2022-04-22 07:40:42

by Lukasz Luba

[permalink] [raw]
Subject: [PATCH] arch_topology: Trace the update thermal pressure

Add trace event to capture the moment of the call for updating the thermal
pressure value. It's helpful to investigate how often those events occurs
in a system dealing with throttling. This trace event is needed since the
old 'cdev_update' might not be used by some drivers. Also, the new trace
event shows capacity value, not a cooling state.

We already have a trace point 'pelt_thermal_tp' which after a change to
trace event can be paired with this new 'thermal_pressure_update' and
derive more insight what is going on in the system under pressure
(and why).

Signed-off-by: Lukasz Luba <[email protected]>
---
drivers/base/arch_topology.c | 4 ++++
include/trace/events/thermal.h | 19 +++++++++++++++++++
2 files changed, 23 insertions(+)

diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
index 1d6636ebaac5..4f0392de3081 100644
--- a/drivers/base/arch_topology.c
+++ b/drivers/base/arch_topology.c
@@ -19,6 +19,8 @@
#include <linux/rcupdate.h>
#include <linux/sched.h>

+#include <trace/events/thermal.h>
+
static DEFINE_PER_CPU(struct scale_freq_data __rcu *, sft_data);
static struct cpumask scale_freq_counters_mask;
static bool scale_freq_invariant;
@@ -195,6 +197,8 @@ void topology_update_thermal_pressure(const struct cpumask *cpus,

th_pressure = max_capacity - capacity;

+ trace_thermal_pressure_update(cpu, th_pressure);
+
for_each_cpu(cpu, cpus)
WRITE_ONCE(per_cpu(thermal_pressure, cpu), th_pressure);
}
diff --git a/include/trace/events/thermal.h b/include/trace/events/thermal.h
index 8a5f04888abd..1bf08ee1a25b 100644
--- a/include/trace/events/thermal.h
+++ b/include/trace/events/thermal.h
@@ -65,6 +65,25 @@ TRACE_EVENT(cdev_update,
TP_printk("type=%s target=%lu", __get_str(type), __entry->target)
);

+TRACE_EVENT(thermal_pressure_update,
+
+ TP_PROTO(int cpu, unsigned long thermal_pressure),
+
+ TP_ARGS(cpu, thermal_pressure),
+
+ TP_STRUCT__entry(
+ __field(int, cpu)
+ __field(unsigned long, thermal_pressure)
+ ),
+
+ TP_fast_assign(
+ __entry->cpu = cpu;
+ __entry->thermal_pressure = thermal_pressure;
+ ),
+
+ TP_printk("cpu=%d thermal_pressure=%lu", __entry->cpu, __entry->thermal_pressure)
+);
+
TRACE_EVENT(thermal_zone_trip,

TP_PROTO(struct thermal_zone_device *tz, int trip,
--
2.17.1


2022-04-22 19:41:54

by Lukasz Luba

[permalink] [raw]
Subject: Re: [PATCH] arch_topology: Trace the update thermal pressure



On 4/21/22 00:44, Steven Rostedt wrote:
> On Tue, 19 Apr 2022 17:48:01 +0100
> Lukasz Luba <[email protected]> wrote:
>
>> diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
>> index 1d6636ebaac5..4f0392de3081 100644
>> --- a/drivers/base/arch_topology.c
>> +++ b/drivers/base/arch_topology.c
>> @@ -19,6 +19,8 @@
>> #include <linux/rcupdate.h>
>> #include <linux/sched.h>
>>
>> +#include <trace/events/thermal.h>
>> +
>> static DEFINE_PER_CPU(struct scale_freq_data __rcu *, sft_data);
>> static struct cpumask scale_freq_counters_mask;
>> static bool scale_freq_invariant;
>> @@ -195,6 +197,8 @@ void topology_update_thermal_pressure(const struct cpumask *cpus,
>>
>> th_pressure = max_capacity - capacity;
>>
>> + trace_thermal_pressure_update(cpu, th_pressure);
>> +
>> for_each_cpu(cpu, cpus)
>> WRITE_ONCE(per_cpu(thermal_pressure, cpu), th_pressure);
>> }
>> diff --git a/include/trace/events/thermal.h b/include/trace/events/thermal.h
>> index 8a5f04888abd..1bf08ee1a25b 100644
>> --- a/include/trace/events/thermal.h
>> +++ b/include/trace/events/thermal.h
>> @@ -65,6 +65,25 @@ TRACE_EVENT(cdev_update,
>> TP_printk("type=%s target=%lu", __get_str(type), __entry->target)
>> );
>>
>> +TRACE_EVENT(thermal_pressure_update,
>> +
>> + TP_PROTO(int cpu, unsigned long thermal_pressure),
>> +
>> + TP_ARGS(cpu, thermal_pressure),
>> +
>> + TP_STRUCT__entry(
>> + __field(int, cpu)
>> + __field(unsigned long, thermal_pressure)
>
> Note, it is always best to place the bigger object before the smaller one
> (when not properly aligned), as that will help to prevent structure
> "holes". That is:
>
> __field(unsigned long, thermal_pressure)
> __field(int, cpu)
>
>
> Otherwise, you are pretty much guaranteed to have a 4 byte hole between cpu
> and thermal_pressure on 64 bit machines.
>
> Also, for the warning you got from the test robot, if you are using this in
> a module and defining it in the core kernel, you need to add:
>
> EXPORT_TRACEPOINT_SYMBOL_GPL(thermal_pressure_update);
>
> Somewhere in the C file that includes this file and defines
> CREATE_TRACE_POINTS.
>
> -- Steve
>
>

Thank you Steve, I'll swap those fields and add the needed export.
I have to go through those older trace events in that file and
understand if they are correct, since I just followed the pattern
in 'cdev_update'.

Regards,
Lukasz

2022-04-22 21:40:36

by Steven Rostedt

[permalink] [raw]
Subject: Re: [PATCH] arch_topology: Trace the update thermal pressure

On Tue, 19 Apr 2022 17:48:01 +0100
Lukasz Luba <[email protected]> wrote:

> diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
> index 1d6636ebaac5..4f0392de3081 100644
> --- a/drivers/base/arch_topology.c
> +++ b/drivers/base/arch_topology.c
> @@ -19,6 +19,8 @@
> #include <linux/rcupdate.h>
> #include <linux/sched.h>
>
> +#include <trace/events/thermal.h>
> +
> static DEFINE_PER_CPU(struct scale_freq_data __rcu *, sft_data);
> static struct cpumask scale_freq_counters_mask;
> static bool scale_freq_invariant;
> @@ -195,6 +197,8 @@ void topology_update_thermal_pressure(const struct cpumask *cpus,
>
> th_pressure = max_capacity - capacity;
>
> + trace_thermal_pressure_update(cpu, th_pressure);
> +
> for_each_cpu(cpu, cpus)
> WRITE_ONCE(per_cpu(thermal_pressure, cpu), th_pressure);
> }
> diff --git a/include/trace/events/thermal.h b/include/trace/events/thermal.h
> index 8a5f04888abd..1bf08ee1a25b 100644
> --- a/include/trace/events/thermal.h
> +++ b/include/trace/events/thermal.h
> @@ -65,6 +65,25 @@ TRACE_EVENT(cdev_update,
> TP_printk("type=%s target=%lu", __get_str(type), __entry->target)
> );
>
> +TRACE_EVENT(thermal_pressure_update,
> +
> + TP_PROTO(int cpu, unsigned long thermal_pressure),
> +
> + TP_ARGS(cpu, thermal_pressure),
> +
> + TP_STRUCT__entry(
> + __field(int, cpu)
> + __field(unsigned long, thermal_pressure)

Note, it is always best to place the bigger object before the smaller one
(when not properly aligned), as that will help to prevent structure
"holes". That is:

__field(unsigned long, thermal_pressure)
__field(int, cpu)


Otherwise, you are pretty much guaranteed to have a 4 byte hole between cpu
and thermal_pressure on 64 bit machines.

Also, for the warning you got from the test robot, if you are using this in
a module and defining it in the core kernel, you need to add:

EXPORT_TRACEPOINT_SYMBOL_GPL(thermal_pressure_update);

Somewhere in the C file that includes this file and defines
CREATE_TRACE_POINTS.

-- Steve



> + ),
> +
> + TP_fast_assign(
> + __entry->cpu = cpu;
> + __entry->thermal_pressure = thermal_pressure;
> + ),
> +
> + TP_printk("cpu=%d thermal_pressure=%lu", __entry->cpu, __entry->thermal_pressure)
> +);
> +
> TRACE_EVENT(thermal_zone_trip,
>
> TP_PROTO(struct thermal_zone_device *tz, int trip,