Messages in this thread | | | Date | Fri, 12 Nov 2010 15:20:47 +0100 | Subject | Re: [PATCH 2/3] PERF(kernel): Cleanup power events | From | Jean Pihet <> |
| |
Thomas,
Thanks for the patches re-spin!
Here are my comments inlined.
On Thu, Nov 11, 2010 at 7:03 PM, Thomas Renninger <trenn@suse.de> wrote: > Recent changes: > - Enable EVENT_POWER_TRACING_DEPRECATED by default > > New power trace events: > power:cpu_idle > power:cpu_frequency > power:machine_suspend > > > C-state/idle accounting events: > power:power_start > power:power_end > are replaced with: > power:cpu_idle > > and > power:power_frequency > is replaced with: > power:cpu_frequency > > power:machine_suspend > is newly introduced. > Jean Pihet has a patch integrated into the generic layer > (kernel/power/suspend.c) which will make use of it. > > the type= field got removed from both, it was never > used and the type is differed by the event type itself. > > perf timechart > userspace tool gets adjusted in a separate patch. > > Signed-off-by: Thomas Renninger <trenn@suse.de> > Acked-by: Arjan van de Ven <arjan@linux.intel.com> > Acked-by: Jean Pihet <jean.pihet@newoldbits.com> > CC: Arjan van de Ven <arjan@linux.intel.com> > CC: Ingo Molnar <mingo@elte.hu> > CC: rjw@sisk.pl > CC: linux-kernel@vger.kernel.org > --- > arch/x86/kernel/process.c | 7 +++- > arch/x86/kernel/process_32.c | 2 +- > arch/x86/kernel/process_64.c | 2 + > drivers/cpufreq/cpufreq.c | 1 + > drivers/cpuidle/cpuidle.c | 1 + > drivers/idle/intel_idle.c | 1 + > include/trace/events/power.h | 87 +++++++++++++++++++++++++++++++++++++++++- > kernel/trace/Kconfig | 15 +++++++ > kernel/trace/power-traces.c | 3 + > 9 files changed, 116 insertions(+), 3 deletions(-) > ... > diff --git a/include/trace/events/power.h b/include/trace/events/power.h > index 286784d..ab26d8e 100644 > --- a/include/trace/events/power.h > +++ b/include/trace/events/power.h > @@ -7,6 +7,67 @@ > #include <linux/ktime.h> > #include <linux/tracepoint.h> > > +DECLARE_EVENT_CLASS(cpu, > + > + TP_PROTO(unsigned int state, unsigned int cpu_id), > + > + TP_ARGS(state, cpu_id), > + > + TP_STRUCT__entry( > + __field( u32, state ) > + __field( u32, cpu_id ) > + ), > + > + TP_fast_assign( > + __entry->state = state; > + __entry->cpu_id = cpu_id; > + ), > + > + TP_printk("state=%lu cpu_id=%lu", (unsigned long)__entry->state, > + (unsigned long)__entry->cpu_id) Using %lu for the state field causes PWR_EVENT_EXIT to appear as 4294967295 instead of -1. Can the field be of a signed type?
> +); > + > +DEFINE_EVENT(cpu, cpu_idle, > + > + TP_PROTO(unsigned int state, unsigned int cpu_id), > + > + TP_ARGS(state, cpu_id) > +); > + > +/* This file can get included multiple times, TRACE_HEADER_MULTI_READ at top */ > +#ifndef _PWR_EVENT_AVOID_DOUBLE_DEFINING > +#define _PWR_EVENT_AVOID_DOUBLE_DEFINING > + > +#define PWR_EVENT_EXIT -1 > + > +#endif > + > +DEFINE_EVENT(cpu, cpu_frequency, > + > + TP_PROTO(unsigned int frequency, unsigned int cpu_id), > + > + TP_ARGS(frequency, cpu_id) > +); > + > +TRACE_EVENT(machine_suspend, > + > + TP_PROTO(unsigned int state), > + > + TP_ARGS(state), > + > + TP_STRUCT__entry( > + __field( u32, state ) > + ), > + > + TP_fast_assign( > + __entry->state = state; > + ), > + > + TP_printk("state=%lu", (unsigned long)__entry->state) Same remark about the unsigned type for the state field.
> +); > + > +#ifdef CONFIG_EVENT_POWER_TRACING_DEPRECATED > + > #ifndef _TRACE_POWER_ENUM_ > #define _TRACE_POWER_ENUM_ > enum { > @@ -153,8 +214,32 @@ DEFINE_EVENT(power_domain, power_domain_target, > > TP_ARGS(name, state, cpu_id) > ); > - > +#endif /* CONFIG_EVENT_POWER_TRACING_DEPRECATED */ The clock and power_domain events have been recently introduced and so must be part of the new API. Can this #endif be moved right after the definition of power_end?
> #endif /* _TRACE_POWER_H */ Should this be at the very end of the file?
> > +/* Deprecated dummy functions must be protected against multi-declartion */ > +#ifndef EVENT_POWER_TRACING_DEPRECATED_PART_H > +#define EVENT_POWER_TRACING_DEPRECATED_PART_H > + > +#ifndef CONFIG_EVENT_POWER_TRACING_DEPRECATED > + > +#ifndef _TRACE_POWER_ENUM_ > +#define _TRACE_POWER_ENUM_ > +enum { > + POWER_NONE = 0, > + POWER_CSTATE = 1, > + POWER_PSTATE = 2, > +}; > +#endif > + > +static inline void trace_power_start(u64 type, u64 state, u64 cpuid) {}; > +static inline void trace_power_end(u64 cpuid) {}; > +static inline void trace_power_frequency(u64 type, u64 state, u64 cpuid) {}; > +#endif /* CONFIG_EVENT_POWER_TRACING_DEPRECATED */ > + > +#endif /* EVENT_POWER_TRACING_DEPRECATED_PART_H */ > + > + > + > /* This part must be outside protection */ > #include <trace/define_trace.h> > diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig > index e04b8bc..0be2e7f 100644 > --- a/kernel/trace/Kconfig > +++ b/kernel/trace/Kconfig > @@ -69,6 +69,21 @@ config EVENT_TRACING > select CONTEXT_SWITCH_TRACER > bool > > +config EVENT_POWER_TRACING_DEPRECATED > + depends on EVENT_TRACING > + bool A string is needed here. Without it it is impossible to have the option unset. This does the trick: +bool "Deprecated power event trace API, to be removed"
> + default y > + help > + Provides old power event types: > + C-state/idle accounting events: > + power:power_start > + power:power_end > + and old cpufreq accounting event: > + power:power_frequency > + This is for userspace compatibility > + and will vanish after 5 kernel iterations, > + namely 2.6.41. > + > config CONTEXT_SWITCH_TRACER > bool > ...
Thanks, Jean -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |