Re: [tip:tracing/core] tracing/bkl: Add bkl ftrace events

Re: [tip:tracing/core] tracing/bkl: Add bkl ftrace events

[Steven Rostedt]

On Fri, 2009-09-25 at 06:19 +0000, tip-bot for Frederic Weisbecker
wrote:
> Commit-ID:  96a2c464de07d7c72988db851c029b204fc59108

> +#if !defined(_TRACE_BKL_H) || defined(TRACE_HEADER_MULTI_READ)
> +#define _TRACE_BKL_H
> +
> +#include <linux/tracepoint.h>
> +
> +TRACE_EVENT(lock_kernel,
> +
> +	TP_PROTO(const char *func, const char *file, int line),
> +
> +	TP_ARGS(func, file, line),
> +
> +	TP_STRUCT__entry(
> +		__field(	int,		lock_depth		)

FYI,

By your request I added this as a common field ;-)

With commit:

 637e7e864103a7a68c1ce43ada27dfc25c0d113f

But since other utilities outside of ftrace may use this trace point, it
may be OK to keep the duplicate information in the kernel lock API trace
points.

-- Steve

> +		__field_ext(	const char *,	func, FILTER_PTR_STRING	)
> +		__field_ext(	const char *,	file, FILTER_PTR_STRING	)
> +		__field(	int,		line			)
> +	),
> +
> +	TP_fast_assign(
> +		/* We want to record the lock_depth after lock is acquired */
> +		__entry->lock_depth = current->lock_depth + 1;
> +		__entry->func = func;
> +		__entry->file = file;
> +		__entry->line = line;
> +	),
> +
> +	TP_printk("depth: %d, %s:%d %s()", __entry->lock_depth,
> +		  __entry->file, __entry->line, __entry->func)
> +);
> +

Re: [tip:tracing/core] tracing/bkl: Add bkl ftrace events

[Frederic Weisbecker]

On Sat, Sep 26, 2009 at 06:05:54AM -0400, Steven Rostedt wrote:
> On Fri, 2009-09-25 at 06:19 +0000, tip-bot for Frederic Weisbecker
> wrote:
> > Commit-ID:  96a2c464de07d7c72988db851c029b204fc59108
> 
> > +#if !defined(_TRACE_BKL_H) || defined(TRACE_HEADER_MULTI_READ)
> > +#define _TRACE_BKL_H
> > +
> > +#include <linux/tracepoint.h>
> > +
> > +TRACE_EVENT(lock_kernel,
> > +
> > +	TP_PROTO(const char *func, const char *file, int line),
> > +
> > +	TP_ARGS(func, file, line),
> > +
> > +	TP_STRUCT__entry(
> > +		__field(	int,		lock_depth		)
> 
> FYI,
> 
> By your request I added this as a common field ;-)
> 
> With commit:
> 
>  637e7e864103a7a68c1ce43ada27dfc25c0d113f
> 
> But since other utilities outside of ftrace may use this trace point, it
> may be OK to keep the duplicate information in the kernel lock API trace
> points.
> 
> -- Steve




Right, or may be we can define this tracepoint as a latency format
user by default. It's not possible yet for an event to do that but
I guess we can set a flag for this purpose or something...

Thanks.