On 12/10/2007, Christian Plattner <plattner@cleondra.ch> wrote:
quoted text
> Hi,
>
> Today something strange happened on one of my Soekris 5501 boxes,
> it runs OpenBSD 4.1-stable. The box is connected with a cross-over cable
> to another machine via the vr1 interface (the box has 4 vr interfaces).
>
> Problem: After having rebooted the machine at the other end of the cable
> multiple times, the Soekris box suddenly stopped receiving packets on
> the vr1 interface.
>
> After playing around with ping and tcpdump on both sides I found out
> that the vr1 interface allowed me to send packets, but incoming packets
> did no show up in tcpdump, even though the LED on the interface was
> flickering.
>
> I changed the cable, connected the vr1 of the Soekris to another
> machine, then to a switch port with lots of broadcast traffic etc.etc.
> nothing helped, ingress traffic on vr1 did not show up in tcpdump.
>
> Solution: Finally, immediatelly after doing "ifconfig vr1 down &&
> ifconfig vr1 up" everything worked again as normal.
>
> The link on vr1 is currently only used to do SSH between the two
> machines, so this is really a low traffic link. On the other hand,
> vr0,vr2,vr3 are heavily used (BGP sessions etc., the Soekris is at the
> border of my AS). btw: the machine at the other end of the cable is a
> Soekris 5501 as well. Had to reboot it to perform a BIOS upgrade.
>
> No suspicious output in dmesg.
>
> Any ideas on how I could further track down the problem?
>
> Thanks,
>   Christian
>
> The vr interfaces in dmesg:
>
> vr0 at pci0 dev 6 function 0 "VIA VT6105M RhineIII" rev 0x96: irq 11,
> address 00:00:24:c8:de:68
> ukphy0 at vr0 phy 1: Generic IEEE 802.3u media interface, rev. 3: OUI
> 0x004063, model 0x0034
> vr1 at pci0 dev 7 function 0 "VIA VT6105M RhineIII" rev 0x96: irq 5,
> address 00:00:24:c8:de:69
> ukphy1 at vr1 phy 1: Generic IEEE 802.3u media interface, rev. 3: OUI
> 0x004063, model 0x0034
> vr2 at pci0 dev 8 function 0 "VIA VT6105M RhineIII" rev 0x96: irq 9,
> address 00:00:24:c8:de:6a
> ukphy2 at vr2 phy 1: Generic IEEE 802.3u media interface, rev. 3: OUI
> 0x004063, model 0x0034
> vr3 at pci0 dev 9 function 0 "VIA VT6105M RhineIII" rev 0x96: irq 12,
> address 00:00:24:c8:de:6b
> ukphy3 at vr3 phy 1: Generic IEEE 802.3u media interface, rev. 3: OUI
> 0x004063, model 0x0034
>
> # ifconfig vr1
>
> vr1: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
>          lladdr 00:00:24:c8:de:69
>          media: Ethernet autoselect (100baseTX full-duplex)
>          status: active
>          inet XX.XX.XX.XX netmask 0xffffffe0 broadcast XX.XX.XX.YY
>          inet6 fe80::200:24ff:fec8:de69%vr1 prefixlen 64 scopeid 0x2

Not sure if related, but something similar has been fixed in
4.2-current already.

http://lists.freebsd.org/pipermail/freebsd-current/2007-August/076486.html
http://www.openbsd.org/cgi-bin/cvsweb/src/sys/dev/pci/if_vr.c#rev1.70

I think you should be able to safely apply the 1.69 to 1.70 diff to
your source tree if this is of a concern. The diff is for vr_attach()
only, so if your system is already up and running and you never reboot
it, then you probably shouldn't bother until your next upgrade.

Cheers,
Constantine.