We're using a very old epoll patch (ep_patch-2.4.18-0.32), and have noticed that if there is a dropped TCP packet followed by a FIN, epoll says the socket is ready for POLLHUP *before* all the data has been read from the socket. As a workaround, we are ignoring POLLHUP and simply reading until we get EOF; this solves the problem nicely.
Hi Dan. The POLLHUP is a condition that does not preclude the socket to have data to be still read. The correct approach, with both and new epoll, is to register the event and read until data is available (if you need to). Then, if a POLLHUP has been received, do not go to wait for POLLINs.
Sounds like I need to go read Stevens again.
For what it's worth, SuSv3 says in the poll() man page:
POLLHUP
The device has been disconnected. This event and POLLOUT are
mutually-exclusive; a stream can never be writable if a hangup has
occurred. However, this event and POLLIN, POLLRDNORM, POLLRDBAND,
or POLLPRI are not mutually-exclusive.which doesn't quite say that POLLHUP may be signalled when there's still bunches of data to read, but it comes close.
Thanks, Dan
-- Dan Kegel http://www.kegel.com http://counter.li.org/cgi-bin/runscript/display-person.cgi?user=78045
- : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
