[PATCH 7/7] um: simplify IRQ handling code

[Johannes Berg]

On Wed, 2020-12-02 at 11:56 +0000, Anton Ivanov wrote:
> 
> > Also, we probably should disable SIGIO if the IRQ is freed, otherwise
> > the FD can keep interrupting us but we don't find anything in the epoll
> > set ... But again, a separate cleanup?
> 
> I think that deleting the fd from the set should stop that.

It doesn't. We'd have to separately call ignore_sigio_fd(), but we only
do that in deactivate_fd(), not in any of the other code paths ...

> Let me think on it. It was clearly being held together by duck tape
> and baling wire and it needs fixing. You got most of it which is
> grand, we just need to finish it.

Indeed.

Will send out this set again with the fix in a few minutes.

johannes