virtualbox.org

Not sure if this applies to anyone here but hopefully it might help someone.

I was beating my head against the same issue today (bridged NIC not working with WinXP guest) and found out the problem: while I have 2 NICs in my linux host, I'm only using one and was accidentally specifying the inactive NIC in the network configuration of the VM .

There's an option in the network configuration that gives you a choice of NICs and it was set to eth0, which looked right, but in fact the active NIC on my host is eth1 -- switched it to eth1, now works beautifully.

NAT worked fine because it doesn't require you to select a specific NIC.

Sasquatch wrote:You can install a packet sniffer pretty easily. Wireshark can capture packets which you can check for proper DHCP communication between the VM and router. Compare it with the results you get when you request a DHCP address on the Host. For maximum debugging, run Wireshark on the Host and Guest when obtaining an IP in the Guest. You can then see where it fails. It is possible that a firewall on either the Host or Guest blocks it, Wireshark should show that.

Thanks for turning me on to Wireshark, much faster end result than what I was planning with cli tools. The good news is that even via bridge it seems to be using the MAC from the guest not the host.

The interesting news is that going through my switch (Cisco 2950T-24) seems to throw off the guest getting to the Firewall to get its IP, but the host has no problem at all. I needed to see from a middle-man sniffer what was going on and knew I needed to change to a basic HUB instead of my switch. Once I swapped the cables from switch to the hub for my PC, Firewall, and MBP (the middle-man), starting up the guest just worked in bridge mode. Power off the guest, change the host to be DHCP as well, and started the guest and still the guest received its IP. Went back to using the 1st NIC and the guest still worked. So, it's not a host DHCP vs guest DHCP. Although it would be interesting to test the host getting static DHCP entry in the FW and test, then add the guest as a static entry and test to see if they still work. (I want that to work at the end)

So, in order to troubleshoot this further, I will need more info as to what the guest in bridge mode using DHCP is doing in detail vs the host and research that info with Cisco. Unless someone might know what is happening so I don't reinvent the troubleshooting wheel - BTW, I did have the FW in a FastEthernet port and my PC in one of the two 1-GigE ports, but as a last test tonight I moved the PC to a FE port and same results, start up guest in bridge mode does not get IP via DHCP.

I can't continue using a hub as a workaround, I need a switch and even thinking of upgrading my 2950T to an all GigE switch. Would be interesting if any switch causes this even the < $50 ones just because it's a switch vs hub. I don't have access to any other switches to test this question.

I do know that managed switches like your Cisco can block certain traffic if it detects more than one MAC address on one port. What you can do, is hook up the second NIC to the switch and let the VM use that. The Host needs to have it disabled in some way. See how that goes. I can't guarantee that it will work, because the switch will probe it's link to learn the MAC address on the other end and you might end up with the same problem.
But the place you need to look at, is at the switch configuration. That's the source of your problem.

Sasquatch wrote:But the place you need to look at, is at the switch configuration. That's the source of your problem.

I agree, but I would need to know what's diff about how the vm does it's DHCP client vs the host (cause host always gets its IP) in order to work with Cisco to figure out what to change if possible to resolve the issue and report back so others might be helped if they run into this issue.

What would be best is someone that's knows the DHCP client code and/or the way it communicates for Windows XP and OpenSolaris 2009.06 (snv 111b) guests since they act the same and do not get an IP if these guests are in bridge mode while physical port is connected to the switch.

In answer to your possible workaround, Ports from the PC go into an outlet in the wall, then down to the patch panel, then into the switch and I don't have an extra port in the room with the PC. So, I can't use your suggestion as a workaround but sounds like it might have worked. (PC is a 10-bay extra large all steel case ie: too large/heavy to move - can't really bring rack-mounted switch up)


BTW, since this is happening to OpenSolaris guest in bridge mode as well, shouldn't this be moved to the Windows Host forum? Just thinking out loud...

Try turning off VT-X in your BIOS.

I'm running 3.0.2 on a WinXP host, Ubuntu 9.04 guest. I discovered that I hadn't enabled VT-X in the BIOS, but when I did, my guest could no longer get an IP address. Network Manager would just time-out. After disabling VT-X again, DHCP works as it should.