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

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

On Mon, Aug 26, 2024 at 12:36:45PM -0400, Will Hawkins wrote:
> On Mon, Aug 26, 2024 at 10:58 AM Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> > 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

oops, "extern"

> > such as C++, so I think this patch needs reworking, there is ambiguity
> > on the "linkage" term, we should use some other term :-\
> 
> Completely understand! I will rework it and propose something different.
> 
> I am just thankful that you are so welcoming of the contribution! Your
> enthusiasm has made a contributor for life!

Another recommendation, please don't point the field to strings like
"extern" or "static", as DWARF is not language specific, pahole manages
to handle C mostly, C++ to a great degree but I also used it with Go,
Rust.

We want to make it more language agnostic, so while the "linkage" as a
mangled name is not language specific, "static" or "extern" may be.

So just keep it using the DWARF constant/enumeration, so that at
dwarves_fprintf.c time we can, based on the language encoded in the
debugging info (for BTF we assume C until we introduce some way to
encode Rust (The kernel can be written using Rust, so we're bound to
have something in that direction introduced in BTF at some point)).

- Arnaldo