On Tue, 8 Jul 2003, Eric Varsanyi wrote:
> On Tue, Jul 08, 2003 at 08:42:29AM -0700, Davide Libenzi wrote:
> > It is not that events are delivered per-fd. If 3 and 4 refer to the same
> > file* and you register both 3 and 4 with EPOLLIN, you'll get two events if
> > an EPOLLIN happen. One for 3 and one for 4.
>
> Agreed 100%, this is roughly what would happen with select() as well which
> IMO is good (not surprising behaviour) for event loop writers: it would
> return with both bits set. The EEXIST we were getting before this patch
> would be analogous to select() returning an error if you set 2 bits that
> where for fd's sharing an object (even across read/write bit vectors).
>
> One could argue at the logic of having 2 fd's get read events on a
> shared underlying object, but one read and the other write certainly
> makes sense as discussed earlier.
I did not have the time to test the patch in your scenario, but if you can
confirm me it is working fine I'll push it.
- 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/
This archive was generated by hypermail 2b29 : Tue Jul 15 2003 - 22:00:28 EST