Re: L2CAP non-blocking socket nasty race conditions

[Jean Tourrilhes <jt@bougret.hpl.hp.com>]

On Fri, Feb 06, 2004 at 12:37:24AM +0100, Marcel Holtmann wrote:
> Hi Jean,
> 
> > 	There is one thing I don't like about your patch : you disable
> > error checking for listening socket. I don't think it's right,
> > especially for checking socket shutdown.
> > 	Let's take the following scenario. Multi-threaded program, one
> > listening l2cap socket. One thread poll/select on the socket. At some
> > point, the second thread close the socket. I would expect the first
> > thread to receive a POLLHUP. Note that you can do the same with a
> > single threaded program, closing the listening socket in the signal
> > handler.
> > 	The attached patch is a minor modification of your patch that
> > should fix this issue.
> 
> take a look at tcp_poll() in net/ipv4/tcp.c, because they do it the same
> and do we really have to care about it? How does TCP handles this case?
> 
> Regards
> 
> Marcel

	Well, it never hurts to be on the safe side. And if you look
at my patch, you would see that it's only a trivial change to your
patch. Anyway, up to you...

	Jean