On Thu, Mar 18, 2021 at 03:26:13PM -0500, Madhavan T. Venkataraman wrote:
> On 3/18/21 10:09 AM, Mark Brown wrote:

> > If we are going to add the extra record there would probably be less
> > potential for confusion if we pointed it at some sensibly named dummy
> > function so anything or anyone that does see it on the stack doesn't get
> > confused by a NULL.

> I agree. I will think about this some more. If no other solution presents
> itself, I will add the dummy function.

After discussing this with Mark Rutland offlist he convinced me that so
long as we ensure the kernel doesn't print the NULL record we're
probably OK here, the effort setting the function pointer up correctly
in all circumstances (especially when we're not in the normal memory
map) is probably not worth it for the limited impact it's likely to have
to see the NULL pointer (probably mainly a person working with some
external debugger).  It should be noted in the changelog though, and/or
merged in with the relevant change to the unwinder.