Re: [PATCH 1/1] eventfd: implementation of EFD_MASK flag

[Eric Wong <normalperson@yhbt.net>]

Andy Lutomirski <luto@amacapital.net> wrote:
> On Thu, Feb 7, 2013 at 12:11 PM, Martin Sustrik <sustrik@250bpm.com> wrote:
> > On 07/02/13 20:12, Andy Lutomirski wrote:
> >> On 02/06/2013 10:41 PM, Martin Sustrik wrote:
> >>> The value of 'events' should be any combination of event flags as defined
> >>> by
> >>> poll(2) function (POLLIN, POLLOUT, POLLERR, POLLHUP etc.) Specified
> >>> events will
> >>> be signaled when polling (select, poll, epoll) on the eventfd is done
> >>> later on.
> >>> 'ptr' is an opaque pointer that is not interpreted by eventfd object.
> >>
> >> How does this interact with EPOLLET?
> >
> > That's an interesting question. The original eventfd code doesn't do
> > anything specific to either edge or level mode. Neither does my patch.
> >
> > Inspection of the code seems to suggest that edge vs. level distinction is
> > handled elsewhere (ep_send_events_proc) where there is a separate list of
> > ready events and the function, after returning the event, decides whether to
> > leave the event in the list (level) or delete it from the list (edge).

Right, the edge vs. level distinction is internal to epoll.

> Hmm.  Having looked at the eventpoll.c source again, I remain
> unconvinced that EPOLLET works the way that any userspace developer
> would expect it to.

As as userspace developer, EPOLLET seems to work as expected/documented;
but I realized EPOLLONESHOT is what I want to be using instead.

> > In any case, review from someone with experience with epoll implementation
> > would help.

I'm no expert, but I don't think eventfd (or any file type) needs to
care about what I/O notification scheme/options it's used with.