Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 11 Aug 2012 17:58:10 +1000 (EST)
From:      Ian Smith <smithi@nimnet.asn.au>
To:        "Christoph P.U. Kukulies" <kuku@kukulies.org>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: weird problem with 9.0 Release and ed0
Message-ID:  <20120811171652.I93465@sola.nimnet.asn.au>
In-Reply-To: <20120810120045.C92A91065691@hub.freebsd.org>
References:  <20120810120045.C92A91065691@hub.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
In freebsd-questions Digest, Vol 427, Issue 6, Message: 16
On Fri, 10 Aug 2012 12:39:36 +0200 "Christoph P.U. Kukulies" <kuku@kukulies.org> wrote:
 > Am 10.08.2012 11:40, schrieb Christoph P.U. Kukulies:
 > > Am 10.08.2012 11:28, schrieb Christoph P.U. Kukulies:
 > >> The problem need not to be confined to 9.0. It stated to develop 
 > >> under 5.1 already.
 > > read: started to develop...
 > >>
 > >> I'm running a natd gateway machine that was developing strange 
 > >> behaviour such that the
 > >> outside interface (ed0, BNC connector) that was connected via a small 
 > >> media converter switch to
 > >> the providers sync line had dropouts. The machine couldn't ping into 
 > >> the Internet and also couldn't be pinged.
 > >>
 > >> I first thought it was the switch/media converter, but another 
 > >> (Windows XP) machine that was on the
 > >> same BNC cable worked flawlessly.

That XP box was directly on the outside, not inside nat'd via this one?

 > >> So I decided to migrate that 5.1 machine to a 9.0 machine. The 
 > >> situation now is that I have the9.0 machine
 > >> at the BNC cable and simultanously the old FreeBSD 5.1 gateway on the 
 > >> same BNC cable but through a
 > >> TP adapter. This was the old machine works fine and I can care about 
 > >> the new machine.

Not quite clear .. can you sketch your network configuration?

 > >> Is there a known problem with ed0 cards that have the Realtek 8029 
 > >> chipset. Do they need some
 > >> special flags like memory mapping or irq?

Long time since I've run anything with 10base2/BNC, but it used to work 
ok, on an ed0.

 > >> When I for example boot the 9.0 machine the comping up of the em0 (on 
 > >> mainboard interface results in a highlighted
 > >> kernel message on the console. The coming up of the ed0 is not 
 > >> flagged this way. And as a result the
 > >> ed0 interface seems to be dead.

Does the outside interface have a static address, or do you use DHCP 
via the provider's switch/hub/whatever?  Show /etc/rc.conf setup.  It 
smells a bit like the interface may not be up soon enough at that time; 
the ntpd message below could also indicate something like that re ipv6.

 > >> Here some excerpts of dmesg:
 > >> em0: <Intel(R) PRO/1000 Network Connection 7.2.3> port 0x4400-0x441f 
 > >> mem 0x93100000-0x9311ffff,0x93124000-0x93124fff irq 20 at device 25.0 
 > >> on pci0
 > >> em0: Using an MSI interrupt
 > >> em0: Ethernet address: 00:1c:c0:37:b2:9f
 > >>
 > >> ed0: <RealTek 8029> port 0x1000-0x101f irq 22 at device 1.0 on pci7
 > >> ed0: Ethernet address: 00:e0:7d:7c:2b:4a
 > >>
 > >> I also see this:
 > >> Jul 30 23:03:54 forum ntpd[1711]: unable to create socket on ed0 (20) 
 > >> for fe80::
 > >> 2e0:7dff:fe7c:2b4a#123

You should get more / better clues if you boot with verbose messages.

 > > Forgot to add this info:
 > >
 > > ed0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
 > >         ether 00:e0:7d:7c:2b:4a
 > >         inet 80.72.44.230 netmask 0xfffffff0 broadcast 80.72.44.239
 > >         inet6 fe80::2e0:7dff:fe7c:2b4a%ed0 prefixlen 64 scopeid 0xa
 > >         nd6 options=29<PERFORMNUD,IFDISABLED,AUTO_LINKLOCAL>
 > >         media: Ethernet autoselect (10base2/BNC)
 > >
 > 
 > Must add some more info:
 > 
 > My kernel config:
 > 
 > cpu             I486_CPU
 > cpu             I586_CPU
 > cpu             I686_CPU
 > ident           DIVERT
 > 
 > makeoptions     DEBUG=-g                # Build kernel with gdb(1) debug 
 > symbols
 > options IPFIREWALL
 > options IPFIREWALL_VERBOSE
 > options IPFIREWALL_VERBOSE_LIMIT=10
 > options IPDIVERT
 > options IPFIREWALL_DEFAULT_TO_ACCEPT
 > 
 > (the rest like in GENERIC).

Just to mention: you don't actually need to include FIREWALL* or DIVERT 
in kernels these days; a GENERIC kernel will work fine, loading modules 
as needed.  Only exception is if you needed FIREWALL_FORWARD, which it 
appears you don't.

 > Strange thing:
 > 
 > I cannot ping neither the outside interface address nor the inside 
 > (172.27.2.115)
 > 
 > --
 > Christoph Kukulies

Please show output from:

# egrep 'ifconfig|firewall|natd|gateway|ntpd' /etc/rc.conf
# cat /etc/natd.conf
# ipfw show
# netstat -finet -rn

cheers, Ian



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120811171652.I93465>