From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753955AbYGGNnR (ORCPT ); Mon, 7 Jul 2008 09:43:17 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752554AbYGGNnH (ORCPT ); Mon, 7 Jul 2008 09:43:07 -0400 Received: from tomts20-srv.bellnexxia.net ([209.226.175.74]:35637 "EHLO tomts20-srv.bellnexxia.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751857AbYGGNnG (ORCPT ); Mon, 7 Jul 2008 09:43:06 -0400 X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: AvsEAAy2cUhMQWVt/2dsb2JhbACBXKx3 Date: Mon, 7 Jul 2008 09:43:02 -0400 From: Mathieu Desnoyers To: Eduard - Gabriel Munteanu Cc: akpm@linux-foundation.org, Ingo Molnar , linux-kernel@vger.kernel.org, Peter Zijlstra , "Frank Ch. Eigler" , Steven Rostedt Subject: Re: [RFC patch 00/12] Tracepoints v2 Message-ID: <20080707134302.GB14704@Krystal> References: <20080704235207.147809973@polymtl.ca> <20080706022711.199401fe@linux360.ro> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline In-Reply-To: <20080706022711.199401fe@linux360.ro> X-Editor: vi X-Info: http://krystal.dyndns.org:8080 X-Operating-System: Linux/2.6.21.3-grsec (i686) X-Uptime: 09:34:04 up 32 days, 18:15, 6 users, load average: 0.82, 0.91, 0.87 User-Agent: Mutt/1.5.16 (2007-06-11) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Eduard - Gabriel Munteanu (eduard.munteanu@linux360.ro) wrote: > On Fri, 04 Jul 2008 19:52:07 -0400 > Mathieu Desnoyers wrote: > > > Hi, > > > > Here is the second release of kernel tracepoints, including the > > architecture independent instrumentation taken from LTTng. I submit > > this for another round of comments. > > Hi, > > Not using format specifiers sounds pretty nice. Are there any other > advantages over markers? What about self-modifying code, does it employ > this technique to eliminate conditional branching (markers did this on > some arches, IIRC)? > The main advantage is to remove the format strings. It also permits to declare the tracepoints in a header file instead of only spreading them all over the kernel code. It aims at providing an internal kernel interface to kernel tracers, compared to markers which are eventually meant to export the trace information, along with the format specifiers, to userspace. The version I sent here does not use immediate values (self-modifying code), but the one I have in my patchset does. > I'm considering using this for kmemtrace, so please keep me Cc-ed if > possible. > Sure, I will, Mathieu > > Cheers, > Eduard > > P.S.: BTW, the To header is either broken or I screwed it (I'm reading > this via Gmane's NNTP): > "Steven Rostedt