Re: [PATCH 0/1] btf_loader support for subprogram linkage

[Arnaldo Carvalho de Melo <acme@kernel.org>]

On Mon, Aug 26, 2024 at 11:53:56AM -0300, Arnaldo Carvalho de Melo wrote:
> On Mon, Aug 26, 2024 at 11:52:13AM -0300, Arnaldo Carvalho de Melo wrote:
> > On Fri, Aug 23, 2024 at 05:28:30PM -0400, Will Hawkins wrote:
> > > First, 
> > > 
> > > I am a huge fan of pahole (and friends). I volunteered to work on
> > > helping to edit the BTF spec for the IETF and so I've started to look
> > > more deeply at BTF and the tools. 
> > > 
> > > Second, I hope that what I am offering is being sent to the right place
> > > and is in the right format. I tried to follow what seems to be the
> > > "right thing" by looking at mailing list archives.
> > > 
> > > This patch add supports to the btf_loader for subprogram (BTF_KIND_FUNC)
> > > linkages. For example,
> > > 
> > > ```
> > > $ cat a.c
> > > static int x() {
> > >   return 5;
> > > }
> > > $ gcc -gbtf -g -O0 -c a.c 
> > > $ ~/code/pahole/build/pfunct -Fbtf --compile a.o
> > > int x(void) /* linkage=static */
> > > {
> > > 	return 0;
> > > }
> > > ```
> > 
> > So, I'm changing this to become:
> > 
> >   $ ~/code/pahole/build/pfunct -Fbtf --compile a.o
> >   static int x(void)
> >   {
> >   	return 0;
> >   }
> > 
> > As --compile is supposed to generate compileable code that is as much as
> > possible from the type information similar to the original code, ok?
> 
> I see, your patch isn't the one adding that linkage comment, you're just
> collecting it from BTF, I'll then add a followup patch to get to the
> output I described.

So I'll looked up what that linkage field comes from and:

commit ce516fb0cf2a3d9fc45c2a9c7ab2cf4bd24965a0
Author: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Date:   Sun Jul 8 19:47:26 2007 -0300

    [LIB]: Support DW_AT_MIPS_linkage_name
    
    Another C++ specific case:
    
    -       class TypeTemplate ByName(const string  &, size_t);
    +       class TypeTemplate ByName(const string  &, size_t); /* linkage=_ZN4ROOT6Reflex12TypeTemplate6ByNameERKSsj */
    
    Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>

So its not really "static" or "stern", but the mangled name in languages
such as C++, so I think this patch needs reworking, there is ambiguity
on the "linkage" term, we should use some other term :-\

- Arnaldo