On Thursday 07 May 2009, Ilpo Järvinen wrote:
quoted text
> On Thu, 7 May 2009, Matthias Andree wrote:
> > I've applied the following patch to net/ipv4/tcp.c for the kernel
> > running on my server (2.6.29-rc8):
> > @@ -1499,8 +1499,9 @@ do_prequeue:
> >    		}
> >    		if ((flags & MSG_PEEK) && peek_seq != tp->copied_seq) {
> >    			if (net_ratelimit())
> > -				printk(KERN_DEBUG "TCP(%s:%d): Application bug, race in MSG_PEEK.\n",
> > -				       current->comm, task_pid_nr(current));
> > +				printk(KERN_DEBUG "TCP(%s:%d): Application bug, race in MSG_PEEK: %x, %x.\n",
> > +				       current->comm, task_pid_nr(current)),
> > +				       peek_seq, tp->copied_seq;
>
> I cannot resist myself from noting that this certainly wasn't the patch
> one got those printks below... It might happily compile though :-).

Can you please elaborate why you think that? It may be horribly broken
(I've never claimed to be a C coder, and probably never will), but it
also really is the patch that generates the printks...

quoted text
> >    			peek_seq = tp->copied_seq;
> >    		}
> >    		continue;
> >
> > So, the values you see at the end of the warning are peek_seq and
> > tp->copied_seq. This gives messages like:
> > kernel: TCP(fetchmail:9311): Application bug, race in MSG_PEEK: 156233, 16a.
> > kernel: TCP(fetchmail:9311): Application bug, race in MSG_PEEK: 71259ac8, 5b4.
> > kernel: TCP(fetchmail:31216): Application bug, race in MSG_PEEK: 833fe5, c0.

[...]

quoted text
> What would you think about the following, untested patch... I suppose
> it is enough to capture the racy situations except with that crazy urg
> hole, grr (I suppose that will need just another variable to do the
> offset of one).

I'll give your patch a try and report back.

Thanks,
FJP
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html