Re: [RFC 0/5] kernel: backtrace unwind support

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Fri, 2012-02-10 at 12:25 +0100, Jiri Olsa wrote:
> I was recently dealing with libunwind and wanted to try out
> the dwarf backtrace unwind in kernel space.
> 
> The attached patchset implements dwarf backtrace unwind
> based on the exception header frames (.eh_frame_hdr and 
> .eh_frame ELF sections). The code is mostly stolen from
> libunwind (git://git.sv.gnu.org/libunwind.git).
> 
> I'm not sure how much of usage this can be given that we
> already have quite reliable stack backtrace, and given
> the complexity of the dwarf unwind. But I might be
> overlooking something and this could be of use for someone
> else.
> 
> Also it needs to be said, that the state of this patchset
> is far from being done. It's in state 'working for me' on
> x86_64 and seems to provide reliable backtrace.
> 
> attached patches:
>  - 1/5 unwind, kconfig: Adding UNWIND* options
>  - 2/5 unwind, x86: Generate exception frames data for UNWIND_EH_FRAME option
>  - 3/5 unwind, dwarf: Add dwarf unwind support
>  - 4/5 unwind, api: Add unwind interface and implementation for x86_64
>  - 5/5 unwind, test: Add backtrace unwind test code 

Right, so last time someone did a x86 dwarf unwinder there was a bit of
a 'discussion' and Linus basically told people to go away. That said,
there are a number of literate dwarfs in various architectures because
that simply the only way to get a backtrace on them.

So I CC'ed Linus who has a strong here, jejb since he's the one that
told me several time there's a number of literate dwarfs already in the
kernel and Jan because I think it was him that tried last on x86.