This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: [PATCH] Linux Kernel Markers 0.8 for 2.6.17 (with near jump for i386)
- From: Mathieu Desnoyers <compudj at krystal dot dyndns dot org>
- To: Karim Yaghmour <karim at opersys dot com>
- Cc: Martin Bligh <mbligh at google dot com>, "Frank Ch. Eigler" <fche at redhat dot com>, Masami Hiramatsu <masami dot hiramatsu dot pt at hitachi dot com>, prasanna at in dot ibm dot com, Andrew Morton <akpm at osdl dot org>, Ingo Molnar <mingo at elte dot hu>, Paul Mundt <lethal at linux-sh dot org>, linux-kernel <linux-kernel at vger dot kernel dot org>, Jes Sorensen <jes at sgi dot com>, Tom Zanussi <zanussi at us dot ibm dot com>, Richard J Moore <richardj_moore at uk dot ibm dot com>, Michel Dagenais <michel dot dagenais at polymtl dot ca>, Christoph Hellwig <hch at infradead dot org>, Greg Kroah-Hartman <gregkh at suse dot de>, Thomas Gleixner <tglx at linutronix dot de>, William Cohen <wcohen at redhat dot com>, ltt-dev at shafik dot org, systemtap at sources dot redhat dot com, Alan Cox <alan at lxorguk dot ukuu dot org dot uk>, Jeremy Fitzhardinge <jeremy at goop dot org>
- Date: Mon, 25 Sep 2006 01:39:40 -0400
- Subject: Re: [PATCH] Linux Kernel Markers 0.8 for 2.6.17 (with near jump for i386)
- References: <20060922214239.GA28625@Krystal> <451591AC.9030703@opersys.com>
* Karim Yaghmour (karim@opersys.com) wrote:
> diffing against an LTT tree ...
>
Ok, I will do the diff against a vanilla kernel, good point.
> > --- /dev/null
> > +++ b/include/asm-alpha/marker.h
> ...
> > --- /dev/null
> > +++ b/include/asm-arm/marker.h
> ...
> > --- /dev/null
> > +++ b/include/asm-arm26/marker.h
> ...
> ...
>
> Not sure about the need for asm-foo/marker.h if the file contains no
> code at all. If there's going to be one marker.h per arch, it might
> as well have a purpose. So instead of:
>
> > --- /dev/null
> > +++ b/include/asm-i386/marker.h
> ...
> > +#define ARCH_HAS_MARK_NEAR_JUMP
>
> and
>
> > --- /dev/null
> > +++ b/include/linux/marker.h
> ...
> > +#ifndef ARCH_HAS_MARK_NEAR_JUMP
> ...
>
> Why not just have asm-foo/marker.h either implement the optimization
> or point to an asm-generic/marker.h which contains the non-optimized
> code. No #ifndefs needed.
>
Ok, good idea.
> > --- /dev/null
> > +++ b/kernel/Kconfig.marker
> ...
> > +config MARK_SYMBOL
> ...
> > +config MARK_JUMP_CALL
> ...
> > +config MARK_JUMP
> ...
>
> My understanding of Ingo's input is that he'd rather not have this
> multiple options. Either the markers are active or they aren't.
> So ...
> MARK_ACTIVE ... speaks for itself, enables both the markers and
> the set/disable infrastructure. Markers are enabled in their
> optimized per-architecture implementation.
>
> MARK_FORCE_DIRECT_CALL ... forces all markers to be non-optimized
> (good for embedded systems where the image is in rom/flash and
> can therefore not have runtime binary modifications.) Maybe this
> should depend on CONFIG_EMBEDDED.
>
>
Ok, now only :
CONFIG_MARKERS
and
CONFIG_MARKERS_FORCE_DIRECT_CALL (only appears is EMBEDDED is enabled)
I have rewritted almost everything else, I will post a patch soon (currently
cleaning up). It now supports inline function, unrolled loops and multiple
definitions of the same marker.
Supporting inline functions has proven more important than I first thought
because of gcc 4.0+ optimisations where it takes a static-only function and
inline it automatically in the body of the caller. It removes the symbols of the
inlined function when it does that.
My new approach is a separate .markers section, everything is there.
Mathieu
OpenPGP public key: http://krystal.dyndns.org:8080/key/compudj.gpg
Key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68