On Wed, 17 Dec 2008 19:00:23 +0100, "Sam Ravnborg" <sam@xxxxxxxxxxxx>
said:
> On Wed, Dec 17, 2008 at 08:38:24PM +0300, Cyrill Gorcunov wrote:
> > [Sam Ravnborg - Wed, Dec 17, 2008 at 06:26:40PM +0100]
> > | On Wed, Dec 17, 2008 at 10:17:54AM +0100, Alexander van Heukelum wrote:
> > | > Introduce the PROC macro in the generic header file
> > | > include/linux/linkage.h to annotate functions in assembly
> > | > files. This is a first step to fully annotate functions
> > | > (procedures) in .S-files. The PROC macro complements the
> > | > already existing and being used ENDPROC macro. The generic
> > | > implementation of PROC is exactly the same as ENTRY.
> > | >
> > | > The goal is to annotate functions, at least those called
> > | > from C code, with PROC at the beginning and ENDPROC at the
> > | > end. This is for the benefit of debugging and tracing. It
> > | > will also allow to introduce a framework to check for
> > | > nesting problems and missing annotations in a later stage
> > | > by overriding ENTRY/END and PROC/ENDPROC in architecture-
> > | > specific code, after the annotation errors have been fixed.
> > | >
> > | > Signed-off-by: Alexander van Heukelum <heukelum@xxxxxxxxxxx>
> > | > Cc: Sam Ravnborg <sam@xxxxxxxxxxxx>
> > | > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> > |
> > | I understand where you are coming from with these.
> > | But what I see now is:
> > |
> > | ENTRY/END
> > | PROC/ENDPROC
> > | KPROBE_ENTRY/KPROBE_END
> > |
> > | And it is not obvious for me reading the comment when I should
> > | expect which one to be used.
> > |
> > | Could we try to keep it down to two variants?
> > | And then document when to use which one.
> > |
> > | Sam
> > |
> >
> > Sam, I think eventually we should get something like this:
> >
> > - KPROBE will be eliminated and explicit section descriptions
> > are to be used
> > - ENTRY could be used / or renamed for something more descriptive
> > and being used aligned jmp targets or in case of procs with
> > shared body
I don't think ENTRY should be used for nested procedures. If the
author wants to do something like that, he better knew something
about the assembler anyhow.
> > - PROC/ENDPROC are to replace old ENTRY/END for procs being called
> > mostly from C code
Currently there is many different patterns. Some functions use ENTRY
without END, some use ENTRY/ENDPROC, some use ENDPROC without annotation
at the start...
> So what prevents us from extending ENTRY/END instead of introducing
> another set?
ENTRY/END alone is not enough if one wants to be able to distinguish
between code (functions) and non-executed data.
> Let us try to extend what we have and not introduce something new.
Agreed. I vote to complement the existing ENDPROC annotation with
the proposed PROC annotation. Let's call that an extension, not
something new ;). As it stands it is not impossible to go with
ENTRY/ENDPROC for code and ENTRY/END for data. However, ENTRY
implies alignment and the prefered alignment for code and data
might differ.
Greetings,
Alexander
> Sam
--
Alexander van Heukelum
heukelum@xxxxxxxxxxx
--
http://www.fastmail.fm - Access your email from home and the web
--
To unsubscribe from this list: send the line "unsubscribe linux-arch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
