Re: [PATCH v13 06/11] arm64: Use stack_trace_consume_fn and rename args to unwind()

[Miroslav Benes]

On Wed, 9 Mar 2022, Mark Brown wrote:

> On Tue, Mar 08, 2022 at 04:00:35PM -0600, Madhavan T. Venkataraman wrote:
> 
> > It is just that patch 11 that defines "select
> > HAVE_RELIABLE_STACKTRACE" did not receive any comments from you
> > (unless I missed a comment that came from you. That is entirely
> > possible. If I missed it, my bad). Since you suggested that change, I
> > just wanted to make sure that that patch looks OK to you.
> 
> I think that's more a question for the livepatch people to be honest -
> it's not entirely a technical one, there's a bunch of confidence level
> stuff going on.  For example there was some suggestion that people might
> insist on having objtool support, though there's also substantial
> pushback on making objtool a requirement for anything from other
> quarters.  I was hoping that posting that patch would provoke some
> discussion about what exactly is needed but that's not happened thus
> far.

I think everyone will be happy with HAVE_RELIABLE_STACKTRACE on arm64 as 
long as there is a guarantee that stack traces are really reliable. My 
understanding is that there is still some work to be done on arm64 arch 
side (but I may have misunderstood what Mark R. said elsewhere). And yes, 
then there is a question of objtool. It is one option but not the only 
one. There have been proposals of implementing guarantees on a compiler 
side and leaving objtool for x86_64 only (albeit objtool may bring more 
features to the table... ORC, arch features checking).

Madhavan also mentioned that he enhanced objtool and he planned to submit 
it eventually 
(https://lore.kernel.org/all/1a0e19db-a7f8-4c8e-0163-398fcd364d54@xxxxxxxxxxxxxxxxxxx/T/#u), 
so maybe arm64 maintainers could decide on a future direction based on 
that?

Regards
Miroslav