Host interface networking crashed the host (CentOS 5.3 x86_64) => Fixed in SVN

[Peter Schneider]

When host interface networking is used, the host crashes on guest VM startup; machine freezes, two keyboard LEDs start blinking and the host even cannot be reached from the network any more.

Host is a Core2Quad Q9400, 8 GB RAM, Mainboard is Asus P5B Deluxe, eth0 is the onboard Marvell 88E8001 Gigabit Ethernet Controller. Host OS is CentOS 5.3 x86_64, and eth0 uses the skge driver. Crash occurs with kernels 2.6.18-128.el5 and 2.6.18-128.1.6.el5.

Guest VM OS is Win2000; I have cloned this from my laptop where this setup with VirtualBox 2.1.4 (64bit) is running without a problem on Ubuntu 8.04.2 (kernel 2.6.24-23-generic). Please note: the crash also occurs if I try a CentOS 5.2 (32bit) guest VM instead, so this does not seem to be related to the guest.

After reboot, /var/log/messages has these entries as the last ones before the crash:

Apr  6 02:53:26 euler kernel: VBoxDrv: dbg - g_abExecMemory=ffffffff88d74ac0
Apr  6 02:53:26 euler kernel: vboxdrv: fAsync=0 offMin=0x488 offMax=0x1aa0
Apr  6 02:53:26 euler kernel: vboxdrv: TSC mode is 'synchronous', kernel timer mode is 'normal'.
Apr  6 02:53:26 euler kernel: VBoxNetFlt: dbg - g_abExecMemory=ffffffff88f16ce0
Apr  6 02:55:36 euler kernel: device eth0 entered promiscuous mode
Apr  6 02:59:20 euler syslogd 1.4.1: restart.

I tried the workaround (I think from tickets 2827/2829) suggested here: ​http://forums.virtualbox.org/viewtopic.php?f=7&t=16082 but this did not help.

[Peter Schneider]

The issue still exists with VirtualBox 2.2.0. This time however, the host does not crash immediately, but very soon after guest VM startup. Judging from /var/log/messages it is even the case that the host (euler, 192.168.1.1, running sama) and the guest (win-client, 192.168.1.150) communicate with each other over the host interface link to negotiate who is master browser.

Apr  9 15:43:08 euler kernel: VBoxDrv: dbg - g_abExecMemory=ffffffff88d93d40
Apr  9 15:43:08 euler kernel: vboxdrv: fAsync=0 offMin=0x510 offMax=0x1428
Apr  9 15:43:08 euler kernel: vboxdrv: TSC mode is 'synchronous', kernel timer mode is 'normal'.
Apr  9 15:43:08 euler kernel: VBoxNetFlt: dbg - g_abExecMemory=ffffffff88f36ec0
Apr  9 15:45:56 euler kernel: device eth0 entered promiscuous mode
Apr  9 15:47:06 euler nmbd[3793]: [2009/04/09 15:47:06, 0] nmbd/nmbd_incomingdgrams.c:process_local_master_ann
ounce(309) 
Apr  9 15:47:06 euler nmbd[3793]:   process_local_master_announce: Server WIN-CLIENT at IP 192.168.1.150 is an
nouncing itself as a local master browser for workgroup ATHOME and we think we are master. Forcing election. 
Apr  9 15:47:06 euler nmbd[3793]: [2009/04/09 15:47:06, 0] nmbd/nmbd_become_lmb.c:unbecome_local_master_succes
s(149) 
Apr  9 15:47:06 euler nmbd[3793]:   ***** 
Apr  9 15:47:06 euler nmbd[3793]:    
Apr  9 15:47:06 euler nmbd[3793]:   Samba name server EULER has stopped being a local master browser for workg
roup ATHOME on subnet 192.168.1.1 
Apr  9 15:47:06 euler nmbd[3793]:    
Apr  9 15:47:06 euler nmbd[3793]:   ***** 
Apr  9 15:50:03 euler syslogd 1.4.1: restart.
Apr  9 15:50:03 euler kernel: klogd 1.4.1, log source = /proc/kmsg started.
Apr  9 15:50:03 euler kernel: Linux version 2.6.18-128.1.6.el5 (mockbuild@builder10.centos.org) (gcc version 4
.1.2 20080704 (Red Hat 4.1.2-44)) #1 SMP Wed Apr 1 09:10:25 EDT 2009

But soon after, the host crashes nevertheless. This time, I can attach guest VM log.

[Frank Mehnert]

Please check if the problem is gone with VirtualBox 2.2.2. There was a small change in the netfilter code which could fix your problem.

[Peter Schneider]

Unfortunately, this bug is still present with VirtualBox 2.2.2. The host machine still crashes, with exactly the same symptoms as before. For the time being, I'm using a bridged tap/br setup instead of eth0 as a workaround; however being able to use eth0 and thus an IP address in the same subnet would be much nicer.

[Frank Mehnert]

Thanks for the feedback. Which kernel version is that (sorry, no CentOS 5.3 handy here)?

[Peter Schneider]

This is the kernel version:

[ps@euler ~]$ uname -rp
2.6.18-128.1.6.el5 x86_64

[Frank Mehnert]

Installed Cent OS 5.3 x86_64, WinXP guest, bridged networking: No problem so far. I fear this is somehow specific to the driver of your onboard network controller. I have only E1000 network cards here...

[Peter Schneider]

Ok, I think I will then try to file a bug against the driver in the CentOS bug tracker and get back to here if this yields some new information. Frank, thanks for your effort!

[Frank Mehnert]

The problem is most probably solved. See ​here fox a fix. Feedback welcome.

[Peter Schneider]

I installed VB 2.2.4 and the problem is gone - everything works fine now with hostif/bridged networking setup using eth0. Thanks very much for fixing this!

[Frank Mehnert]

Thanks for the feedback!