> > +
> > + TP_PROTO(unsigned int min_freq, unsigned int max_freq,
> > + unsigned int cpu_id),
> > +
> > + TP_ARGS(min_freq, max_freq, cpu_id),
> > +
> > + TP_STRUCT__entry(
> > + __field(u32, min_freq)
> > + __field(u32, max_freq)
> > + __field(u32, cpu_id)
> > + ),
> > +
> > + TP_fast_assign(
> > + __entry->min_freq = min_freq;
> > + __entry->max_freq = max_freq;
> > + __entry->cpu_id = cpu_id;
>
> Then have here:
>
> __entry->min_freq = policy->min;
> __entry->max_freq = policy->max;
> __entry->puc_id = policy->cpu;
>
> It would also make the footprint of the tracepoint in the code smaller
> as it would pass fewer parameters to the trace event.

Yes, that's a great idea. I'll make the change and post v3.

-Joel