> > Anything that "exposes" a file* interface that support f_op->poll() is
> > usable with epoll. File rocks !! :)
>
> I agree, fds are pretty good, and it's nice that they work equally
> well with poll/select/sigio as with epoll.
>
> It's just that to write an ideal, general event loop, pollable fds
> need to be created on the fly for futexes, signals, timers and aio
> requests at least. Currently this is already done for futexes. In
> addition, some kinds of event result need to return a few words of
> data with each event (for example, SIGCHLD events).
>
> Perhaps it's not a bad idea, but I do wonder whether fds created on
> the fly for every requested event, and events than can only report
> "readable" not anything richer than that are a good solution.
IMHO it is very elegant to have these objects to be a file* internally.
Common event retrival and automatic cleanup are just the first two points
that come in my mind.
- Davide
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/