Date: Tue, 10 Jul 2012 09:29:31 +0400 From: Gleb Smirnoff <glebius@FreeBSD.org> To: Eugene Grosbein <egrosbein@rdtc.ru> Cc: bzeeb-lists@lists.zabbadoz.net, Przemyslaw Frasunek <venglin@freebsd.lublin.pl>, Mike Tancsa <mike@sentex.net>, Ryan Stone <rysto32@gmail.com>, freebsd-net@FreeBSD.org Subject: Re: mpd5/Netgraph issues after upgrading to 7.4 Message-ID: <20120710052931.GV21957@glebius.int.ru> In-Reply-To: <4FFBB7A8.90201@rdtc.ru> References: <4FDB1D71.6050908@freebsd.lublin.pl> <20120615203142.GW28613@glebius.int.ru> <4FDBAFD7.9020606@freebsd.lublin.pl> <4FDF2F81.6030307@sentex.net> <4FDF3097.6080701@freebsd.lublin.pl> <4FE0EE62.5070905@freebsd.lublin.pl> <4FF7F2C6.5070401@freebsd.lublin.pl> <20120709081225.GJ21957@glebius.int.ru> <CAFMmRNydo=Bci=WM5mODS0vhiv8%2B2Oj39XKTXkEymYOJunhPxg@mail.gmail.com> <4FFBB7A8.90201@rdtc.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Jul 10, 2012 at 12:03:36PM +0700, Eugene Grosbein wrote: E> 10.07.2012 03:25, Ryan Stone пишет: E> > On Mon, Jul 9, 2012 at 4:12 AM, Gleb Smirnoff <glebius@freebsd.org> wrote: E> >> This looks very much related to a known race in ARP code. E> >> E> >> See this email and related thread: E> >> E> >> http://lists.freebsd.org/pipermail/freebsd-net/2012-March/031865.html E> >> E> >> Ryan didn't check in any patches since, and I failed to follow on this E> >> problem due to ENOTIME. E> >> E> >> I've added Ryan to Cc. Ryan, what's the status of the problem at your E> >> side? Did you come to any solution? E> > E> > Unfortunately I was never able to come to a satisfactory solution. As E> > I recall, in the end I ran headlong into problems with making the E> > locking sane. The big problem was with arpresolve. At one point it E> > calls callout_reset to schedule the LLE's la_timer. In my patch this E> > would have to be done with a write lock help on the afdata lock. E> > However, this acquisition would have to be done before taking the E> > LLE_LOCK to prevent a LOR, and in the end you conclude that you have E> > to take a write lock on the ifnet's afdata lock for every packet that E> > goes through arpresolve, which was a non-starter. That's the point E> > that I reached before I got distracted by other things at $WORK. E> > E> > As I recall, the in6 case was even worse, as the in6 equivalent of E> > arptimer is significantly more complicated and likes to do crazy E> > things like dropping locks. E> E> It seems, Przemyslaw Frasunek uses proxyarp? E> I have no such problems but I do not use proxyarp. E> Could you get rid of it, Przemyslaw? I'm not sure this is related. The race happens w/o proxy arp as well. -- Totus tuus, Glebius.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120710052931.GV21957>