Re: [PATCH v7 03/14] perf tools: Introducing struct unwind_libunwind_ops for local unwind

From: Jiri Olsa
Date: Wed Jun 01 2016 - 04:40:02 EST


On Tue, May 31, 2016 at 11:19:01AM +0000, He Kuang wrote:
> Currently, libunwind operations are fixed, and they are chosen
> according to the host architecture. This will lead a problem that if a
> thread is run as x86_32 on x86_64 machine, perf will use libunwind
> methods for x86_64 to parse the callchain and get wrong result.
>
> This patch changes the fixed methods of libunwind operations to
> thread/map related, and each thread can have indivadual libunwind
> operations. Local libunwind methods are registered as default value.
>
> Signed-off-by: He Kuang <hekuang@xxxxxxxxxx>
> ---
> tools/perf/util/thread.c | 6 ++++
> tools/perf/util/thread.h | 4 ++-
> tools/perf/util/unwind-libunwind.c | 58 +++++++++++++++++++++++++++++++++++---
> tools/perf/util/unwind.h | 19 +++++++++++++
> 4 files changed, 82 insertions(+), 5 deletions(-)
>
> diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
> index 45fcb71..95ff1b8 100644
> --- a/tools/perf/util/thread.c
> +++ b/tools/perf/util/thread.c
> @@ -43,6 +43,12 @@ struct thread *thread__new(pid_t pid, pid_t tid)
> thread->cpu = -1;
> INIT_LIST_HEAD(&thread->comm_list);
>
> +#ifdef HAVE_LIBUNWIND_SUPPORT
> + unwind__register_ops(thread, local_unwind_libunwind_ops);
> +#else
> + unwind__register_ops(thread, NULL);
> +#endif
> +

I think we should put unwind__register_ops call into unwind__prepare_access
in this patch (like you do in patch 4) and without the #else clause

this way you can keep unwind__register_ops local

thanks,
jirka