[PATCH] um: fix stub location calculation

[Johannes Berg]

On Tue, 2021-07-13 at 17:44 -0500, YiFei Zhu wrote:
> 
> The store is optimized away -> the faultinfo is unmodified -> the segv
> handler treats the fault as unfixable -> init dead.

Yeah. I was confused about this in the clone case, but even there of
course the store is important because that's the child's PID in the
parent ... I somehow looked at the code last night and thought it was
only a broken error case.

I'll leave it to Richard to figure out which of the two things to apply,
I guess. And it should probably come with a Cc: stable, which I forgot
on my patch.

johannes