Message-ID: <CAP-5=fVcF+F7ST3Ya0_3hXq69ArhZv0gy30U4SPC7Cqih7HAWA@mail.gmail.com>
Date: Wed, 29 Jan 2025 11:36:02 -0800
From: Ian Rogers <irogers@...gle.com>
To: Daniel Xu <dxu@...uu.xyz>
Cc: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>, 
	Arnaldo Carvalho de Melo <acme@...nel.org>, Namhyung Kim <namhyung@...nel.org>, 
	Mark Rutland <mark.rutland@....com>, 
	Alexander Shishkin <alexander.shishkin@...ux.intel.com>, Jiri Olsa <jolsa@...nel.org>, 
	Adrian Hunter <adrian.hunter@...el.com>, Kan Liang <kan.liang@...ux.intel.com>, 
	Nathan Chancellor <nathan@...nel.org>, Nick Desaulniers <ndesaulniers@...gle.com>, 
	Bill Wendling <morbo@...gle.com>, Justin Stitt <justinstitt@...gle.com>, 
	Aditya Gupta <adityag@...ux.ibm.com>, "Steinar H. Gunderson" <sesse@...gle.com>, 
	Charlie Jenkins <charlie@...osinc.com>, Changbin Du <changbin.du@...wei.com>, 
	"Masami Hiramatsu (Google)" <mhiramat@...nel.org>, James Clark <james.clark@...aro.org>, 
	Kajol Jain <kjain@...ux.ibm.com>, Athira Rajeev <atrajeev@...ux.vnet.ibm.com>, 
	Li Huafei <lihuafei1@...wei.com>, Dmitry Vyukov <dvyukov@...gle.com>, 
	Andi Kleen <ak@...ux.intel.com>, Chaitanya S Prakash <chaitanyas.prakash@....com>, 
	linux-kernel@...r.kernel.org, linux-perf-users@...r.kernel.org, 
	llvm@...ts.linux.dev, Song Liu <song@...nel.org>, bpf@...r.kernel.org
Subject: Re: [PATCH v3 13/18] perf build: Remove libbfd support

On Wed, Jan 29, 2025 at 8:47 AM Daniel Xu <dxu@...uu.xyz> wrote:
>
> On Tue, Jan 28, 2025 at 05:40:44PM -0800, Ian Rogers wrote:
> > On Tue, Jan 28, 2025 at 4:31 PM Daniel Xu <dxu@...uu.xyz> wrote:
> > >
> > > Hi Ian,
> > >
> > > On Wed, Jan 22, 2025 at 09:43:03AM -0800, Ian Rogers wrote:
> > > > libbfd is license incompatible with perf and building requires the
> > > > BUILD_NONDISTRO=1 build flag. Remove the code to simplify the code
> > > > base.
> > > >
> > > > Signed-off-by: Ian Rogers <irogers@...gle.com>
> > > > ---
> > > >  tools/perf/Documentation/perf-check.txt |   1 -
> > > >  tools/perf/Makefile.config              |  38 +---
> > > >  tools/perf/builtin-check.c              |   1 -
> > > >  tools/perf/tests/Build                  |   1 -
> > > >  tools/perf/tests/builtin-test.c         |   1 -
> > > >  tools/perf/tests/pe-file-parsing.c      | 101 ----------
> > > >  tools/perf/tests/tests.h                |   1 -
> > > >  tools/perf/util/demangle-cxx.cpp        |  13 +-
> > > >  tools/perf/util/disasm_bpf.c            | 166 ----------------
> > > >  tools/perf/util/srcline.c               | 243 +-----------------------
> > > >  tools/perf/util/symbol-elf.c            |  86 +--------
> > > >  tools/perf/util/symbol.c                | 135 -------------
> > > >  tools/perf/util/symbol.h                |   4 -
> > > >  13 files changed, 7 insertions(+), 784 deletions(-)
> > > >  delete mode 100644 tools/perf/tests/pe-file-parsing.c
> > >
> > > [..]
> > >
> > > I was briefly investigating why the centos build of perf was not
> > > demangling rust v0 symbols [0]. From looking at the rust issue [1], it
> > > appears the rust team somehow delivered support for v0 demangling
> > > through libbfd. The code itself looked a bit odd (relying on cxx
> > > demangle to run first?), but that's a separate thing.
> >
> > There is still C++ demangling support by way of cxxabi:
> > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/demangle-cxx.cpp?h=perf-tools-next#n44
> > that was in libstdc++ (GNU) and libcxx (LLVM) when I looked.
> >
> > > The centos build does not build with libbfd for the license issues you
> > > mentioned. So your change probably won't regress any distro use cases.
> > > But it does remove support for motivated users who don't have
> > > re-distribution requirements.
> > >
> > > But since this patchset came up first in my search, I thought it'd be
> > > good to mention that someone probably needs to add v0 support to
> > > tools/perf/util/demangle-rust.c.
> >
> > So I don't see any libbfd dependencies in demangle-rust.c:
> > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/util/demangle-rust.c?h=perf-tools-next#n8
> > Unusually we don't have any tests on the Rust demangling, we do for
> > Java and OCaml:
> > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/tests/demangle-java-test.c?h=perf-tools-next
> > https://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/tree/tools/perf/tests/demangle-ocaml-test.c?h=perf-tools-next
> >
> > Reading a bit more it seems that previous libiberty was coming to the
> > rescue by way of C++ demangling. I'll see if I can write a demangler
> > by way of lex and yacc.
>
> Cool :)

Not by way of lex and yacc, as it seemed overkill, but I sent out:
https://lore.kernel.org/lkml/20250129193037.573431-1-irogers@google.com/
I only tested with the examples from the doc. If you could take a look.

> > If we have a v0 standard one is there any
> > value in the existing demangler or legacy demangling? It seems this
> > has been broken for the best part of 5 years.
>
> I believe the "legacy" symbol format is still the rust default. So
> probably can't remove that. Looks like there's some desire to change
> that, probably probably not very soon [0].
>
> That probably also explains why nobody reported the breakage - only very
> cool kids are using v0 scheme currently.

Ok. I wasn't sure on the status so I've tried to incorporate the
previous legacy support and the v0 support in my patch.

Thanks,
Ian

> Thanks,
> Daniel
>
>
> [0]: https://github.com/rust-lang/rust/pull/89917

Powered by blists - more mailing lists