Re: [PATCH v1] PM / Domains: Add tracepoints

[Matthias Kaehlcke]

On Fri, Sep 27, 2019 at 04:42:39AM -0400, Steven Rostedt wrote:
> On Thu, 26 Sep 2019 15:04:38 -0700
> Matthias Kaehlcke <mka@xxxxxxxxxxxx> wrote:
> 
> > Define genpd_power_on/off and genpd_set_performance_state
> > tracepoints and use them.
> 
> I agree with Greg about adding a "why" you need this. But, in case
> there's a good reason to have this, I have comments about the code
> below.

Thanks Greg and Steven for your comments.

How about this instead:

  Add tracepoints for genpd_power_on, genpd_power_off and
  genpd_set_performance_state. The tracepoints can help with
  understanding power domain behavior of a given device, which
  may be particularly interesting for battery powered devices
  and suspend/resume.

> > Signed-off-by: Matthias Kaehlcke <mka@xxxxxxxxxxxx>
> > ---
> > 
> >  drivers/base/power/domain.c  | 27 +++++++++++++++++---
> >  include/trace/events/power.h | 49 ++++++++++++++++++++++++++++++++++++
> >  2 files changed, 72 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
> > index cc85e87eaf05..aee988c112e5 100644
> > --- a/drivers/base/power/domain.c
> > +++ b/drivers/base/power/domain.c
> > @@ -21,6 +21,7 @@
> >  #include <linux/suspend.h>
> >  #include <linux/export.h>
> >  #include <linux/cpu.h>
> > +#include <trace/events/power.h>
> >  
> >  #include "power.h"
> >  
> > @@ -329,6 +330,9 @@ static int _genpd_set_performance_state(struct generic_pm_domain *genpd,
> >  		goto err;
> >  
> >  	genpd->performance_state = state;
> > +
> > +	trace_genpd_set_performance_state(genpd);
> > +
> >  	return 0;
> >  
> >  err:
> > @@ -418,14 +422,21 @@ static int _genpd_power_on(struct generic_pm_domain *genpd, bool timed)
> >  	if (!genpd->power_on)
> >  		return 0;
> >  
> > -	if (!timed)
> > -		return genpd->power_on(genpd);
> > +	if (!timed) {
> > +		ret = genpd->power_on(genpd);
> > +		if (!ret)
> > +			trace_genpd_power_on(genpd);
> > +
> > +		return ret;
> > +	}
> >  
> >  	time_start = ktime_get();
> >  	ret = genpd->power_on(genpd);
> >  	if (ret)
> >  		return ret;
> >  
> > +	trace_genpd_power_on(genpd);
> > +
> >  	elapsed_ns = ktime_to_ns(ktime_sub(ktime_get(), time_start));
> >  	if (elapsed_ns <= genpd->states[state_idx].power_on_latency_ns)
> >  		return ret;
> > @@ -448,14 +459,22 @@ static int _genpd_power_off(struct generic_pm_domain *genpd, bool timed)
> >  	if (!genpd->power_off)
> >  		return 0;
> >  
> > -	if (!timed)
> > -		return genpd->power_off(genpd);
> > +	if (!timed) {
> > +		ret = genpd->power_off(genpd);
> > +		if (!ret)
> 
> Here
> 
> > +			trace_genpd_power_off(genpd);
> > +
> > +		return ret;
> > +	}
> >  
> >  	time_start = ktime_get();
> >  	ret = genpd->power_off(genpd);
> >  	if (ret)
> >  		return ret;
> >  
> > +	if (!ret)
> 
> And here add a conditional branch for only a tracepoint. To eliminate
> the branch when tracepoints are not enabled, please do it this way
> instead:
> 
> 	if (trace_genpd_power_off_enabled() && !ret)
> 
> The above is a static branch (nop when disabled, and jmp when enabled),
> and the above should move the conditional branch on !ret into the
> section that is only called when the tracepoint is enabled.

ok, will do in future versions.