On Wed, Nov 25, 2020 at 1:30 PM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> I'm not sure I'm willing to write and test the real patch, but it
> doesn't look _too_ nasty from just looking at the code. The bookmark
> thing makes it important to only actually clear the bit at the end (as
> does the handoff case anyway), but the way wake_up_page_bit() is
> written, that's actually very straightforward - just after the
> while-loop. That's when we've woken up everybody.
Actually, there's a problem. We don't know if we've done the hand-off
or not, so we don't know if we should clear the bit after waking
everybody up or not.
We set that WQ_FLAG_DONE bit for the hand-0off case, but only the
woken party sees that - the waker itself doesn't know about it (and we
have no good way to return it in that call chain: wake_up_page_bit ->
__wake_up_locked_key_bookmark -> __wake_up_common ->
wake_page_function().
We could easily hide the flag in the "bookmark" wait queue entry, but
that smells a bit hacky to me.
So I don't think it's worth it, unless somebody really wants to give it a try.
But if it turns out that the page ref change from Hugh causes some
unexpected problem, we do have this model as a backup.
Linus
