Re: [rfc] new poll callback'd wake up hell ...

[Davide Libenzi <davidel@xmailserver.org>]

On Wed, 20 Nov 2002, Mark Mielke wrote:

> > 2) Use this infrastructure to perform safe poll wakeups
> > ...
> > static void ep_poll_safe_wakeup(struct poll_safewake *psw, wait_queue_head_t *wq)
> > {
> >         atomic_inc(&psw->count);
> >         do {
> >                 if (!xchg(&psw->wakedoor, 0))
> >                         break;
> >                 wake_up(wq);
> >                 xchg(&psw->wakedoor, 1);
> >         } while (!atomic_dec_and_test(&psw->count));
> > }
> > Does anyone foresee problem in this implementation ?
> > Another ( crappy ) solution might be to avoid the epoll fd to drop inside
> > its poll wait queue head, wait queues that has the function pointer != NULL
>
> Clever. (I think the second xchg() can just be atomic_set()) Without actually
> playing with it, it looks good to me.
>
> If the problem is too hard to solve - it isn't that bad if one can't
> epoll recursively. If the functionality was added later, it is
> doubtful that the API itself would need to change.

This should cover the wake_up() cycle case IMHO. We should be able to
stock epoll fds inside epoll fd. In theory ... :)




- Davide