Re: [PATCH] Linux Kernel Markers 0.5 for Linux 2.6.17 (with probe management)

From: Frank Ch. Eigler
Date: Mon Sep 25 2006 - 13:48:46 EST


Hi -

On Thu, Sep 21, 2006 at 08:50:29PM +0200, Ingo Molnar wrote:

> [...] let me qualify that: parameters must be prepared there too -
> but no actual function call inserted. (at most a NOP
> inserted). [...] Does a simple asm() that takes read-only
> parameters but only adds a NOP achieve this result?

You mean something like this?

#define MARK(n,v1,v2,v3) asm ("__mark_" #n ": nop" :: \
"X" (v1), "X" (v2), "X" (v3))

I haven't been able to get gcc to emit any better debuginfo for
parameters pseudo-passed like this.

(I've tested such a marker inserted into an inner loop of dhrystone.
It was compiled with "-ggdb -O3". Neither gdb nor systemtap could
resolve the same values/symbols being passed as MARK() arguments,
though at least the breakpoint address was nicely marked.)

- FChE

Attachment: pgp00000.pgp
Description: PGP signature