VirtualBox

Ticket #5415 (new defect)

Opened 4 years ago

Last modified 15 months ago

VLAN tagged packets pass through fails

Reported by: vlabmichl Owned by:
Priority: major Component: network/hostif
Version: VirtualBox 3.0.10 Keywords: VLAN bridge trunk pass
Cc: Guest type: BSD
Host type: Windows

Description

My configuration: Host(s): Windows XP Pro SP 3 and Mac OS X 10.5.8 Guest: pfSense 1.2.2 and 1.2.3RC3 Networking: Interface #1: Bridge to the physcal host network, Card: Intel Pro Server Interface #2: Internal network connection to other hosts, Card: Intel Pro Server

Neither hosts NIC nor the host system itself are configured to access any of the VLANs. The physical NIC receives the plain trunk data and passes it to VirtualBox.

My target: Passing incoming tagged packets of several VLANs (as a trunk) on the hosts physical NIC into the VM through the bridge of interface1 and allow "Inter-VLAN routing" inside it.

My conclusion: I can't get ANY tagged packets inside of the VM and don't know why. I tested the EXACTLY same configuration on a physical test machine and everything works just fine. So, I'd say that the problem must reside somewhere inbetween the VirtualBox host and the guest.

Change History

comment:1 Changed 3 years ago by Technologov

Added myself as CC

-Technologov

comment:2 Changed 3 years ago by frank

  • Component changed from network to network/hostif

comment:3 Changed 15 months ago by FredrikWendt

Adding myself to CC. I have the exact same need and I can't get it to work with a different setup:

  • HP ProCurve for trunk
  • Ubuntu 12.04 as host OS, virtualbox 4.1.12-dfsg-2ubuntu0.2, physical NIC is "eth0: Intel(R) PRO/1000 Network Connection"
  • Ubuntu 12.04 as guest OS

Using tcpdump on host's VLAN interface, I see the ARP request from the guest, and the ARP reply from a machine on the network (that has passed the HP switch), but the ARP reply is never dumped by tcpdump on the guest OS' VLAN interface. eth0 on the host sees one untagged VLAN's traffic, and several tagged VLANs without any problems. (So it's not a "clean" trunk, one VLAN is sent untagged.)

If I instead bridge host OS' VLAN interface (such as eth0:100) to guest OS' second NIC (eth1), then I can configure eth1 in the guest and have perfect communication. Ie, the guest OS talks to the VLAN on eth1 as were it untagged. This is the fallback/workaround solution I'm using now in production. (I just pray I won't ever need to use more than 8 VLANs in a guest OS.)

Update: I believe this all might relate to the network card I have in the server, Intel PRO/1000, which seems to "strip VLAN Tags in hardware".

Last edited 15 months ago by FredrikWendt (previous) (diff)
Note: See TracTickets for help on using tickets.

www.oracle.com
ContactPrivacy policyTerms of Use