Messages in this thread | | | Date | Wed, 20 May 2015 21:10:16 +0200 (CEST) | From | Jiri Kosina <> | Subject | Re: [PATCH v4 0/3] Compile-time stack frame pointer validation |
| |
On Wed, 20 May 2015, Peter Zijlstra wrote:
> > I think it would be nice to have full DWARF unwind support for > > everything at some point. Unfortunately, I don't see any easy path to > > getting there. It doesn't help that AFAIK no one has ever proposed a > > usable in-kernel DWARF unwinder. > > There's a bit of history here; SuSE (iirc) actually has one, however: > > https://lkml.org/lkml/2012/2/10/356
Oh absolutely, there are stories behind this :)
Just for the sake of completness -- the current implementation can be found in our public GIT repository, for not-really-complete picture see [1] [2] [3] [4].
It turned out to be rather useful on many ocasions when debugging customer reports, but I of course also understand what Linus is saying above. The bugs in unwinder can be *really* painful. Our experience so far has been that it did pay off at the end of the day (and of course analyzing stacktraces is our daily bread).
[1] http://kernel.suse.com/cgit/kernel-source/tree/patches.suse/stack-unwind?h=SLE12 [2] http://kernel.suse.com/cgit/kernel-source/tree/patches.suse/no-frame-pointer-select?h=SLE12 [3] http://kernel.suse.com/cgit/kernel-source/tree/patches.arch/stack-unwind-cfi_ignore-takes-more-arguments?h=SLE12 [4] http://kernel.suse.com/cgit/kernel-source/tree/patches.arch/x86_64-unwind-annotations?h=SLE12
-- Jiri Kosina SUSE Labs
| |