On Fri, Jul 24, 2020 at 8:24 AM Oleg Nesterov <oleg@redhat.com> wrote:
>
> not sure this makes any sense, but this looks like another user of
> trylock_page_bit_common(), see the patch below on top of 1/2.

Ok, that makes sense. Except you did it on top of the original patch
without the fix to set WQ_FLAG_WOKEN for the non-wakeup case.

And in fact, once you do it on top of that, it becomes obvious that we
can share even more code: move the WQ_FLAG_WOKEN logic _into_ the
trylock_page_bit_common() function.

Then the whole thing becomes something like the attached.

I added your reviewed-by, but maybe you should double-check my changes.

                Linus