Re: [RFC PATCH] ftrace trace event: introduce assignment macros

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

* Frederic Weisbecker (fweisbec@gmail.com) wrote:
> On Thu, Dec 09, 2010 at 11:22:11AM -0500, Mathieu Desnoyers wrote:
> > This patch proposes encapsulation of the raw assignments within TP_fast_assign()
> > by introducing tp_assign() and tp_memcpy() macros. This will allow us to:
> > 
> > - generically filter from input fields,
> > - redefine the field write primitives.
> > 
> > The current macros map directly to the old code. I changed the documentation in
> > tracepoint.txt to reflect these new primitives, but all TRACE_EVENT() users
> > should gradually update their code to use these macro wrappers rather than raw
> > "=" assignments or mempcy() calls.
> > 
[...]
> > ---
> >  include/linux/tracepoint.h |   12 ++++++------
> >  include/trace/ftrace.h     |    6 ++++++
> >  2 files changed, 12 insertions(+), 6 deletions(-)
> > 
> > Index: linux-2.6-lttng.git/include/trace/ftrace.h
> > ===================================================================
> > --- linux-2.6-lttng.git.orig/include/trace/ftrace.h
> > +++ linux-2.6-lttng.git/include/trace/ftrace.h
> > @@ -1,4 +1,10 @@
> >  /*
> > + * Macros mapping tp_assign() and tp_memcpy() to "=" and memcpy.
> > + */
> > +#define tp_assign(dest, src)		__entry->dest = src
> > +#define tp_memcpy(dest, src, len)	memcpy(__entry->dest, src, len)
> 
> Looks good.

Great! Can I resend it without "RFC" with your Acked-by ?

If there is wide-spread agreement on this change, I might soon start rolling out
patches that migrate TRACE_EVENT users to these new primitives.

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com