[PATCH] tracing, perf: add more power related events

[PATCH] tracing, perf: add more power related events

[Jean Pihet]

Hi,

Here is a patch proposal for adding new trace events for power management.
Note: thread restarted after the initial discussions on LKML.

Jean

>From 6768b88e8133129fa847dd7a95dc6dd17c0662d2 Mon Sep 17 00:00:00 2001
From: Jean Pihet <jean.pihet@newoldbits.com>
Date: Tue, 7 Sep 2010 09:12:48 +0200
Subject: [PATCH] [PATCH] tracing, perf: add more power related events

This patch adds new generic events for dynamic power management
tracing:
- clock events class: used for clock enable/disable and for
  clock rate change,
- power_domain events class: used for power domains transitions.

The OMAP architecture is using the new events for PM debugging, however
the new events are made generic enough to be used by all platforms.

Signed-off-by: Jean Pihet <j-pihet@ti.com>
---
 arch/arm/mach-omap2/cpuidle34xx.c |    2 +
 arch/arm/mach-omap2/pm34xx.c      |    4 ++
 arch/arm/mach-omap2/powerdomain.c |    3 +
 arch/arm/plat-omap/clock.c        |   13 ++++-
 arch/arm/plat-omap/cpu-omap.c     |    5 ++-
 include/trace/events/power.h      |   90 +++++++++++++++++++++++++++++++++++-
 6 files changed, 110 insertions(+), 7 deletions(-)

diff --git a/arch/arm/mach-omap2/cpuidle34xx.c
b/arch/arm/mach-omap2/cpuidle34xx.c
index 3d3d035..6113bd9 100644
--- a/arch/arm/mach-omap2/cpuidle34xx.c
+++ b/arch/arm/mach-omap2/cpuidle34xx.c
@@ -24,6 +24,7 @@

 #include <linux/sched.h>
 #include <linux/cpuidle.h>
+#include <trace/events/power.h>

 #include <plat/prcm.h>
 #include <plat/irqs.h>
@@ -130,6 +131,7 @@ static int omap3_enter_idle(struct cpuidle_device *dev,
 	local_irq_disable();
 	local_fiq_disable();

+	trace_power_start(POWER_CSTATE, cx->type, smp_processor_id());
 	pwrdm_set_next_pwrst(mpu_pd, mpu_state);
 	pwrdm_set_next_pwrst(core_pd, core_state);

diff --git a/arch/arm/mach-omap2/pm34xx.c b/arch/arm/mach-omap2/pm34xx.c
index f25bc3d..7bf8a87 100644
--- a/arch/arm/mach-omap2/pm34xx.c
+++ b/arch/arm/mach-omap2/pm34xx.c
@@ -28,6 +28,7 @@
 #include <linux/clk.h>
 #include <linux/delay.h>
 #include <linux/slab.h>
+#include <trace/events/power.h>

 #include <plat/sram.h>
 #include <plat/clockdomain.h>
@@ -588,6 +589,7 @@ static void omap3_pm_idle(void)
 	if (omap_irq_pending() || need_resched())
 		goto out;

+	trace_power_start(POWER_CSTATE, 1, smp_processor_id());
 	omap_sram_idle();

 out:
@@ -628,6 +630,8 @@ static int omap3_pm_suspend(void)
 		omap2_pm_wakeup_on_timer(wakeup_timer_seconds,
 					 wakeup_timer_milliseconds);

+	trace_power_start(POWER_SSTATE, 1, smp_processor_id());
+
 	/* Read current next_pwrsts */
 	list_for_each_entry(pwrst, &pwrst_list, node)
 		pwrst->saved_state = pwrdm_read_next_pwrst(pwrst->pwrdm);
diff --git a/arch/arm/mach-omap2/powerdomain.c
b/arch/arm/mach-omap2/powerdomain.c
index 6527ec3..73cbe9a 100644
--- a/arch/arm/mach-omap2/powerdomain.c
+++ b/arch/arm/mach-omap2/powerdomain.c
@@ -23,6 +23,7 @@
 #include <linux/errno.h>
 #include <linux/err.h>
 #include <linux/io.h>
+#include <trace/events/power.h>

 #include <asm/atomic.h>

@@ -440,6 +441,8 @@ int pwrdm_set_next_pwrst(struct powerdomain
*pwrdm, u8 pwrst)
 	pr_debug("powerdomain: setting next powerstate for %s to %0x\n",
 		 pwrdm->name, pwrst);

+	trace_power_domain_target(pwrdm->name, pwrst, smp_processor_id());
+
 	prm_rmw_mod_reg_bits(OMAP_POWERSTATE_MASK,
 			     (pwrst << OMAP_POWERSTATE_SHIFT),
 			     pwrdm->prcm_offs, pwrstctrl_reg_offs);
diff --git a/arch/arm/plat-omap/clock.c b/arch/arm/plat-omap/clock.c
index 7190cbd..d6518f5 100644
--- a/arch/arm/plat-omap/clock.c
+++ b/arch/arm/plat-omap/clock.c
@@ -21,6 +21,7 @@
 #include <linux/cpufreq.h>
 #include <linux/debugfs.h>
 #include <linux/io.h>
+#include <trace/events/power.h>

 #include <plat/clock.h>

@@ -43,8 +44,10 @@ int clk_enable(struct clk *clk)
 		return -EINVAL;

 	spin_lock_irqsave(&clockfw_lock, flags);
-	if (arch_clock->clk_enable)
+	if (arch_clock->clk_enable) {
+		trace_clock_enable(clk->name, 1, smp_processor_id());
 		ret = arch_clock->clk_enable(clk);
+	}
 	spin_unlock_irqrestore(&clockfw_lock, flags);

 	return ret;
@@ -66,8 +69,10 @@ void clk_disable(struct clk *clk)
 		goto out;
 	}

-	if (arch_clock->clk_disable)
+	if (arch_clock->clk_disable) {
+		trace_clock_disable(clk->name, 0, smp_processor_id());
 		arch_clock->clk_disable(clk);
+	}

 out:
 	spin_unlock_irqrestore(&clockfw_lock, flags);
@@ -120,8 +125,10 @@ int clk_set_rate(struct clk *clk, unsigned long rate)
 		return ret;

 	spin_lock_irqsave(&clockfw_lock, flags);
-	if (arch_clock->clk_set_rate)
+	if (arch_clock->clk_set_rate) {
+		trace_clock_set_rate(clk->name, rate, smp_processor_id());
 		ret = arch_clock->clk_set_rate(clk, rate);
+	}
 	if (ret == 0) {
 		if (clk->recalc)
 			clk->rate = clk->recalc(clk);
diff --git a/arch/arm/plat-omap/cpu-omap.c b/arch/arm/plat-omap/cpu-omap.c
index df08829..cc4e41f 100644
--- a/arch/arm/plat-omap/cpu-omap.c
+++ b/arch/arm/plat-omap/cpu-omap.c
@@ -25,6 +25,7 @@
 #include <linux/err.h>
 #include <linux/clk.h>
 #include <linux/io.h>
+#include <trace/events/power.h>

 #include <mach/hardware.h>
 #include <plat/clock.h>
@@ -116,8 +117,10 @@ static int omap_target(struct cpufreq_policy *policy,
 	cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
 #elif defined(CONFIG_ARCH_OMAP3) && !defined(CONFIG_OMAP_PM_NONE)
 	freq = target_freq * 1000;
-	if (opp_find_freq_ceil(mpu_dev, &freq))
+	if (opp_find_freq_ceil(mpu_dev, &freq)) {
+		trace_power_frequency(POWER_PSTATE, freq, smp_processor_id());
 		omap_pm_cpu_set_freq(freq);
+	}
 #endif
 	return ret;
 }
diff --git a/include/trace/events/power.h b/include/trace/events/power.h
index 35a2a6e..286784d 100644
--- a/include/trace/events/power.h
+++ b/include/trace/events/power.h
@@ -10,12 +10,17 @@
 #ifndef _TRACE_POWER_ENUM_
 #define _TRACE_POWER_ENUM_
 enum {
-	POWER_NONE = 0,
-	POWER_CSTATE = 1,
-	POWER_PSTATE = 2,
+	POWER_NONE	= 0,
+	POWER_CSTATE	= 1,	/* C-State */
+	POWER_PSTATE	= 2,	/* Fequency change or DVFS */
+	POWER_SSTATE	= 3,	/* Suspend */
 };
 #endif

+/*
+ * The power events are used for cpuidle & suspend (power_start, power_end)
+ *  and for cpufreq (power_frequency)
+ */
 DECLARE_EVENT_CLASS(power,

 	TP_PROTO(unsigned int type, unsigned int state, unsigned int cpu_id),
@@ -70,6 +75,85 @@ TRACE_EVENT(power_end,

 );

+/*
+ * The clock events are used for clock enable/disable and for
+ *  clock rate change
+ */
+DECLARE_EVENT_CLASS(clock,
+
+	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+
+	TP_ARGS(name, state, cpu_id),
+
+	TP_STRUCT__entry(
+		__string(       name,           name            )
+		__field(        u64,            state           )
+		__field(        u64,            cpu_id          )
+	),
+
+	TP_fast_assign(
+		__assign_str(name, name);
+		__entry->state = state;
+		__entry->cpu_id = cpu_id;
+	),
+
+	TP_printk("%s state=%lu cpu_id=%lu", __get_str(name),
+		(unsigned long)__entry->state, (unsigned long)__entry->cpu_id)
+);
+
+DEFINE_EVENT(clock, clock_enable,
+
+	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+
+	TP_ARGS(name, state, cpu_id)
+);
+
+DEFINE_EVENT(clock, clock_disable,
+
+	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+
+	TP_ARGS(name, state, cpu_id)
+);
+
+DEFINE_EVENT(clock, clock_set_rate,
+
+	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+
+	TP_ARGS(name, state, cpu_id)
+);
+
+/*
+ * The power domain events are used for power domains transitions
+ */
+DECLARE_EVENT_CLASS(power_domain,
+
+	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+
+	TP_ARGS(name, state, cpu_id),
+
+	TP_STRUCT__entry(
+		__string(       name,           name            )
+		__field(        u64,            state           )
+		__field(        u64,            cpu_id          )
+	),
+
+	TP_fast_assign(
+		__assign_str(name, name);
+		__entry->state = state;
+		__entry->cpu_id = cpu_id;
+),
+
+	TP_printk("%s state=%lu cpu_id=%lu", __get_str(name),
+		(unsigned long)__entry->state, (unsigned long)__entry->cpu_id)
+);
+
+DEFINE_EVENT(power_domain, power_domain_target,
+
+	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+
+	TP_ARGS(name, state, cpu_id)
+);
+
 #endif /* _TRACE_POWER_H */

 /* This part must be outside protection */
-- 
1.7.1

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

Here is some more information about the patch:

Initial discussion on LKML: cf.
http://marc.info/?l=linux-kernel&m=128195697205096&w=4,
PM debug and profiling wiki page with rationale, todo, patches, tools
screenshots ...:
http://omappedia.org/wiki/Power_Management_Debug_and_Profiling

On Tue, Sep 7, 2010 at 9:21 AM, Jean Pihet <jean.pihet@newoldbits.com> wrote:
> Hi,
>
> Here is a patch proposal for adding new trace events for power management.
> Note: thread restarted after the initial discussions on LKML.
Sorry the thread did not get restarted because I am using the same
e-mail subject.

Jean

Re: [PATCH] tracing, perf: add more power related events

[Ingo Molnar]

* Jean Pihet <jean.pihet@newoldbits.com> wrote:

> Hi,
> 
> Here is a patch proposal for adding new trace events for power management.
> Note: thread restarted after the initial discussions on LKML.

Also mind including the ACPI tracepoints we talked about? That way a lot 
more people could test it, etc.

	Ingo

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

On Wed, Sep 8, 2010 at 8:53 AM, Ingo Molnar <mingo@elte.hu> wrote:
>
> * Jean Pihet <jean.pihet@newoldbits.com> wrote:
>
>> Hi,
>>
>> Here is a patch proposal for adding new trace events for power management.
>> Note: thread restarted after the initial discussions on LKML.
>
> Also mind including the ACPI tracepoints we talked about? That way a lot
> more people could test it, etc.
I think the ACPI tracepoints can be added on top of the proposed
patch. Is that ok?

Jean

Re: [PATCH] tracing, perf: add more power related events

[Ingo Molnar]

* Jean Pihet <jean.pihet@newoldbits.com> wrote:

> On Wed, Sep 8, 2010 at 8:53 AM, Ingo Molnar <mingo@elte.hu> wrote:
> >
> > * Jean Pihet <jean.pihet@newoldbits.com> wrote:
> >
> >> Hi,
> >>
> >> Here is a patch proposal for adding new trace events for power management.
> >> Note: thread restarted after the initial discussions on LKML.
> >
> > Also mind including the ACPI tracepoints we talked about? That way a lot
> > more people could test it, etc.
>
> I think the ACPI tracepoints can be added on top of the proposed 
> patch. Is that ok?

Yeah - and the OMAP thing can be split up too if the OMAP folks prefer 
it that way, but we still want to _see_ all the patches in this thread 
together so that we have a critical mass of people interested in all 
this ...

Thanks,

	Ingo

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

On Thu, Sep 9, 2010 at 9:54 AM, Ingo Molnar <mingo@elte.hu> wrote:
>> I think the ACPI tracepoints can be added on top of the proposed
>> patch. Is that ok?
>
> Yeah - and the OMAP thing can be split up too if the OMAP folks prefer
> it that way, but we still want to _see_ all the patches in this thread
> together so that we have a critical mass of people interested in all
> this ...

Any chance to get the patch merged in?
What is needed wrt to ACPI support?

>
> Thanks,
>
>        Ingo

Thanks,
Jean

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

Hi,

I had a quick look at this and it's amazing how broken
the whole power event tracing interfaces are.
It's not your fault, Jean, they always were and adding your stuff is 
fine.

Some questions, maybe I've overseen something:

Why does this event:
DEFINE_EVENT(power, power_frequency,
exist and takes a C-/P-state, now also an S-state type as argument?

It should be named more generic, like:
DEFINE_EVENT(power, power_switch,
then it could get invoked when any P-/C-/S-/X- state
switch happened.

What kind of hack is this:
TRACE_EVENT(power_end,
showing up as:
power_end: dummy=65535
What ends here?
I know it's a workaround/hack for userspace apps to be
able to detect leaving of a sleep state, but how would someone
know or how would someone describe this in a proper API documentation?
Apropos documentation..., are the power trace events documented 
somewhere?

At least the state should still be passed, then the _start/_end thing 
can be reused by something else than C-states.

I can't see the use of having _start/_end events at all.
You are always in a state, having one:
power_switch
as suggested above, is enough to track everything.

Examples:

Today's C-state tracking:
power_start: type=1 state=1  -> C1 entered
power_end: dummy=65535       -> C-state left
power_start: type=1 state=2  -> C2 entered
power_end: dummy=65535       -> C-state left
would then be:
power_switch: type=1 state=1  -> C1 entered
power_switch: type=1 state=0  -> C0 entered -> means C1 left...
power_switch: type=1 state=2  -> C2 entered
power_switch: type=1 state=0  -> C0 entered -> means C2 left...
...

Todays P-state tracking:
power_frequency: type=2 state=125000000     -> P-state change to 125 MHz
power_frequency: type=2 state=90000000      -> P-state change to 90 MHz
would then be:
power_switch: type=2 state=125000000       -> P-state change to 125 MHz
power_switch: type=2 state=90000000        -> P-state change to 90 MHz

The S-state and T-state tracking would fit into that without 
modification.
Thinking one step further, a possibility to track D-states would
need an additional field, a pointer to the device, best a sysfs path
or similar.

Jean, I do not think tracing the S-state with power_start is a
good idea. Best would be the power_frequency gets renamed (yes, breaks
userspace, but those have to be adjusted) and used for P- and S-
state tracking (and C-state tracking as well, but this would need
additional userspace modifications).
How do you track when the S-states end?

        Thomas

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

Hi Thomas,

On Fri, Sep 17, 2010 at 3:08 PM, Thomas Renninger <trenn@suse.de> wrote:
> Hi,
>
> I had a quick look at this and it's amazing how broken
> the whole power event tracing interfaces are.
> It's not your fault, Jean, they always were and adding your stuff is
> fine.
That is the whole point! This code needs a serious clean-up and reorganization.
The patch I submitted is now merged in the tip tree, but the can be
corrected once the new power trace API is agreed on.

> Some questions, maybe I've overseen something:
>
> Why does this event:
> DEFINE_EVENT(power, power_frequency,
> exist and takes a C-/P-state, now also an S-state type as argument?
There is no clear link between the POWER_ types enum and the API, that
needs to change.

> It should be named more generic, like:
> DEFINE_EVENT(power, power_switch,
> then it could get invoked when any P-/C-/S-/X- state
> switch happened.
Agree. We need some better definitions for the events and at the same
time a more flexible way to pass extra arguments (e.g. like a return
code or the real HW state in case the desired state is not reached).
Also it is required to be able to track sub-states.
The idea is to identify the C-states latency in the low power entry
and exit paths, using 2 new macros in the enum and a sub state
argument. Cf. http://omappedia.org/images/b/b9/Pytimechart_screenshot9_rs.jpg
(from http://omappedia.org/wiki/Power_Management_Debug_and_Profiling).

> What kind of hack is this:
> TRACE_EVENT(power_end,
> showing up as:
> power_end: dummy=65535
> What ends here?
> I know it's a workaround/hack for userspace apps to be
> able to detect leaving of a sleep state, but how would someone
> know or how would someone describe this in a proper API documentation?
That was a hack, it is now gone. The new power_end only takes the
cpu_id argument.

> Apropos documentation..., are the power trace events documented
> somewhere?
No. We need something like Documentation/trace/events-kmem.txt. I can
write that with for the new power API.

> At least the state should still be passed, then the _start/_end thing
> can be reused by something else than C-states.
>
> I can't see the use of having _start/_end events at all.
> You are always in a state, having one:
> power_switch
> as suggested above, is enough to track everything.
Most of the events can be converted to power_switch, but not all.
Example: you need to know when the suspend ends.

>
> Examples:
>
> Today's C-state tracking:
> power_start: type=1 state=1  -> C1 entered
> power_end: dummy=65535       -> C-state left
> power_start: type=1 state=2  -> C2 entered
> power_end: dummy=65535       -> C-state left
> would then be:
> power_switch: type=1 state=1  -> C1 entered
> power_switch: type=1 state=0  -> C0 entered -> means C1 left...
> power_switch: type=1 state=2  -> C2 entered
> power_switch: type=1 state=0  -> C0 entered -> means C2 left...
> ...
>
> Todays P-state tracking:
> power_frequency: type=2 state=125000000     -> P-state change to 125 MHz
> power_frequency: type=2 state=90000000      -> P-state change to 90 MHz
> would then be:
> power_switch: type=2 state=125000000       -> P-state change to 125 MHz
> power_switch: type=2 state=90000000        -> P-state change to 90 MHz
>
> The S-state and T-state tracking would fit into that without
> modification.
> Thinking one step further, a possibility to track D-states would
What are the T- and D- states?

> need an additional field, a pointer to the device, best a sysfs path
> or similar.
Agree on that. As said above I would like to have extra args.

> Jean, I do not think tracing the S-state with power_start is a
> good idea. Best would be the power_frequency gets renamed

>(yes, breaks
> userspace, but those have to be adjusted)
BTW I can patch pytimechart. What other tools have to change? powertop ...?

> and used for P- and S-
> state tracking (and C-state tracking as well, but this would need
> additional userspace modifications).
OK

> How do you track when the S-states end?
Two options: 1) have new arguments that indicates enter/exit and a
sub-state; 2) a new event fot the exit.
I am in favor of 1 which is more generic.

Ok thanks for those remarks and suggetions.
Let me come back soon with a new power API proposal.

>
>        Thomas
>

Jean

Re: [PATCH] tracing, perf: add more power related events

[Ingo Molnar]

* Jean Pihet <jean.pihet@newoldbits.com> wrote:

> > Apropos documentation..., are the power trace events documented
> > somewhere?
>
> No. We need something like Documentation/trace/events-kmem.txt. I can 
> write that with for the new power API.

Such a patch introducing events-power.txt would be most welcome!

	Ingo

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

On Friday 17 September 2010 16:24:59 Ingo Molnar wrote:
> 
> * Jean Pihet <jean.pihet@newoldbits.com> wrote:
> 
> > > Apropos documentation..., are the power trace events documented
> > > somewhere?
> >
> > No. We need something like Documentation/trace/events-kmem.txt. I
> > can write that with for the new power API.
> Such a patch introducing events-power.txt would be most welcome!
Better wait a bit...
As soon as it exists, things are harder to change.
As long as there isn't an API documented, there doesn't exist one
(yes we want to have perf timechart, etc. still functional/compatible,
but still, better let's do not the same mistake with some quick shots).

Whatabout adding perf/trace interfaces to:
Documentation/ABI/*/perf-xy

    Thomas

Re: [PATCH] tracing, perf: add more power related events

[Ingo Molnar]

* Thomas Renninger <trenn@suse.de> wrote:

> On Friday 17 September 2010 16:24:59 Ingo Molnar wrote:
> > 
> > * Jean Pihet <jean.pihet@newoldbits.com> wrote:
> > 
> > > > Apropos documentation..., are the power trace events documented
> > > > somewhere?
> > >
> > > No. We need something like Documentation/trace/events-kmem.txt. I
> > > can write that with for the new power API.
> > Such a patch introducing events-power.txt would be most welcome!
> Better wait a bit...
> As soon as it exists, things are harder to change.
> As long as there isn't an API documented, there doesn't exist one
> (yes we want to have perf timechart, etc. still functional/compatible,
> but still, better let's do not the same mistake with some quick shots).

Music to my ears ;-) I was waiting for ACPI side feedback to all this, 
so whatever we do it would be nice to cover ARM and x86 as well - and in 
a single coherent framework.

[ You dont even have to document it, as good code is self-explanatory ;-) ]

Thanks,

	Ingo

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

On Friday 17 September 2010 18:24:12 Ingo Molnar wrote:
> 
> * Thomas Renninger <trenn@suse.de> wrote:
> 
> > On Friday 17 September 2010 16:24:59 Ingo Molnar wrote:

> [ You dont even have to document it, as good code is self-explanatory ;-) ]
I recently posted a patch exporting some things through /sys/kernel/debug/...
Greg complained that a file for Documentation/ABI/{testing,stable}/* is missing
and I fully agree.
If different userspace apps should make use of this (in above case nobody
than my debug userspace tool will...) and this should be called something like an API,
it should be documented and if something changes, it should
first be marked deprecated, etc.

     Thomas

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

Hi,

Here is a patch that redefines the power events API. The advantages
are: easier maintainance of the kernel and the
user space tools, a cleaner and more generic interface, more
parameters for fine tracing and even documentation!

Thomas, this patch has your patch above merged in ('power-trace: Use
power_switch_state instead of power_start and power_end'). The revised
ACPI patch is coming asap.

The trace points for x86 and OMAP are also udated accordingly.

The pytimechart tool needs an update for the new API. This can be done
as soon as the kernel code gets merged in.
Please note the point below about the existing code in builtin-timechart.c.

On Sat, Sep 18, 2010 at 12:26 AM, Thomas Renninger <trenn@suse.de> wrote:
> On Friday 17 September 2010 18:24:12 Ingo Molnar wrote:
>>
>> * Thomas Renninger <trenn@suse.de> wrote:
>>
>> > On Friday 17 September 2010 16:24:59 Ingo Molnar wrote:
>
>> [ You dont even have to document it, as good code is self-explanatory ;-) ]
> I recently posted a patch exporting some things through /sys/kernel/debug/...
> Greg complained that a file for Documentation/ABI/{testing,stable}/* is missing
> and I fully agree.
The proposed patch has the documentation in
Documentation/trace/events-power.txt.

> If different userspace apps should make use of this (in above case nobody
> than my debug userspace tool will...) and this should be called something like an API,
> it should be documented and if something changes, it should
> first be marked deprecated, etc.
Note: the exsiting code in tools/perf/builtin-timechart.c needs an
update for the new events API. Is this code still maintained? I not,
could pytimechart be merged in instead?

Feedback is welcome!

>
>     Thomas
>

Thanks,
Jean

---
From f37d1875050d33273b10e99497b4059b37e6680d Mon Sep 17 00:00:00 2001
From: Jean Pihet <jean.pihet@newoldbits.com>
Date: Wed, 22 Sep 2010 17:10:47 +0200
Subject: [PATCH] tools, perf: redefine the power events API

Redefine the API with:
- power_switch_state for C-, P- and S-states,
- clock and power_domain events

The new API allows easier maintainance of the kernel and the
user space tools, a cleaner and more generic interface,
more parameters for fine tracing and even documentation.

The new events are used by the x86 and OMAP platforms.

Signed-off-by: Jean Pihet <j-pihet@ti.com>
---
 Documentation/trace/events-power.txt |   70 +++++++++++++++++++++++++
 arch/arm/mach-omap2/cpuidle34xx.c    |    3 +
 arch/arm/mach-omap2/pm34xx.c         |   11 ++++
 arch/arm/mach-omap2/powerdomain.c    |    3 +
 arch/arm/plat-omap/clock.c           |   13 ++++-
 arch/arm/plat-omap/cpu-omap.c        |    3 +
 arch/x86/kernel/process.c            |   13 +++--
 arch/x86/kernel/process_32.c         |    3 +-
 arch/x86/kernel/process_64.c         |    3 +-
 drivers/cpufreq/cpufreq.c            |    3 +-
 drivers/cpuidle/cpuidle.c            |    2 +-
 drivers/idle/intel_idle.c            |    2 +-
 include/trace/events/power.h         |   95 ++++++++++++++++------------------
 kernel/trace/power-traces.c          |    2 -
 14 files changed, 161 insertions(+), 65 deletions(-)
 create mode 100644 Documentation/trace/events-power.txt

diff --git a/Documentation/trace/events-power.txt
b/Documentation/trace/events-power.txt
new file mode 100644
index 0000000..967f842
--- /dev/null
+++ b/Documentation/trace/events-power.txt
@@ -0,0 +1,70 @@
+
+			Subsystem Trace Points: power
+
+The power tracing system captures events related to power transitions
+within the kernel. Broadly speaking there are three major subheadings:
+
+  o Power state switch which reports events related to suspend (S-states),
+     cpuidle (C-states) and cpufreq (P-states)
+  o System clock related changes
+  o Power domains related changes and transitions
+
+This document describes what each of the tracepoints is and why they
+might be useful.
+
+Cf. include/trace/events/power.h for the events definitions.
+
+1. Power state switch events
+============================
+
+power_switch_state	type=%lu state=%lu sub_state=%lu cpu_id=%lu
+
+The 'type' parameter takes one of those macros:
+ . POWER_NONE	= 0,
+ . POWER_CSTATE	= 1,	/* C-State */
+ . POWER_PSTATE	= 2,	/* Fequency change or DVFS */
+ . POWER_SSTATE	= 3,	/* Suspend */
+
+The 'state' parameter is set depending on the type:
+ . Target C-state for type=POWER_CSTATE,
+ . Target frequency for type=POWER_PSTATE,
+ . Target suspend state for type=POWER_SSTATE
+
+Note: the value of '0' for state means an exit from the current state,
+i.e. power_switch_state(POWER_SSTATE, 1, 0, cpu) means 'the system enters
+the suspend state' while power_switch_state(POWER_SSTATE, 0, 0, cpu) means
+'the system exits the suspend state).
+
+The event which has 'state=0' is very important to the user space tools
+which are using it to detect the end of the current state, and so to correctly
+draw the states diagrams and to calculate accurate statistics etc.
+
+The 'sub_state' parameter is optional and is used as a sub state in the low
+power mode enter or exit, a return code or the real state the system
went into etc.
+
+2. Clocks events
+================
+The clock events are used for clock enable/disable and for
+clock rate change.
+
+clock_enable	%s state=%lu sub_state=%lu cpu_id=%lu
+clock_disable	%s state=%lu sub_state=%lu cpu_id=%lu
+clock_set_rate	%s state=%lu sub_state=%lu cpu_id=%lu
+
+The first parameter gives the clock name (e.g. "gpio1_iclk").
+The second parameter is '1' for enable, '0' for disable, the target
+clock rate for set_rate.
+The 'sub_state' parameter is optional and is used as a sub state in the low
+power mode enter or exit, a return code, the real state the system
went into etc.
+
+3. Power domains events
+=======================
+The power domain events are used for power domains transitions
+
+power_domain_target	"%s state=%lu sub_state=%lu cpu_id=%lu"
+
+The first parameter gives the power domain name (e.g. "mpu_pwrdm").
+The second parameter is the power domain target state.
+The 'sub_state' parameter is optional and is used as a sub state in the low
+power mode enter or exit, a return code, the real state the system
went into etc.
+
diff --git a/arch/arm/mach-omap2/cpuidle34xx.c
b/arch/arm/mach-omap2/cpuidle34xx.c
index 3d3d035..8fa317c 100644
--- a/arch/arm/mach-omap2/cpuidle34xx.c
+++ b/arch/arm/mach-omap2/cpuidle34xx.c
@@ -24,6 +24,7 @@

 #include <linux/sched.h>
 #include <linux/cpuidle.h>
+#include <trace/events/power.h>

 #include <plat/prcm.h>
 #include <plat/irqs.h>
@@ -122,6 +123,8 @@ static int omap3_enter_idle(struct cpuidle_device *dev,
 	struct timespec ts_preidle, ts_postidle, ts_idle;
 	u32 mpu_state = cx->mpu_state, core_state = cx->core_state;

+	trace_power_switch_state(POWER_CSTATE, cx->type, 0, smp_processor_id());
+
 	current_cx_state = *cx;

 	/* Used to keep track of the total time in idle */
diff --git a/arch/arm/mach-omap2/pm34xx.c b/arch/arm/mach-omap2/pm34xx.c
index 7b03426..8b774b1 100644
--- a/arch/arm/mach-omap2/pm34xx.c
+++ b/arch/arm/mach-omap2/pm34xx.c
@@ -28,6 +28,7 @@
 #include <linux/clk.h>
 #include <linux/delay.h>
 #include <linux/slab.h>
+#include <trace/events/power.h>

 #include <plat/sram.h>
 #include <plat/clockdomain.h>
@@ -557,6 +558,8 @@ static void omap3_pm_idle(void)
 	if (omap_irq_pending() || need_resched())
 		goto out;

+	trace_power_switch_state(POWER_CSTATE, 1, 0, smp_processor_id());
+
 	omap_sram_idle();

 out:
@@ -595,6 +598,8 @@ static int omap3_pm_suspend(void)
 	struct power_state *pwrst;
 	int state, ret = 0;

+	trace_power_switch_state(POWER_SSTATE, 1, 0, smp_processor_id());
+
 	if (wakeup_timer_seconds || wakeup_timer_milliseconds)
 		omap2_pm_wakeup_on_timer(wakeup_timer_seconds,
 					 wakeup_timer_milliseconds);
@@ -623,6 +628,10 @@ restore:
 			printk(KERN_INFO "Powerdomain (%s) didn't enter "
 			       "target state %d\n",
 			       pwrst->pwrdm->name, pwrst->next_state);
+			trace_power_domain_target(pwrst->pwrdm->name,
+						  pwrst->next_state,
+						  state,
+						  smp_processor_id());
 			ret = -1;
 		}
 		set_pwrdm_state(pwrst->pwrdm, pwrst->saved_state);
@@ -633,6 +642,8 @@ restore:
 		printk(KERN_INFO "Successfully put all powerdomains "
 		       "to target state\n");

+	trace_power_switch_state(POWER_SSTATE, 0, 0, smp_processor_id());
+
 	return ret;
 }

diff --git a/arch/arm/mach-omap2/powerdomain.c
b/arch/arm/mach-omap2/powerdomain.c
index 6527ec3..03d286e 100644
--- a/arch/arm/mach-omap2/powerdomain.c
+++ b/arch/arm/mach-omap2/powerdomain.c
@@ -23,6 +23,7 @@
 #include <linux/errno.h>
 #include <linux/err.h>
 #include <linux/io.h>
+#include <trace/events/power.h>

 #include <asm/atomic.h>

@@ -440,6 +441,8 @@ int pwrdm_set_next_pwrst(struct powerdomain
*pwrdm, u8 pwrst)
 	pr_debug("powerdomain: setting next powerstate for %s to %0x\n",
 		 pwrdm->name, pwrst);

+	trace_power_domain_target(pwrdm->name, pwrst, 0, smp_processor_id());
+
 	prm_rmw_mod_reg_bits(OMAP_POWERSTATE_MASK,
 			     (pwrst << OMAP_POWERSTATE_SHIFT),
 			     pwrdm->prcm_offs, pwrstctrl_reg_offs);
diff --git a/arch/arm/plat-omap/clock.c b/arch/arm/plat-omap/clock.c
index 7190cbd..d42739d 100644
--- a/arch/arm/plat-omap/clock.c
+++ b/arch/arm/plat-omap/clock.c
@@ -21,6 +21,7 @@
 #include <linux/cpufreq.h>
 #include <linux/debugfs.h>
 #include <linux/io.h>
+#include <trace/events/power.h>

 #include <plat/clock.h>

@@ -43,8 +44,10 @@ int clk_enable(struct clk *clk)
 		return -EINVAL;

 	spin_lock_irqsave(&clockfw_lock, flags);
-	if (arch_clock->clk_enable)
+	if (arch_clock->clk_enable) {
+		trace_clock_enable(clk->name, 1, 0, smp_processor_id());
 		ret = arch_clock->clk_enable(clk);
+	}
 	spin_unlock_irqrestore(&clockfw_lock, flags);

 	return ret;
@@ -66,8 +69,10 @@ void clk_disable(struct clk *clk)
 		goto out;
 	}

-	if (arch_clock->clk_disable)
+	if (arch_clock->clk_disable) {
+		trace_clock_disable(clk->name, 0, 0, smp_processor_id());
 		arch_clock->clk_disable(clk);
+	}

 out:
 	spin_unlock_irqrestore(&clockfw_lock, flags);
@@ -120,8 +125,10 @@ int clk_set_rate(struct clk *clk, unsigned long rate)
 		return ret;

 	spin_lock_irqsave(&clockfw_lock, flags);
-	if (arch_clock->clk_set_rate)
+	if (arch_clock->clk_set_rate) {
+		trace_clock_set_rate(clk->name, rate, 0, smp_processor_id());
 		ret = arch_clock->clk_set_rate(clk, rate);
+	}
 	if (ret == 0) {
 		if (clk->recalc)
 			clk->rate = clk->recalc(clk);
diff --git a/arch/arm/plat-omap/cpu-omap.c b/arch/arm/plat-omap/cpu-omap.c
index 6d3d333..5b6aa61 100644
--- a/arch/arm/plat-omap/cpu-omap.c
+++ b/arch/arm/plat-omap/cpu-omap.c
@@ -21,6 +21,7 @@
 #include <linux/err.h>
 #include <linux/clk.h>
 #include <linux/io.h>
+#include <trace/events/power.h>

 #include <mach/hardware.h>
 #include <plat/clock.h>
@@ -95,6 +96,8 @@ static int omap_target(struct cpufreq_policy *policy,
 	printk(KERN_DEBUG "cpufreq-omap: transition: %u --> %u\n",
 	       freqs.old, freqs.new);
 #endif
+	trace_power_switch_state(POWER_PSTATE, freqs.new, 0,
+				 smp_processor_id());
 	ret = clk_set_rate(mpu_clk, freqs.new * 1000);
 	cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);

diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index e8a5ad1..496d4af 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -372,7 +372,8 @@ static inline int hlt_use_halt(void)
 void default_idle(void)
 {
 	if (hlt_use_halt()) {
-		trace_power_start(POWER_CSTATE, 1, smp_processor_id());
+		trace_power_switch_state(POWER_CSTATE, 1, 0,
+					 smp_processor_id());
 		current_thread_info()->status &= ~TS_POLLING;
 		/*
 		 * TS_POLLING-cleared state must be visible before we
@@ -442,7 +443,8 @@ EXPORT_SYMBOL_GPL(cpu_idle_wait);
  */
 void mwait_idle_with_hints(unsigned long ax, unsigned long cx)
 {
-	trace_power_start(POWER_CSTATE, (ax>>4)+1, smp_processor_id());
+	trace_power_switch_state(POWER_CSTATE, (ax>>4)+1, 0,
+				 smp_processor_id());
 	if (!need_resched()) {
 		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))
 			clflush((void *)&current_thread_info()->flags);
@@ -458,7 +460,8 @@ void mwait_idle_with_hints(unsigned long ax,
unsigned long cx)
 static void mwait_idle(void)
 {
 	if (!need_resched()) {
-		trace_power_start(POWER_CSTATE, 1, smp_processor_id());
+		trace_power_switch_state(POWER_CSTATE, 1, 0,
+					 smp_processor_id());
 		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))
 			clflush((void *)&current_thread_info()->flags);

@@ -479,11 +482,11 @@ static void mwait_idle(void)
  */
 static void poll_idle(void)
 {
-	trace_power_start(POWER_CSTATE, 0, smp_processor_id());
+	trace_power_switch_state(POWER_CSTATE, 0, 0, smp_processor_id());
 	local_irq_enable();
 	while (!need_resched())
 		cpu_relax();
-	trace_power_end(0);
+	trace_power_switch_state(POWER_CSTATE, 0, 0, smp_processor_id());
 }

 /*
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index 96586c3..b2415b2 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -114,7 +114,8 @@ void cpu_idle(void)
 			pm_idle();
 			start_critical_timings();

-			trace_power_end(smp_processor_id());
+			trace_power_switch_state(POWER_CSTATE, 0, 0,
+						 smp_processor_id());
 		}
 		tick_nohz_restart_sched_tick();
 		preempt_enable_no_resched();
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index b3d7a3a..64279b1 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -141,7 +141,8 @@ void cpu_idle(void)
 			pm_idle();
 			start_critical_timings();

-			trace_power_end(smp_processor_id());
+			trace_power_switch_state(POWER_CSTATE, 0, 0,
+						 smp_processor_id());

 			/* In many cases the interrupt that ended idle
 			   has already called exit_idle. But some idle
diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index 199dcb9..013c274 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -354,7 +354,8 @@ void cpufreq_notify_transition(struct
cpufreq_freqs *freqs, unsigned int state)
 		adjust_jiffies(CPUFREQ_POSTCHANGE, freqs);
 		dprintk("FREQ: %lu - CPU: %lu", (unsigned long)freqs->new,
 			(unsigned long)freqs->cpu);
-		trace_power_frequency(POWER_PSTATE, freqs->new, freqs->cpu);
+		trace_power_switch_state(POWER_PSTATE, freqs->new, freqs->cpu,
+					 smp_processor_id());
 		srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
 				CPUFREQ_POSTCHANGE, freqs);
 		if (likely(policy) && likely(policy->cpu == freqs->cpu))
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
index a507108..301261a 100644
--- a/drivers/cpuidle/cpuidle.c
+++ b/drivers/cpuidle/cpuidle.c
@@ -106,7 +106,7 @@ static void cpuidle_idle_call(void)
 	/* give the governor an opportunity to reflect on the outcome */
 	if (cpuidle_curr_governor->reflect)
 		cpuidle_curr_governor->reflect(dev);
-	trace_power_end(smp_processor_id());
+	trace_power_switch_state(POWER_CSTATE, 0, 0, smp_processor_id());
 }

 /**
diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
index 065c804..2b122f1 100755
--- a/drivers/idle/intel_idle.c
+++ b/drivers/idle/intel_idle.c
@@ -185,7 +185,7 @@ static int intel_idle(struct cpuidle_device *dev,
struct cpuidle_state *state)

 	stop_critical_timings();
 #ifndef MODULE
-	trace_power_start(POWER_CSTATE, (eax >> 4) + 1, cpu);
+	trace_power_switch_state(POWER_CSTATE, (eax >> 4) + 1, 0, cpu);
 #endif
 	if (!need_resched()) {

diff --git a/include/trace/events/power.h b/include/trace/events/power.h
index 286784d..192543a 100644
--- a/include/trace/events/power.h
+++ b/include/trace/events/power.h
@@ -18,61 +18,42 @@ enum {
 #endif

 /*
- * The power events are used for cpuidle & suspend (power_start, power_end)
- *  and for cpufreq (power_frequency)
+ * The power events are used for cpuidle & suspend and for cpufreq
  */
 DECLARE_EVENT_CLASS(power,

-	TP_PROTO(unsigned int type, unsigned int state, unsigned int cpu_id),
+	TP_PROTO(unsigned int type, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

-	TP_ARGS(type, state, cpu_id),
+	TP_ARGS(type, state, sub_state, cpu_id),

 	TP_STRUCT__entry(
 		__field(	u64,		type		)
 		__field(	u64,		state		)
+		__field(	u64,		sub_state	)
 		__field(	u64,		cpu_id		)
 	),

 	TP_fast_assign(
 		__entry->type = type;
 		__entry->state = state;
+		__entry->sub_state = sub_state;
 		__entry->cpu_id = cpu_id;
 	),

-	TP_printk("type=%lu state=%lu cpu_id=%lu", (unsigned long)__entry->type,
-		(unsigned long)__entry->state, (unsigned long)__entry->cpu_id)
+	TP_printk("type=%lu state=%lu sub_state=%lu cpu_id=%lu",
+		  (unsigned long)__entry->type,
+		  (unsigned long)__entry->state,
+		  (unsigned long)__entry->sub_state,
+		  (unsigned long)__entry->cpu_id)
 );

-DEFINE_EVENT(power, power_start,
+DEFINE_EVENT(power, power_switch_state,

-	TP_PROTO(unsigned int type, unsigned int state, unsigned int cpu_id),
-
-	TP_ARGS(type, state, cpu_id)
-);
-
-DEFINE_EVENT(power, power_frequency,
-
-	TP_PROTO(unsigned int type, unsigned int state, unsigned int cpu_id),
-
-	TP_ARGS(type, state, cpu_id)
-);
-
-TRACE_EVENT(power_end,
-
-	TP_PROTO(unsigned int cpu_id),
-
-	TP_ARGS(cpu_id),
-
-	TP_STRUCT__entry(
-		__field(	u64,		cpu_id		)
-	),
-
-	TP_fast_assign(
-		__entry->cpu_id = cpu_id;
-	),
-
-	TP_printk("cpu_id=%lu", (unsigned long)__entry->cpu_id)
+	TP_PROTO(unsigned int type, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

+	TP_ARGS(type, state, sub_state, cpu_id)
 );

 /*
@@ -81,45 +62,53 @@ TRACE_EVENT(power_end,
  */
 DECLARE_EVENT_CLASS(clock,

-	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+	TP_PROTO(const char *name, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

-	TP_ARGS(name, state, cpu_id),
+	TP_ARGS(name, state, sub_state, cpu_id),

 	TP_STRUCT__entry(
 		__string(       name,           name            )
 		__field(        u64,            state           )
+		__field(        u64,            sub_state       )
 		__field(        u64,            cpu_id          )
 	),

 	TP_fast_assign(
 		__assign_str(name, name);
 		__entry->state = state;
+		__entry->sub_state = sub_state;
 		__entry->cpu_id = cpu_id;
 	),

-	TP_printk("%s state=%lu cpu_id=%lu", __get_str(name),
-		(unsigned long)__entry->state, (unsigned long)__entry->cpu_id)
+	TP_printk("%s state=%lu sub_state=%lu cpu_id=%lu", __get_str(name),
+		  (unsigned long)__entry->state,
+		  (unsigned long)__entry->sub_state,
+		  (unsigned long)__entry->cpu_id)
 );

 DEFINE_EVENT(clock, clock_enable,

-	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+	TP_PROTO(const char *name, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

-	TP_ARGS(name, state, cpu_id)
+	TP_ARGS(name, state, sub_state, cpu_id)
 );

 DEFINE_EVENT(clock, clock_disable,

-	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+	TP_PROTO(const char *name, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

-	TP_ARGS(name, state, cpu_id)
+	TP_ARGS(name, state, sub_state, cpu_id)
 );

 DEFINE_EVENT(clock, clock_set_rate,

-	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+	TP_PROTO(const char *name, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

-	TP_ARGS(name, state, cpu_id)
+	TP_ARGS(name, state, sub_state, cpu_id)
 );

 /*
@@ -127,31 +116,37 @@ DEFINE_EVENT(clock, clock_set_rate,
  */
 DECLARE_EVENT_CLASS(power_domain,

-	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+	TP_PROTO(const char *name, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

-	TP_ARGS(name, state, cpu_id),
+	TP_ARGS(name, state, sub_state, cpu_id),

 	TP_STRUCT__entry(
 		__string(       name,           name            )
 		__field(        u64,            state           )
+		__field(        u64,            sub_state       )
 		__field(        u64,            cpu_id          )
 	),

 	TP_fast_assign(
 		__assign_str(name, name);
 		__entry->state = state;
+		__entry->sub_state = sub_state;
 		__entry->cpu_id = cpu_id;
 ),

-	TP_printk("%s state=%lu cpu_id=%lu", __get_str(name),
-		(unsigned long)__entry->state, (unsigned long)__entry->cpu_id)
+	TP_printk("%s state=%lu sub_state=%lu cpu_id=%lu", __get_str(name),
+		  (unsigned long)__entry->state,
+		  (unsigned long)__entry->sub_state,
+		  (unsigned long)__entry->cpu_id)
 );

 DEFINE_EVENT(power_domain, power_domain_target,

-	TP_PROTO(const char *name, unsigned int state, unsigned int cpu_id),
+	TP_PROTO(const char *name, unsigned int state,
+		 unsigned int sub_state, unsigned int cpu_id),

-	TP_ARGS(name, state, cpu_id)
+	TP_ARGS(name, state, sub_state, cpu_id)
 );

 #endif /* _TRACE_POWER_H */
diff --git a/kernel/trace/power-traces.c b/kernel/trace/power-traces.c
index a22582a..7da72b0 100644
--- a/kernel/trace/power-traces.c
+++ b/kernel/trace/power-traces.c
@@ -13,5 +13,3 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/power.h>

-EXPORT_TRACEPOINT_SYMBOL_GPL(power_frequency);
-
-- 
1.7.1
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] tracing, perf: add more power related events

[Arjan van de Ven]

  On 9/22/2010 8:31 AM, Jean Pihet wrote:
> Hi,
>
> Here is a patch that redefines the power events API. The advantages
> are: easier maintainance of the kernel and the
> user space tools, a cleaner and more generic interface, more
> parameters for fine tracing and even documentation!
>
> Thomas, this patch has your patch above merged in ('power-trace: Use
> power_switch_state instead of power_start and power_end'). The revised
> ACPI patch is coming asap.
>
> The trace points for x86 and OMAP are also udated accordingly.
>
> The pytimechart tool needs an update for the new API. This can be done
> as soon as the kernel code gets merged in.

unfortunately this code is changing a userspace ABI... we really 
shouldn't do that if we can avoid it,
and here we can avoid it.

applications ARE using this stuff!

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

On Wed, Sep 22, 2010 at 5:33 PM, Arjan van de Ven <arjan@linux.intel.com> wrote:
>  On 9/22/2010 8:31 AM, Jean Pihet wrote:
>>
>> Hi,
>>
>> Here is a patch that redefines the power events API. The advantages
>> are: easier maintainance of the kernel and the
>> user space tools, a cleaner and more generic interface, more
>> parameters for fine tracing and even documentation!
>>
>> Thomas, this patch has your patch above merged in ('power-trace: Use
>> power_switch_state instead of power_start and power_end'). The revised
>> ACPI patch is coming asap.
>>
>> The trace points for x86 and OMAP are also udated accordingly.
>>
>> The pytimechart tool needs an update for the new API. This can be done
>> as soon as the kernel code gets merged in.
>
> unfortunately this code is changing a userspace ABI... we really shouldn't
> do that if we can avoid it,
> and here we can avoid it.
>
> applications ARE using this stuff!
What are the apps that are using it? I know about builtin-timechart,
pytimechart. Is powertop using this as well?

Is it better to go for a 3 steps approach (add new API, change tools,
deprecate old API) like proposed above?

Thanks,
Jean
--
To unsubscribe from this list: send the line "unsubscribe linux-trace-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] tracing, perf: add more power related events

[Arjan van de Ven]

  On 9/22/2010 8:36 AM, Jean Pihet wrote:
> On Wed, Sep 22, 2010 at 5:33 PM, Arjan van de Ven<arjan@linux.intel.com>  wrote:
>>   On 9/22/2010 8:31 AM, Jean Pihet wrote:
>>> Hi,
>>>
>>> Here is a patch that redefines the power events API. The advantages
>>> are: easier maintainance of the kernel and the
>>> user space tools, a cleaner and more generic interface, more
>>> parameters for fine tracing and even documentation!
>>>
>>> Thomas, this patch has your patch above merged in ('power-trace: Use
>>> power_switch_state instead of power_start and power_end'). The revised
>>> ACPI patch is coming asap.
>>>
>>> The trace points for x86 and OMAP are also udated accordingly.
>>>
>>> The pytimechart tool needs an update for the new API. This can be done
>>> as soon as the kernel code gets merged in.
>> unfortunately this code is changing a userspace ABI... we really shouldn't
>> do that if we can avoid it,
>> and here we can avoid it.
>>
>> applications ARE using this stuff!
> What are the apps that are using it? I know about builtin-timechart,
> pytimechart. Is powertop using this as well?

powertop 2.x codebase is as well.

and a bunch of tools we have internal here at Intel.

the thing with ABIs is that you don't know how many users you have.. at 
least here you know the lower bound is 3 different tools that are open 
source.
.... and likely many local tools that aren't.

Re: [PATCH] tracing, perf: add more power related events

[Peter Zijlstra]

On Wed, 2010-09-22 at 09:32 -0700, Arjan van de Ven wrote:
> > What are the apps that are using it? I know about builtin-timechart,
> > pytimechart. Is powertop using this as well?
> 
> powertop 2.x codebase is as well.
> 
> and a bunch of tools we have internal here at Intel.
> 
> the thing with ABIs is that you don't know how many users you have.. at 
> least here you know the lower bound is 3 different tools that are open 
> source.
> .... and likely many local tools that aren't.

These tools should be smart enough to look up the tracepoint name, fail
it its not available, read the tracepoint format, again fail if not
compatible.

I really object to treating tracepoints as ABI and being tied to any
implementation details due to that.

Tools really should be flexible and allow for change.

Re: [PATCH] tracing, perf: add more power related events

[Arjan van de Ven]

  On 9/22/2010 9:43 AM, Peter Zijlstra wrote:
> On Wed, 2010-09-22 at 09:32 -0700, Arjan van de Ven wrote:
>>> What are the apps that are using it? I know about builtin-timechart,
>>> pytimechart. Is powertop using this as well?
>> powertop 2.x codebase is as well.
>>
>> and a bunch of tools we have internal here at Intel.
>>
>> the thing with ABIs is that you don't know how many users you have.. at
>> least here you know the lower bound is 3 different tools that are open
>> source.
>> .... and likely many local tools that aren't.
> These tools should be smart enough to look up the tracepoint name, fail
> it its not available, read the tracepoint format, again fail if not
> compatible.
it's not very useful if none of the critical information is available.

you can't at the one hand push people to use perf for critical pieces 
(like machine checks etc etc) and on
the other hand say that it's not ABI stable and should not be used really.

In this case we're talking about basically a suprious rename of 
something that isn't really an improvement
(because it makes it harder to subscribe to only one type of event)... 
that's not a good thing.

Re: [PATCH] tracing, perf: add more power related events

[Peter Zijlstra]

On Wed, 2010-09-22 at 10:06 -0700, Arjan van de Ven wrote:

> In this case we're talking about basically a suprious rename of 
> something that isn't really an improvement
> (because it makes it harder to subscribe to only one type of event)... 
> that's not a good thing.

People have been talking about more/more comprehensive power tracepoints
for a while, and I think that's a valid goal, if its really a rename I'm
sure you can work it out.

That said, I really didn't read this discussion much, but your stance
seems to be that any tracepoint you use must stay valid, and I object to
that.

What will do you do when we include a new scheduling policy and all the
scheduler tracepoints need to change? (yes that's really going to
happen)

I'm not going to carry double tracepoints, and I'm not going to not
merge that policy. 

Re: [PATCH] tracing, perf: add more power related events

[Steven Rostedt]

On Wed, 2010-09-22 at 19:30 +0200, Peter Zijlstra wrote:
> On Wed, 2010-09-22 at 10:06 -0700, Arjan van de Ven wrote:
> 

> That said, I really didn't read this discussion much, but your stance
> seems to be that any tracepoint you use must stay valid, and I object to
> that.

We could add a TRACE_EVENT_ABI() as Ingo has been suggesting. If
anything, it could mean that the given tracepoint will always have the
same name. And perhaps the data it holds will always be there, but may
also be extended.

> 
> What will do you do when we include a new scheduling policy and all the
> scheduler tracepoints need to change? (yes that's really going to
> happen)

The tracepoint sched_switch should stay the same. We may add more data,
but the comm, pid, prio => comm, pid, prio, I don't see going away.

> 
> I'm not going to carry double tracepoints, and I'm not going to not
> merge that policy. 

Not sure what you mean by "double tracepoints"

-- Steve

Re: [PATCH] tracing, perf: add more power related events

[Ingo Molnar]

* Steven Rostedt <rostedt@goodmis.org> wrote:

> We could add a TRACE_EVENT_ABI() as Ingo has been suggesting. [...]

That would be rather useful.

We could still be flexible about experimental tracepoints (they dont 
hurt), but apps should stick to ABI bits - or at least not complain when 
non-ABI tracepoints change for good reasons. (arbitrary, avoidable 
changes are still bad obviously, regardless of the type of the 
tracepoint)

We'd be more stringent with marking tracepoints an ABI.

So yes, this would be lovely to have.

	Ingo

Re: [PATCH] tracing, perf: add more power related events

[Peter Zijlstra]

On Wed, 2010-09-22 at 20:23 +0200, Ingo Molnar wrote:
> * Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > We could add a TRACE_EVENT_ABI() as Ingo has been suggesting. [...]
> 
> That would be rather useful.
> 
> We could still be flexible about experimental tracepoints (they dont 
> hurt), but apps should stick to ABI bits - or at least not complain when 
> non-ABI tracepoints change for good reasons. (arbitrary, avoidable 
> changes are still bad obviously, regardless of the type of the 
> tracepoint)
> 
> We'd be more stringent with marking tracepoints an ABI.
> 
> So yes, this would be lovely to have.

Why? As long as we mandate that apps need to fully parse the format
stuff anyway, it doesn't buy us much.

Re: [PATCH] tracing, perf: add more power related events

[Peter Zijlstra]

On Wed, 2010-09-22 at 14:15 -0400, Steven Rostedt wrote:
> On Wed, 2010-09-22 at 19:30 +0200, Peter Zijlstra wrote:
> > On Wed, 2010-09-22 at 10:06 -0700, Arjan van de Ven wrote:
> > 
> 
> > That said, I really didn't read this discussion much, but your stance
> > seems to be that any tracepoint you use must stay valid, and I object to
> > that.
> 
> We could add a TRACE_EVENT_ABI() as Ingo has been suggesting. If
> anything, it could mean that the given tracepoint will always have the
> same name. And perhaps the data it holds will always be there, but may
> also be extended.

I still don't see why you need TRACE_EVENT_ABI for that, if its the same
name and the format can be extended you get the same results with what
we've got. Apps need to read/parse the format thing anyway.

> > 
> > What will do you do when we include a new scheduling policy and all the
> > scheduler tracepoints need to change? (yes that's really going to
> > happen)
> 
> The tracepoint sched_switch should stay the same. We may add more data,
> but the comm, pid, prio => comm, pid, prio, I don't see going away.

Right, it would need additional fields. Preferably not only at the end.

> > I'm not going to carry double tracepoints, and I'm not going to not
> > merge that policy. 
> 
> Not sure what you mean by "double tracepoints"

Two different tracepoints in the same location.

Re: [PATCH] tracing, perf: add more power related events

[Steven Rostedt]

On Wed, 2010-09-22 at 20:26 +0200, Peter Zijlstra wrote:
> On Wed, 2010-09-22 at 14:15 -0400, Steven Rostedt wrote:
> > On Wed, 2010-09-22 at 19:30 +0200, Peter Zijlstra wrote:
> > > On Wed, 2010-09-22 at 10:06 -0700, Arjan van de Ven wrote:
> > > 
> > 
> > > That said, I really didn't read this discussion much, but your stance
> > > seems to be that any tracepoint you use must stay valid, and I object to
> > > that.
> > 
> > We could add a TRACE_EVENT_ABI() as Ingo has been suggesting. If
> > anything, it could mean that the given tracepoint will always have the
> > same name. And perhaps the data it holds will always be there, but may
> > also be extended.
> 
> I still don't see why you need TRACE_EVENT_ABI for that, if its the same
> name and the format can be extended you get the same results with what
> we've got. Apps need to read/parse the format thing anyway.

Just a marker that these trace points are being used by apps.

> 
> > > 
> > > What will do you do when we include a new scheduling policy and all the
> > > scheduler tracepoints need to change? (yes that's really going to
> > > happen)
> > 
> > The tracepoint sched_switch should stay the same. We may add more data,
> > but the comm, pid, prio => comm, pid, prio, I don't see going away.
> 
> Right, it would need additional fields. Preferably not only at the end.

Why not at the end? The tools should easily be able to represent them in
anyway they want. The print-fmt field helps in this regard too.

But then again, we present the fields in the data. The tools should use
a parse library (which a generic one will soon be out too). This way, we
don't need them at the "end" but the parsing tools could find the fields
no matter where they are in the record.

> 
> > > I'm not going to carry double tracepoints, and I'm not going to not

"not going to not" eeek! double negative!

> > > merge that policy. 
> > 
> > Not sure what you mean by "double tracepoints"
> 
> Two different tracepoints in the same location.

Agreed, that is wrong to have.

-- Steve

Re: [PATCH] tracing, perf: add more power related events

[Peter Zijlstra]

On Wed, 2010-09-22 at 14:36 -0400, Steven Rostedt wrote:
> > I still don't see why you need TRACE_EVENT_ABI for that, if its the same
> > name and the format can be extended you get the same results with what
> > we've got. Apps need to read/parse the format thing anyway.
> 
> Just a marker that these trace points are being used by apps.

I really don't see any additional value in that. Its not like we won't
change them if we have to.

> But then again, we present the fields in the data. The tools should use
> a parse library (which a generic one will soon be out too). This way, we
> don't need them at the "end" but the parsing tools could find the fields
> no matter where they are in the record. 

Right. They should use the full format specification.

Re: [PATCH] tracing, perf: add more power related events

[Frank Ch. Eigler]

Hi -

On Wed, Sep 22, 2010 at 08:26:40PM +0200, Peter Zijlstra wrote:
> [...]
> > Not sure what you mean by "double tracepoints"
> Two different tracepoints in the same location.

Another possibility may be to provide a backward-compatibility module
that maps new tracepoints to old ones.  On demand, it could attach to
the new tracepoints, and export those callbacks as clones of the old
pseudo-ABI: same old names & parameters.  If the new ones supply a
superset of events & data, this is technically possible.

- FChE

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

On Wednesday 22 September 2010 19:06:54 Arjan van de Ven wrote:
>   On 9/22/2010 9:43 AM, Peter Zijlstra wrote:
> > On Wed, 2010-09-22 at 09:32 -0700, Arjan van de Ven wrote:
> >>> What are the apps that are using it? I know about builtin-timechart,
> >>> pytimechart. Is powertop using this as well?
> >> powertop 2.x codebase is as well.
> >>
> >> and a bunch of tools we have internal here at Intel.
> >>
> >> the thing with ABIs is that you don't know how many users you have.. at
> >> least here you know the lower bound is 3 different tools that are open
> >> source.
> >> .... and likely many local tools that aren't.
> > These tools should be smart enough to look up the tracepoint name, fail
> > it its not available, read the tracepoint format, again fail if not
> > compatible.
> it's not very useful if none of the critical information is available.
> 
> you can't at the one hand push people to use perf for critical pieces 
> (like machine checks etc etc) and on
> the other hand say that it's not ABI stable and should not be used really.
I provided an ABI stable solution, by marking the broken parts deprecated, etc.
I'll rework the CONFIG_DEPRECATED_POWER_EVENTS (or similar config).

> In this case we're talking about basically a suprious rename of 
> something that isn't really an improvement
...

> (because it makes it harder to subscribe to only one type of event)... 
> that's not a good thing.
Finally there is some talking about the details.
You say they should be differed by the name and not the type?
Why does the type= attribute exist at all then?

/sys/kernel/debug/tracing/:[0]# cat available_events  |grep power
power:power_start
power:power_frequency
power:power_end

So which one do I set to get C-, processor sleep, state traces?

What you mean is that instead of passing the type= as attribute, an
additional macro layer could be put in or each type is statically defined.
The type itself needs not to get passed anymore then, because this
info is already in the name and you get:
power:power_cstates
power:power_pstates
power:power_sstates
or
power:power_processor_sleep
power:power_processor_frequency
power:power_machine_suspend
...
right?
This more looks like an interface that can get exposed
to userspace...

       Thomas

Re: [PATCH] tracing, perf: add more power related events

[Peter Zijlstra]

On Wed, 2010-09-22 at 09:32 -0700, Arjan van de Ven wrote:

Also, please don't cross-post to subscriber only lists, that's annoying
as hell.

(removed the discuss@lesswatts list)

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

On Wednesday 22 September 2010 17:33:01 Arjan van de Ven wrote:
>   On 9/22/2010 8:31 AM, Jean Pihet wrote:
> > Hi,
> >
> > Here is a patch that redefines the power events API. The advantages
> > are: easier maintainance of the kernel and the
> > user space tools, a cleaner and more generic interface, more
> > parameters for fine tracing and even documentation!
> >
> > Thomas, this patch has your patch above merged in ('power-trace: Use
> > power_switch_state instead of power_start and power_end'). The revised
> > ACPI patch is coming asap.
> >
> > The trace points for x86 and OMAP are also udated accordingly.
> >
> > The pytimechart tool needs an update for the new API. This can be done
> > as soon as the kernel code gets merged in.
> 
> unfortunately this code is changing a userspace ABI... we really 
> shouldn't do that if we can avoid it,
> and here we can avoid it.
Wow, this is your first post on this thread and you needed 1 min, 20 secs
to reply and comment on work that probably took several weeks and which is quite
an improvement.
You did not even start to read the documentation/patch in this minute, did you?

> applications ARE using this stuff!
The current interfaces do not make sense at all.
Things must be changed now before even more ARE using the broken stuff
and things get more complicated.
You should be happy that someone is helping/doing it.

    Thomas

Re: [PATCH] tracing, perf: add more power related events

[Arjan van de Ven]

  On 9/22/2010 10:57 AM, Thomas Renninger wrote:
> On Wed
>> unfortunately this code is changing a userspace ABI... we really
>> shouldn't do that if we can avoid it,
>> and here we can avoid it.

[do you really need to get personal?]

> Wow, this is your first post on this thread

it isn't.

(but I've been traveling too much in the last few weeks and have been 
rather overwhelmed in mail as a result)

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

On Wednesday 22 September 2010 20:13:19 Arjan van de Ven wrote:
>   On 9/22/2010 10:57 AM, Thomas Renninger wrote:
> > On Wed
> >> unfortunately this code is changing a userspace ABI... we really
> >> shouldn't do that if we can avoid it,
> >> and here we can avoid it.
> 
> [do you really need to get personal?]
Is that personal already?
But discussing details makes more sense and I have to appologize for
the needless comment.

> (but I've been traveling too much in the last few weeks and have been 
> rather overwhelmed in mail as a result)
Yes, time is short.
Sorry about that one. And thanks for your input, even your time is 
probably shorter than mine,

      Thomas

Re: [PATCH] tracing, perf: add more power related events

[Rafael J. Wysocki]

On Wednesday, September 22, 2010, Arjan van de Ven wrote:
>   On 9/22/2010 8:31 AM, Jean Pihet wrote:
> > Hi,
> >
> > Here is a patch that redefines the power events API. The advantages
> > are: easier maintainance of the kernel and the
> > user space tools, a cleaner and more generic interface, more
> > parameters for fine tracing and even documentation!
> >
> > Thomas, this patch has your patch above merged in ('power-trace: Use
> > power_switch_state instead of power_start and power_end'). The revised
> > ACPI patch is coming asap.
> >
> > The trace points for x86 and OMAP are also udated accordingly.
> >
> > The pytimechart tool needs an update for the new API. This can be done
> > as soon as the kernel code gets merged in.
> 
> unfortunately this code is changing a userspace ABI... we really 
> shouldn't do that if we can avoid it,
> and here we can avoid it.
> 
> applications ARE using this stuff!

Apart from this, could we rename things like POWER_CSTATE to CPU_POWER_CSTATE
and state clearly in the docs that this whole thing is about CPU power?

Thanks,
Rafael

Re: [PATCH] tracing, perf: add more power related events

[Rafael J. Wysocki]

[Dropping discuss@lesswatts.org from the CC list.]

On Wednesday, September 22, 2010, Jean Pihet wrote:
> Hi,
> 
> Here is a patch that redefines the power events API. The advantages
> are: easier maintainance of the kernel and the
> user space tools, a cleaner and more generic interface, more
> parameters for fine tracing and even documentation!
> 
> Thomas, this patch has your patch above merged in ('power-trace: Use
> power_switch_state instead of power_start and power_end'). The revised
> ACPI patch is coming asap.
> 
> The trace points for x86 and OMAP are also udated accordingly.
> 
> The pytimechart tool needs an update for the new API. This can be done
> as soon as the kernel code gets merged in.
> Please note the point below about the existing code in builtin-timechart.c.
> 
> On Sat, Sep 18, 2010 at 12:26 AM, Thomas Renninger <trenn@suse.de> wrote:
> > On Friday 17 September 2010 18:24:12 Ingo Molnar wrote:
> >>
> >> * Thomas Renninger <trenn@suse.de> wrote:
> >>
> >> > On Friday 17 September 2010 16:24:59 Ingo Molnar wrote:
> >
> >> [ You dont even have to document it, as good code is self-explanatory ;-) ]
> > I recently posted a patch exporting some things through /sys/kernel/debug/...
> > Greg complained that a file for Documentation/ABI/{testing,stable}/* is missing
> > and I fully agree.
> The proposed patch has the documentation in
> Documentation/trace/events-power.txt.
> 
> > If different userspace apps should make use of this (in above case nobody
> > than my debug userspace tool will...) and this should be called something like an API,
> > it should be documented and if something changes, it should
> > first be marked deprecated, etc.
> Note: the exsiting code in tools/perf/builtin-timechart.c needs an
> update for the new events API. Is this code still maintained? I not,
> could pytimechart be merged in instead?
> 
> Feedback is welcome!
> 
> >
> >     Thomas
> >
> 
> Thanks,
> Jean
> 
> ---
> From f37d1875050d33273b10e99497b4059b37e6680d Mon Sep 17 00:00:00 2001
> From: Jean Pihet <jean.pihet@newoldbits.com>
> Date: Wed, 22 Sep 2010 17:10:47 +0200
> Subject: [PATCH] tools, perf: redefine the power events API
> 
> Redefine the API with:
> - power_switch_state for C-, P- and S-states,
> - clock and power_domain events
> 
> The new API allows easier maintainance of the kernel and the
> user space tools, a cleaner and more generic interface,
> more parameters for fine tracing and even documentation.
> 
> The new events are used by the x86 and OMAP platforms.
> 
> Signed-off-by: Jean Pihet <j-pihet@ti.com>
> ---
>  Documentation/trace/events-power.txt |   70 +++++++++++++++++++++++++
>  arch/arm/mach-omap2/cpuidle34xx.c    |    3 +
>  arch/arm/mach-omap2/pm34xx.c         |   11 ++++
>  arch/arm/mach-omap2/powerdomain.c    |    3 +
>  arch/arm/plat-omap/clock.c           |   13 ++++-
>  arch/arm/plat-omap/cpu-omap.c        |    3 +
>  arch/x86/kernel/process.c            |   13 +++--
>  arch/x86/kernel/process_32.c         |    3 +-
>  arch/x86/kernel/process_64.c         |    3 +-
>  drivers/cpufreq/cpufreq.c            |    3 +-
>  drivers/cpuidle/cpuidle.c            |    2 +-
>  drivers/idle/intel_idle.c            |    2 +-
>  include/trace/events/power.h         |   95 ++++++++++++++++------------------
>  kernel/trace/power-traces.c          |    2 -
>  14 files changed, 161 insertions(+), 65 deletions(-)
>  create mode 100644 Documentation/trace/events-power.txt
> 
> diff --git a/Documentation/trace/events-power.txt
> b/Documentation/trace/events-power.txt
> new file mode 100644
> index 0000000..967f842
> --- /dev/null
> +++ b/Documentation/trace/events-power.txt
> @@ -0,0 +1,70 @@
> +
> +			Subsystem Trace Points: power
> +
> +The power tracing system captures events related to power transitions
> +within the kernel. Broadly speaking there are three major subheadings:
> +
> +  o Power state switch which reports events related to suspend (S-states),
> +     cpuidle (C-states) and cpufreq (P-states)
> +  o System clock related changes
> +  o Power domains related changes and transitions
> +
> +This document describes what each of the tracepoints is and why they
> +might be useful.
> +
> +Cf. include/trace/events/power.h for the events definitions.
> +
> +1. Power state switch events
> +============================
> +
> +power_switch_state	type=%lu state=%lu sub_state=%lu cpu_id=%lu
> +
> +The 'type' parameter takes one of those macros:
> + . POWER_NONE	= 0,
> + . POWER_CSTATE	= 1,	/* C-State */
> + . POWER_PSTATE	= 2,	/* Fequency change or DVFS */
> + . POWER_SSTATE	= 3,	/* Suspend */

This POWER_SSTATE thing seems to be totally artificial and omap-specific.

Why do you want it to be done this way?

Or is the ACPI handling added in the ACPI patch?  In which case, why don't you
put that power_switch_state(POWER_SSTATE, 1, 0, cpu) into
kernel/power/suspend.c:suspend_enter() (and analogously for
power_switch_state(POWER_SSTATE, 0, 0, cpu)).

Moreover, why is the cpu argument necessary for POWER_SSTATE at all?

Thanks,
Rafael

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

Hi,

Here is what I am proposing, in reply to all your comments:

1) rename the events to match Thomas's proposal:
   power:power_cpu_cstate
   power:power_cpu_pstate
   power:power_cpu_sstate
   ...

2) introduce a new Kconfig option CONFIG_DEPRECATED_POWER_EVENTS and
conditionally map a subset of the new events to the old ones for
backward compatibility with the existing user apps. The apps should be
converted to the new API asap,

3) update documentation

Other remarks here below:

On Wed, Sep 22, 2010 at 8:49 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
...
> This POWER_SSTATE thing seems to be totally artificial and omap-specific.
>
> Why do you want it to be done this way?
>
> Or is the ACPI handling added in the ACPI patch?  In which case, why don't you
> put that power_switch_state(POWER_SSTATE, 1, 0, cpu) into
> kernel/power/suspend.c:suspend_enter() (and analogously for
> power_switch_state(POWER_SSTATE, 0, 0, cpu)).
The ACPI code is not using the SSTATE event.
Indeed inserting a tracepoint at
kernel/power/suspend.c:suspend_enter() is more generic. I will correct
this.

> Moreover, why is the cpu argument necessary for POWER_SSTATE at all?
The cpu_id parameter is present in all events prototypes. This is not
needed. I will correct this.

>
> Thanks,
> Rafael
>

A new patch should be ready in the coming days. I would like to have
your opinion on the proposed rework before re-doing it all again.

Thanks,
Jean
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] tracing, perf: add more power related events

[Rafael J. Wysocki]

On Tuesday, September 28, 2010, Jean Pihet wrote:
> Hi,

Hi,

> Here is what I am proposing, in reply to all your comments:
> 
> 1) rename the events to match Thomas's proposal:
>    power:power_cpu_cstate
>    power:power_cpu_pstate
>    power:power_cpu_sstate

If that sstate thing is going to mean "suspend", then please drop it.
"Suspend" is not a state, let alone a CPU state.  It is a procedure by which
the (entire) system is put into a sleep state (that is not confined to CPUs).

>    ...
> 
> 2) introduce a new Kconfig option CONFIG_DEPRECATED_POWER_EVENTS and
> conditionally map a subset of the new events to the old ones for
> backward compatibility with the existing user apps. The apps should be
> converted to the new API asap,
> 
> 3) update documentation

Sounds reasonable.

> Other remarks here below:
> 
> On Wed, Sep 22, 2010 at 8:49 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> ...
> > This POWER_SSTATE thing seems to be totally artificial and omap-specific.
> >
> > Why do you want it to be done this way?
> >
> > Or is the ACPI handling added in the ACPI patch?  In which case, why don't you
> > put that power_switch_state(POWER_SSTATE, 1, 0, cpu) into
> > kernel/power/suspend.c:suspend_enter() (and analogously for
> > power_switch_state(POWER_SSTATE, 0, 0, cpu)).
> The ACPI code is not using the SSTATE event.
> Indeed inserting a tracepoint at
> kernel/power/suspend.c:suspend_enter() is more generic. I will correct
> this.

OK

> > Moreover, why is the cpu argument necessary for POWER_SSTATE at all?
> The cpu_id parameter is present in all events prototypes. This is not
> needed. I will correct this.

OK

Thanks,
Rafael

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

Hi,

On Tue, Sep 28, 2010 at 11:22 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> On Tuesday, September 28, 2010, Jean Pihet wrote:
>> Hi,
>
> Hi,
>
>> Here is what I am proposing, in reply to all your comments:
>>
>> 1) rename the events to match Thomas's proposal:
>>    power:power_cpu_cstate
>>    power:power_cpu_pstate
>>    power:power_cpu_sstate
>
> If that sstate thing is going to mean "suspend", then please drop it.
> "Suspend" is not a state, let alone a CPU state.  It is a procedure by which
> the (entire) system is put into a sleep state (that is not confined to CPUs).

Since suspend is tied to the power management of the system, is it ok
to rename it to e.g. power:system_suspend?

>
>>    ...
>>
>> 2) introduce a new Kconfig option CONFIG_DEPRECATED_POWER_EVENTS and
>> conditionally map a subset of the new events to the old ones for
>> backward compatibility with the existing user apps. The apps should be
>> converted to the new API asap,
>>
>> 3) update documentation
>
> Sounds reasonable.
OK

>
>> Other remarks here below:
>>
>> On Wed, Sep 22, 2010 at 8:49 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
>> ...
>> > This POWER_SSTATE thing seems to be totally artificial and omap-specific.
>> >
>> > Why do you want it to be done this way?
>> >
>> > Or is the ACPI handling added in the ACPI patch?  In which case, why don't you
>> > put that power_switch_state(POWER_SSTATE, 1, 0, cpu) into
>> > kernel/power/suspend.c:suspend_enter() (and analogously for
>> > power_switch_state(POWER_SSTATE, 0, 0, cpu)).
>> The ACPI code is not using the SSTATE event.
>> Indeed inserting a tracepoint at
>> kernel/power/suspend.c:suspend_enter() is more generic. I will correct
>> this.
>
> OK
>
>> > Moreover, why is the cpu argument necessary for POWER_SSTATE at all?
>> The cpu_id parameter is present in all events prototypes. This is not
>> needed. I will correct this.
>
> OK
>
> Thanks,
> Rafael
>

Thanks,
Jean
--
To unsubscribe from this list: send the line "unsubscribe linux-trace-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] tracing, perf: add more power related events

[Rafael J. Wysocki]

On Tuesday, September 28, 2010, Jean Pihet wrote:
> Hi,
> 
> On Tue, Sep 28, 2010 at 11:22 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> > On Tuesday, September 28, 2010, Jean Pihet wrote:
> >> Hi,
> >
> > Hi,
> >
> >> Here is what I am proposing, in reply to all your comments:
> >>
> >> 1) rename the events to match Thomas's proposal:
> >>    power:power_cpu_cstate
> >>    power:power_cpu_pstate
> >>    power:power_cpu_sstate
> >
> > If that sstate thing is going to mean "suspend", then please drop it.
> > "Suspend" is not a state, let alone a CPU state.  It is a procedure by which
> > the (entire) system is put into a sleep state (that is not confined to CPUs).
> 
> Since suspend is tied to the power management of the system, is it ok
> to rename it to e.g. power:system_suspend?

I think so.

Thanks,
Rafael

Re: [PATCH] tracing, perf: add more power related events

[Arjan van de Ven]

  On 9/28/2010 2:22 PM, Rafael J. Wysocki wrote:
> On Tuesday, September 28, 2010, Jean Pihet wrote:
>> Hi,
> Hi,
>
>> Here is what I am proposing, in reply to all your comments:
>>
>> 1) rename the events to match Thomas's proposal:
>>     power:power_cpu_cstate
>>     power:power_cpu_pstate
>>     power:power_cpu_sstate
> If that sstate thing is going to mean "suspend", then please drop it.
> "Suspend" is not a state, let alone a CPU state.  It is a procedure by which
> the (entire) system is put into a sleep state (that is not confined to CPUs).

there are also non-suspend S states, like S0i1 and S0i3 (supported in 
the current Intel "Moorestown" platform)

so it's slightly more complex than "just" suspend :)

Re: [PATCH] tracing, perf: add more power related events

[Rafael J. Wysocki]

On Tuesday, September 28, 2010, Arjan van de Ven wrote:
>   On 9/28/2010 2:22 PM, Rafael J. Wysocki wrote:
> > On Tuesday, September 28, 2010, Jean Pihet wrote:
> >> Hi,
> > Hi,
> >
> >> Here is what I am proposing, in reply to all your comments:
> >>
> >> 1) rename the events to match Thomas's proposal:
> >>     power:power_cpu_cstate
> >>     power:power_cpu_pstate
> >>     power:power_cpu_sstate
> > If that sstate thing is going to mean "suspend", then please drop it.
> > "Suspend" is not a state, let alone a CPU state.  It is a procedure by which
> > the (entire) system is put into a sleep state (that is not confined to CPUs).
> 
> there are also non-suspend S states, like S0i1 and S0i3 (supported in 
> the current Intel "Moorestown" platform)
> 
> so it's slightly more complex than "just" suspend :)

That's exactly why I used the conditional above. :-)

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

On Tuesday 28 September 2010 23:45:24 Arjan van de Ven wrote:
>   On 9/28/2010 2:22 PM, Rafael J. Wysocki wrote:
> > On Tuesday, September 28, 2010, Jean Pihet wrote:
> >> Hi,
> > Hi,
> >
> >> Here is what I am proposing, in reply to all your comments:
> >>
> >> 1) rename the events to match Thomas's proposal:
> >>     power:power_cpu_cstate
> >>     power:power_cpu_pstate
> >>     power:power_cpu_sstate
I'd not name it that X86/ACPI specific.
    power:processor_sleep
    power:processor_frequency
    power:system_suspend
This would map with X86/ACPI c/p/s states but the name would
also match fine with ARM and other archs.

> > If that sstate thing is going to mean "suspend", then please drop
> > it.
> > "Suspend" is not a state, let alone a CPU state.  It is a procedure
> > by which the (entire) system is put into a sleep state (that is not
> > confined to CPUs).
> 
> there are also non-suspend S states, like S0i1 and S0i3 (supported in 
> the current Intel "Moorestown" platform)
> 
> so it's slightly more complex than "just" suspend :)
Something specific for this arch could get introduced, similar as
Jean did for the ARM specifics, e.g.:
power:moorestown_suspend
Intel probably invented three names for this new technique, one
might fit as an event name?
Depending whether extra info needs passed through this event it
could also use
    power:system_suspend
and pass a suspend state of #define S0i1 0x100, #define S0i2 0x101...

I try to find time to come up with another cleanup patch.
I also want to look at perf timechart then where I remember some ugly
hacks with C-state accounting and the broken state_start, state_end and 
frequency_switch events. Hope it won't get too ugly and perf timechart 
can support both, the old and the cleaned up events for a while.

    Thomas

Re: [PATCH] tracing, perf: add more power related events

[Jean Pihet]

Hi,

On Wed, Sep 29, 2010 at 9:49 AM, Thomas Renninger <trenn@suse.de> wrote:
> On Tuesday 28 September 2010 23:45:24 Arjan van de Ven wrote:
>>   On 9/28/2010 2:22 PM, Rafael J. Wysocki wrote:
>> > On Tuesday, September 28, 2010, Jean Pihet wrote:
>> >> Hi,
>> > Hi,
>> >
>> >> Here is what I am proposing, in reply to all your comments:
>> >>
>> >> 1) rename the events to match Thomas's proposal:
>> >>     power:power_cpu_cstate
>> >>     power:power_cpu_pstate
>> >>     power:power_cpu_sstate
> I'd not name it that X86/ACPI specific.
>    power:processor_sleep
>    power:processor_frequency
>    power:system_suspend
> This would map with X86/ACPI c/p/s states but the name would
> also match fine with ARM and other archs.
Good for me!

>
>> > If that sstate thing is going to mean "suspend", then please drop
>> > it.
>> > "Suspend" is not a state, let alone a CPU state.  It is a procedure
>> > by which the (entire) system is put into a sleep state (that is not
>> > confined to CPUs).
>>
>> there are also non-suspend S states, like S0i1 and S0i3 (supported in
>> the current Intel "Moorestown" platform)
>>
>> so it's slightly more complex than "just" suspend :)
> Something specific for this arch could get introduced, similar as
> Jean did for the ARM specifics, e.g.:
> power:moorestown_suspend
I would not introduce arch specific events. Your other proposal below
is more generic.

> Intel probably invented three names for this new technique, one
> might fit as an event name?
> Depending whether extra info needs passed through this event it
> could also use
>    power:system_suspend
> and pass a suspend state of #define S0i1 0x100, #define S0i2 0x101...
I am OK with that.

>
> I try to find time to come up with another cleanup patch.
> I also want to look at perf timechart then where I remember some ugly
> hacks with C-state accounting and the broken state_start, state_end and
> frequency_switch events. Hope it won't get too ugly and perf timechart
> can support both, the old and the cleaned up events for a while.
About pytimechart, I think it should be updated with the support for
the new events only. The current version is not perfect but supports
the current events correctly. I am planning to celan up and upgrade
that tool when the new API is out.

If that could force people to upgrade to the new events API asap...

>
>    Thomas
>
I know you want to see real code. Let me come with a respin of the
patch asap (it is a matter of days).

Thanks,
Jean
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [linux-pm] [PATCH] tracing, perf: add more power related events

[Pavel Machek]

Hi!

> >> Here is what I am proposing, in reply to all your comments:
> >>
> >> 1) rename the events to match Thomas's proposal:
> >>     power:power_cpu_cstate
> >>     power:power_cpu_pstate
> >>     power:power_cpu_sstate
> > If that sstate thing is going to mean "suspend", then please drop it.
> > "Suspend" is not a state, let alone a CPU state.  It is a procedure by which
> > the (entire) system is put into a sleep state (that is not confined to CPUs).
> 
> there are also non-suspend S states, like S0i1 and S0i3 (supported in 
> the current Intel "Moorestown" platform)

Where can one learn more?
								Pavel


-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

On Friday 17 September 2010 16:05:51 Jean Pihet wrote:
> Hi Thomas,
> 
> On Fri, Sep 17, 2010 at 3:08 PM, Thomas Renninger <trenn@suse.de> wrote:
> > Hi,
> >
> > I had a quick look at this and it's amazing how broken
> > the whole power event tracing interfaces are.
> > It's not your fault, Jean, they always were and adding your stuff is
> > fine.
> That is the whole point! This code needs a serious clean-up and reorganization.
> The patch I submitted is now merged in the tip tree, but the can be
> corrected once the new power trace API is agreed on.
Your generic patch is ok. Not that sure about the clock stuff.
How many different clocks do exist on ARM/OMAP?
Possibly this can all go into a:
power_switch_state(type=, state=)
interface and you define some ARM specific static ones like:
+       POWER_NONE      = 0,
+       POWER_CSTATE    = 1,    /* C-State */
+       POWER_PSTATE    = 2,    /* Fequency change or DVFS */
+       POWER_SSTATE    = 3,    /* Suspend */
+       POWER_OMAP_CLOCK_1   = 4,    /* OMAP clock XY */
+       POWER_OMAP_CLOCK_2   = 5,    /* OMAP clock XYZ */
...
> 
> > Some questions, maybe I've overseen something:
> >
> > Why does this event:
> > DEFINE_EVENT(power, power_frequency,
> > exist and takes a C-/P-state, now also an S-state type as argument?
> There is no clear link between the POWER_ types enum and the API, that
> needs to change.
Yep. At least for the frequency (P-state) and idle (C-state) states
I have a reasonable solution, see below.
> 
> > It should be named more generic, like:
> > DEFINE_EVENT(power, power_switch,
> > then it could get invoked when any P-/C-/S-/X- state
> > switch happened.
> Agree. We need some better definitions for the events and at the same
> time a more flexible way to pass extra arguments (e.g. like a return
> code or the real HW state in case the desired state is not reached).
> Also it is required to be able to track sub-states.
> The idea is to identify the C-states latency in the low power entry
> and exit paths, using 2 new macros in the enum and a sub state
> argument. Cf. http://omappedia.org/images/b/b9/Pytimechart_screenshot9_rs.jpg
> (from http://omappedia.org/wiki/Power_Management_Debug_and_Profiling).
IMO one does not need an exit path, see below.
 
> > What kind of hack is this:
> > TRACE_EVENT(power_end,
> > showing up as:
> > power_end: dummy=65535
> > What ends here?
> > I know it's a workaround/hack for userspace apps to be
> > able to detect leaving of a sleep state, but how would someone
> > know or how would someone describe this in a proper API documentation?
> That was a hack, it is now gone. The new power_end only takes the
> cpu_id argument.
The CPU/machine/device always is in a C-/P-/T-/S- whatever state.
Typically 0 means it's in a non-power saving state.
So issueing a power_switch_state(type=X, state=0)
would be the exit path. There is no need for a separate:
power_end(type=X) call.
Compare with my S-state ACPI example, I'll send in a minute.

> > Apropos documentation..., are the power trace events documented
> > somewhere?
> No. We need something like Documentation/trace/events-kmem.txt. I can
> write that with for the new power API.
> 
> > At least the state should still be passed, then the _start/_end thing
> > can be reused by something else than C-states.
> >
> > I can't see the use of having _start/_end events at all.
> > You are always in a state, having one:
> > power_switch
> > as suggested above, is enough to track everything.
> Most of the events can be converted to power_switch, but not all.
> Example: you need to know when the suspend ends.
power_switch_state(type=POWER_SSTATE, state=0)

> >
> > Examples:
> >
> > Today's C-state tracking:
> > power_start: type=1 state=1  -> C1 entered
> > power_end: dummy=65535       -> C-state left
> > power_start: type=1 state=2  -> C2 entered
> > power_end: dummy=65535       -> C-state left
> > would then be:
> > power_switch: type=1 state=1  -> C1 entered
> > power_switch: type=1 state=0  -> C0 entered -> means C1 left...
> > power_switch: type=1 state=2  -> C2 entered
> > power_switch: type=1 state=0  -> C0 entered -> means C2 left...
> > ...
> >
> > Todays P-state tracking:
> > power_frequency: type=2 state=125000000     -> P-state change to 125 MHz
> > power_frequency: type=2 state=90000000      -> P-state change to 90 MHz
> > would then be:
> > power_switch: type=2 state=125000000       -> P-state change to 125 MHz
> > power_switch: type=2 state=90000000        -> P-state change to 90 MHz
> >
> > The S-state and T-state tracking would fit into that without
> > modification.
> > Thinking one step further, a possibility to track D-states would
> What are the T- and D- states?
T-states are processor throttling. Should only get used in critical
thermal conditions, fits nice into this interface.

D-states are device states, defined in ACPI from D0 (no power savings
active, up to D3 (max powersavings active). It's up to the device what
it's doing with it.
This one does not fit into a generic power_switch_state, because:
   - you want a connection to the device exported
   - possibly you want to be smarter than ACPI and tell what exactly
     a device switches off.
I'd leave device specific events out of the discussion for now.

> > need an additional field, a pointer to the device, best a sysfs path
> > or similar.
> Agree on that. As said above I would like to have extra args.
> 
> > Jean, I do not think tracing the S-state with power_start is a
> > good idea. Best would be the power_frequency gets renamed
> 
> >(yes, breaks
> > userspace, but those have to be adjusted)
> BTW I can patch pytimechart. What other tools have to change? powertop ...?
That would be great.
AFAIK the code use trace/perf events in powertop is not yet released.

> > and used for P- and S-
> > state tracking (and C-state tracking as well, but this would need
> > additional userspace modifications).
> OK
> 
> > How do you track when the S-states end?
> Two options: 1) have new arguments that indicates enter/exit and a
> sub-state; 2) a new event fot the exit.
> I am in favor of 1 which is more generic.
See above, I'd use:
power_switch_state(type=POWER_SSTATE, state=0)

I'll post patches to get C-states and S-states making use of
a generic:
power_switch_state(...)
function with still providing old events to not break existing
tools.

    Thomas

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

Some patches for cleanup...
compile tested only...
Should not break existing user space apps, but they should
get converted asap to use power_swtich_state...

---

power-trace: Rename power frequency to power_switch_state

this interface/function is not intended for frequency changes
only, but should get used for any P- (processor frequency),
C- (processor sleep), T- (processor throttling), or S- (machine sleep)
state.

Since it's used in cpufreq.c which must be compiled in
and not in acpi-cpufreq.c anymore there is no need to export it
for modules.

Signed-off-by: Thomas Renninger <trenn@suse.de>

---
 drivers/cpufreq/cpufreq.c    |    1 +
 include/trace/events/power.h |    7 +++++++
 kernel/trace/power-traces.c  |    1 -
 3 files changed, 8 insertions(+), 1 deletion(-)

Index: linux-2.6.35-master/drivers/cpufreq/cpufreq.c
===================================================================
--- linux-2.6.35-master.orig/drivers/cpufreq/cpufreq.c
+++ linux-2.6.35-master/drivers/cpufreq/cpufreq.c
@@ -354,6 +354,7 @@ void cpufreq_notify_transition(struct cp
 		adjust_jiffies(CPUFREQ_POSTCHANGE, freqs);
 		dprintk("FREQ: %lu - CPU: %lu", (unsigned long)freqs->new,
 			(unsigned long)freqs->cpu);
+		trace_power_switch_state(POWER_PSTATE, freqs->new, freqs->cpu);
 		trace_power_frequency(POWER_PSTATE, freqs->new, freqs->cpu);
 		srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
 				CPUFREQ_POSTCHANGE, freqs);
Index: linux-2.6.35-master/include/trace/events/power.h
===================================================================
--- linux-2.6.35-master.orig/include/trace/events/power.h
+++ linux-2.6.35-master/include/trace/events/power.h
@@ -38,6 +38,13 @@ DECLARE_EVENT_CLASS(power,
 		(unsigned long)__entry->state, (unsigned long)__entry->cpu_id)
 );
 
+DEFINE_EVENT(power, power_switch_state,
+
+	TP_PROTO(unsigned int type, unsigned int state, unsigned int cpu_id),
+
+	TP_ARGS(type, state, cpu_id)
+);
+
 DEFINE_EVENT(power, power_start,
 
 	TP_PROTO(unsigned int type, unsigned int state, unsigned int cpu_id),
Index: linux-2.6.35-master/kernel/trace/power-traces.c
===================================================================
--- linux-2.6.35-master.orig/kernel/trace/power-traces.c
+++ linux-2.6.35-master/kernel/trace/power-traces.c
@@ -13,5 +13,4 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/power.h>
 
-EXPORT_TRACEPOINT_SYMBOL_GPL(power_frequency);
 

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

power-trace: Use power_switch_state instead of power_start and power_end

No need to have power_start and power_end. power_switch_state of state=0
means we exited power saving state. Userspace has all the information
it needs to detect power enter/exit case.

Export it, so that intel_idle can make use of it, even if compiled as module.

Signed-off-by: Thomas Renninger <trenn@suse.de>

---
 arch/x86/kernel/process.c   |    9 +++++++--
 drivers/cpuidle/cpuidle.c   |    3 +++
 drivers/idle/intel_idle.c   |    3 +++
 kernel/trace/power-traces.c |    2 +-
 4 files changed, 14 insertions(+), 3 deletions(-)

Index: linux-2.6.35-master/arch/x86/kernel/process.c
===================================================================
--- linux-2.6.35-master.orig/arch/x86/kernel/process.c
+++ linux-2.6.35-master/arch/x86/kernel/process.c
@@ -375,7 +375,10 @@ static inline int hlt_use_halt(void)
 void default_idle(void)
 {
 	if (hlt_use_halt()) {
+		/* trace_power_start is deprecated, remove it after a while */
 		trace_power_start(POWER_CSTATE, 1, smp_processor_id());
+		trace_power_switch_state(POWER_CSTATE, 1, smp_processor_id());
+
 		current_thread_info()->status &= ~TS_POLLING;
 		/*
 		 * TS_POLLING-cleared state must be visible before we
@@ -446,6 +449,8 @@ EXPORT_SYMBOL_GPL(cpu_idle_wait);
 void mwait_idle_with_hints(unsigned long ax, unsigned long cx)
 {
 	trace_power_start(POWER_CSTATE, (ax>>4)+1, smp_processor_id());
+	trace_power_switch_state(POWER_CSTATE, (ax>>4)+1, smp_processor_id());
+
 	if (!need_resched()) {
 		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))
 			clflush((void *)&current_thread_info()->flags);
@@ -462,6 +467,8 @@ static void mwait_idle(void)
 {
 	if (!need_resched()) {
 		trace_power_start(POWER_CSTATE, 1, smp_processor_id());
+		trace_power_switch_state(POWER_CSTATE, 1, smp_processor_id());
+
 		if (cpu_has(&current_cpu_data, X86_FEATURE_CLFLUSH_MONITOR))
 			clflush((void *)&current_thread_info()->flags);
 
@@ -482,11 +489,9 @@ static void mwait_idle(void)
  */
 static void poll_idle(void)
 {
-	trace_power_start(POWER_CSTATE, 0, smp_processor_id());
 	local_irq_enable();
 	while (!need_resched())
 		cpu_relax();
-	trace_power_end(0);
 }
 
 /*
Index: linux-2.6.35-master/drivers/cpuidle/cpuidle.c
===================================================================
--- linux-2.6.35-master.orig/drivers/cpuidle/cpuidle.c
+++ linux-2.6.35-master/drivers/cpuidle/cpuidle.c
@@ -106,7 +106,10 @@ static void cpuidle_idle_call(void)
 	/* give the governor an opportunity to reflect on the outcome */
 	if (cpuidle_curr_governor->reflect)
 		cpuidle_curr_governor->reflect(dev);
+
+	/* trace_power_end is deprecated, remove it after a while */
 	trace_power_end(smp_processor_id());
+	trace_power_switch_state(POWER_CSTATE, 0, smp_processor_id());
 }
 
 /**
Index: linux-2.6.35-master/drivers/idle/intel_idle.c
===================================================================
--- linux-2.6.35-master.orig/drivers/idle/intel_idle.c
+++ linux-2.6.35-master/drivers/idle/intel_idle.c
@@ -192,8 +192,11 @@ static int intel_idle(struct cpuidle_dev
 
 	stop_critical_timings();
 #ifndef MODULE
+	/* trace_power_start is deprecated, remove it after a while */
 	trace_power_start(POWER_CSTATE, (eax >> 4) + 1, cpu);
 #endif
+	trace_power_switch_state(POWER_CSTATE, (eax >> 4) + 1, smp_processor_id());
+
 	if (!need_resched()) {
 
 		__monitor((void *)&current_thread_info()->flags, 0, 0);
Index: linux-2.6.35-master/kernel/trace/power-traces.c
===================================================================
--- linux-2.6.35-master.orig/kernel/trace/power-traces.c
+++ linux-2.6.35-master/kernel/trace/power-traces.c
@@ -13,4 +13,4 @@
 #define CREATE_TRACE_POINTS
 #include <trace/events/power.h>
 
-
+EXPORT_TRACEPOINT_SYMBOL_GPL(power_switch_state);

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

power-trace: Add x86 ACPI S- (machine sleep) state events.

Signed-off-by: Thomas Renninger <trenn@suse.de>

---
 drivers/acpi/sleep.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

Index: linux-2.6.35-master/drivers/acpi/sleep.c
===================================================================
--- linux-2.6.35-master.orig/drivers/acpi/sleep.c
+++ linux-2.6.35-master/drivers/acpi/sleep.c
@@ -17,6 +17,8 @@
 #include <linux/suspend.h>
 #include <linux/reboot.h>
 
+#include <trace/events/power.h>
+
 #include <asm/io.h>
 
 #include <acpi/acpi_bus.h>
@@ -249,14 +251,19 @@ static int acpi_suspend_enter(suspend_st
 	unsigned long flags = 0;
 	u32 acpi_state = acpi_target_sleep_state;
 
+	trace_power_switch_state(POWER_SSTATE, acpi_state, 0);
+
 	ACPI_FLUSH_CPU_CACHE();
 
 	/* Do arch specific saving of state. */
 	if (acpi_state == ACPI_STATE_S3) {
 		int error = acpi_save_state_mem();
 
-		if (error)
+		if (error) {
+			trace_power_switch_state(POWER_SSTATE,
+						 ACPI_STATE_S0, 0);
 			return error;
+		}
 	}
 
 	local_irq_save(flags);
@@ -309,6 +316,8 @@ static int acpi_suspend_enter(suspend_st
 
 	suspend_nvs_restore();
 
+	trace_power_switch_state(POWER_SSTATE, ACPI_STATE_S0, 0);
+
 	return ACPI_SUCCESS(status) ? 0 : -EFAULT;
 }
 

Re: [PATCH] tracing, perf: add more power related events

[Thomas Renninger]

DO NOT APPLY THIS ONE!!!
	
The others should go into a mainline tree if Jean is ok with them.

This one does not work, due to some include dependencies or whatever
else I can't see right now.

The idea: Provide old trace power interfaces via .config option to not break
existing perf/powertop/.. binaries.
Not sure whether it's worth looking at it further. It shouldn't be needed...

       Thomas

---
 include/trace/events/power.h |   11 +++++++++++
 kernel/trace/Kconfig         |    4 ++++
 2 files changed, 15 insertions(+)

Index: linux-2.6.35-master/kernel/trace/Kconfig
===================================================================
--- linux-2.6.35-master.orig/kernel/trace/Kconfig
+++ linux-2.6.35-master/kernel/trace/Kconfig
@@ -64,6 +64,10 @@ config EVENT_TRACING
 	select CONTEXT_SWITCH_TRACER
 	bool
 
+config DEPRECATED_POWER_EVENT_TRACING
+       bool
+       default n
+
 config CONTEXT_SWITCH_TRACER
 	bool
 
Index: linux-2.6.35-master/include/trace/events/power.h
===================================================================
--- linux-2.6.35-master.orig/include/trace/events/power.h
+++ linux-2.6.35-master/include/trace/events/power.h
@@ -50,6 +50,7 @@ DEFINE_EVENT(power, power_switch_state,
 	TP_ARGS(type, state, cpu_id)
 );
 
+#ifdef CONFIG_DEPRECATED_POWER_EVENT_TRACING
 DEFINE_EVENT(power, power_start,
 
 	TP_PROTO(unsigned int type, unsigned int state, unsigned int cpu_id),
@@ -81,6 +82,16 @@ TRACE_EVENT(power_end,
 	TP_printk("cpu_id=%lu", (unsigned long)__entry->cpu_id)
 
 );
+#else
+inline void trace_power_end(unsigned int cpu_id)
+{}
+inline void trace_power_start(unsigned int type, unsigned int state,
+			      unsigned int cpu_id)
+{}
+inline void trace_power_frequency(unsigned int type, unsigned int state,
+				  unsigned int cpu_id)
+{}
+#endif
 
 /*
  * The clock events are used for clock enable/disable and for