Re: [PATCH v2] power: new trace event to print device suspend and resume time

[Steven Rostedt <rostedt@goodmis.org>]

On Wed, 2013-07-17 at 09:57 -0600, Shuah Khan wrote:

> +TRACE_EVENT(device_pm_report_time,
> +
> +	TP_PROTO(const char *device, const char *driver, const char *parent,
> +		 const char *pm_ops, ktime_t ops_time, char *pm_event_str,
> +		 int error),
> +
> +	TP_ARGS(device, driver, parent, pm_ops, ops_time, pm_event_str, error),
> +
> +	TP_STRUCT__entry(
> +		__string(device, device)
> +		__string(driver, driver)
> +		__string(parent, parent)
> +		__string(pm_ops, pm_ops)
> +		__string(pm_event_str, pm_event_str)
> +		__field(s64, ops_time)
> +		__field(int, error)
> +	),
> +
> +	TP_fast_assign(
> +		__assign_str(device, device);
> +		__assign_str(driver, driver);
> +		__assign_str(parent, parent);
> +		__assign_str(pm_ops, pm_ops);
> +		__assign_str(pm_event_str, pm_event_str);
> +		__entry->ops_time = ktime_to_ns(ops_time) >> 10;

You may want to just record the ops_time, and do the conversion in the
TP_printk(). This is the fast path, and any calculations done here just
increases the overhead of the trace point.

-- Steve

> +		__entry->error = error;
> +	),
> +
> +	TP_printk("%s %s parent=%s state=%s ops=%s usecs=%lld err=%d\n",
> +		__get_str(driver), __get_str(device), __get_str(parent),
> +		__get_str(pm_event_str), __get_str(pm_ops),
> +		__entry->ops_time, __entry->error)
> +);
> +
>  DECLARE_EVENT_CLASS(wakeup_source,
>  
>  	TP_PROTO(const char *name, unsigned int state),