Interfaces cannot receive problem

[abuse007]

Sometimes when I boot a guest VM, one or more of its interfaces will be unreachable. It appears that the guest's interface can transmit, but cannot receive. The problem generally clears if I reboot the VM with the problem. It is rare that a reboot or two will not clear the problem. It appears to be a race condition related to load. If I start multiple VMs simultaneously its nearly always happens.

I am using Virtual Box OSE 2.0.6 on Ubuntu 8.04 (32-bit). The same issue was occurring with VB 2.0.4. I am running multiple Vyatta VC4 guests. The guests generally have 2 or sometimes 3 Ethernet interfaces. These are PCnet-FAST III (AM79C973) host interfaces connected to a tun/tap interfaces. The tun/tap interfaces are bridged using brctl.

In my topology I have 5 routers. Right now R4 is having the problem, and it will not clear. R4's eth0 is connected to the hosts vbox41 tun/tap interface. vbox41 and vbox14 (which is an interface on R1) are bridged to br14 on the host. br14 has the IP address 10.1.14.254/24 on the host. R4 eth0 has IP 10.1.14.4/24, and R1 has IP 10.1.14.1/24.

R1 can ping 10.1.14.254 with no problems.
R4 cannot ping 10.1.14.254.

If I sniff on R4's eth0 (the broken interface), and do a ping to the host's bridge (10.1.14.254) I will see it send out ARP requests, but there are no ARP replies coming back. At the same time if I sniff on the host's br14 bridge interface, I see the ARP requests come in and I see the ARP replies go back out. The ARP replies that I see on the bridge are not making it to the guest's eth0 interface.

I have no idea how to troubleshoot this issue. It appears to be a VirtualBox issue as I have used the same setup with other VMMs with no problems. If I bounce the interfaces, remove the bridge etc, I have the same broken behaviour. However if I restart the VM it generally clears. Tonight however the problem is constant, I cannot clear it.

How do I troubleshoot this issue, and why is it happening? I am my wits end, and I'm about to ditch VirtualBox if I cannot resolve this issue.

[Sasquatch]

It's too bad that you use the OSE, as the PUEL version has the Intel/PRO 1000 NICs available, which you could try to use. You can switch to the PUEL without much problems, but you need to completely remove it prior to installing the PUEL version. You also need to recreate the Host Interfaces, as it's part of the config.

[abuse007]

I have changed the interface and it seems to have stoped. I have not noticed this problem so far, but time will tell. Thank you for the suggestion. Perhaps this issue is caused by a timing issue in that driver.