Re: [PATCH 13/14] x86/unwind/orc: Add more unwinder warnings

[Josh Poimboeuf]

On Thu, Mar 12, 2020 at 08:12:16PM +0100, Jann Horn wrote:
> On Thu, Mar 12, 2020 at 6:31 PM Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote:
> > Make sure warnings are displayed for all error scenarios (except when
> > encountering an empty unwind hint).
> [...]
> >         /* End-of-stack check for kernel threads: */
> >         if (orc->sp_reg == ORC_REG_UNDEFINED) {
> > -               if (!orc->end)
> > +               if (!orc->end) {
> > +                       /*
> > +                        * This is reported as an error for the caller, but
> > +                        * otherwise it isn't worth warning about.  In theory
> > +                        * it can only happen when hitting UNWIND_HINT_EMPTY in
> > +                        * entry code, close to a kernel exit point.
> > +                        */
> >                         goto err;
> 
> But UNWIND_HINT_EMPTY sets end=1, right? And this is the branch for
> end==0. What am I missing?

You're right.  I need to revisit that comment...

-- 
Josh