IPv6 issues with local-link addresses #11

Closed
sjorge opened this Issue Jul 3, 2012 · 5 comments

3 participants

@sjorge

I was seeing no IPv6 in my KVM guests on OpenIndiana,
They where not seeing my routers RA's.

rmustacc recommended a qemu-kvm with 6df50a0 applied to it.
After switching to 151a5 which included the said commit, I am indeed seeing the RA's using tcpdump -vvv ip6.

So that is good! However I keep hitting this problem: (from dmesg)
[ 5.016947] eth0: IPv6 duplicate address fe80::8:20ff:fe92:d520 detected!

eth0 then fails go achieve a link UP, this was an arch vm, I tried using virtio and e1000 nics.
This was an existing vm... so I didn't trust it.

I tried with a clean install of fedora:
[ 4.962467] e1000 0000:00:03.0: �eth0: (PCI:33MHz:32-bit) 02:08:20:16:ca:c8
[ 4.962472] e1000 0000:00:03.0: �eth0: Intel(R) PRO/1000 Network Connection
[ 36.175892] e1000: �eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX
[ 36.176501] ADDRCONF(NETDEV_UP): �eth0: link is not ready
[ 36.176656] ADDRCONF(NETDEV_CHANGE): �eth0�: link becomes ready
[ 36.748687] �eth0: IPv6 duplicate address fe80::8:20ff:fe16:cac8 detected!

Ok, a bit more verbose! Nice, but essentially, the same error.
So it is already seeing a local-link address for it's own mac and refuses to bind it.

gravity energy-test # traceroute fe80::8:20ff:fe16:cac8
traceroute: Warning: Multiple interfaces found; using fe80::6a05:caff:fe01:f340 @ host0
traceroute to fe80::8:20ff:fe16:cac8, 30 hops max, 60 byte packets
1 fe80::6a05:caff:fe01:f340 2999.976 ms !A 2994.905 ms !A 2999.923 ms !A

But the address is visible on the host. What about a different machine?
I tried to ping it using my gateways diagnostics page:
Ping check did not deliver a result, because of a probably non-existing ip address / hostname.

My initial reaction was that the interface was pumped in solaris it self:
gravity energy-test # ipadm show-addr
ADDROBJ TYPE STATE ADDR
lo0/v4 static ok 127.0.0.1/8
host0/v4static static ok 172.16.20.100/24
lo0/v6 static ok ::1/128
host0/v6auto addrconf ok fe80::6a05:caff:fe01:f340/10
host0/v6static static ok 2001:6f8:1480:20::100/64

But this turned out not to be the case.

I guess this 'is' progress to not seeing the RA's at all, but there may still be a bug somewhere.
I'll try to keep this up to date if I find new things.

I can also be found in #illumos if needed.

--- configuration for fedora vm ---
gravity energy-test # dladm show-vnic
LINK OVER SPEED MACADDRESS MACADDRTYPE VID
...
venergy0 trunk0 1000 2:8:20:92:d5:20 random 300
venergy1 trunk0 1000 2:8:20:16:ca:c8 random 300
...

gravity energy-test # cat start_test.sh
#!/usr/bin/bash

configuration

VNIC=venergy1
HD=/dev/zvol/dsk/main/vms/kvm/energy-test/disk0
CD=/main/vms/kvm/energy-test/Fedora-17-x86_64-Live-XFCE.iso
VNC=10
VCPU=2
MEM=1024

variables

MAC=dladm show-vnic -po macaddress $VNIC

run

CDROM=
if [ -f ${CD} ]; then
CDROM="-drive file=$CD,media=cdrom,if=ide,index=2"
fi

/usr/bin/qemu-kvm \
-boot cd \
-enable-kvm \
-vnc 127.0.0.1:$VNC \
-cpu host \
-smp $VCPU \
-m $MEM \
$CDROM \
-drive file=$HD,if=virtio,index=0 \
-net nic,vlan=300,name=net0,model=e1000,macaddr=$MAC \
-net vnic,vlan=300,name=net0,ifname=$VNIC,macaddr=$MAC \
-usb -usbdevice tablet \
-vga std -daemonize \
-name test -pidfile test.pid

@sjorge

Pretty sure it is related to Issue #13
IPv6 fails to configure correctly due to the address being in use. (Which makes sense because it is seeing it's own traffic)

@git-jonathankinney

Watching, I have the same issue, as I am sure anyone does who is using KVM on illumos. I am running on oi_151a7 to be specific. The fix mentioned http://fnordig.de/2012/09/16/getting-ipvworking-with-smartos-and-kvm/ does seem to get it functioning, the short of it is to enter "net.ipv6.conf.eth0.accept_dad=0" into the "/etc/sysctl.conf" and reboot, no more messages, and functional (at least pingable) IPv6. It does not actually solve the cause of the issue though.

@rmustacc
Joyent member

I've just verified that this is solved with the recent bardiche work. I'll close this ticket when that lands.

@sjorge

I think this one can be closed already, since this is clearly the result of #13

@rmustacc
Joyent member

Resolved through joyent/illumos-joyent@ad740f5 and 31a39c6.

@rmustacc rmustacc closed this Mar 20, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment