usbfs used by VB process fills syslog after suspend-resume of host (until disk full!)

[Oliver]

I use VirtualBox 3.1.8 (non-OSE) on a laptop host (HP Compaq nx8220) under Ubuntu 9.10. And I run Ubuntu 10.04 as client OS.

When I connect the USB device "Bluetooth adapter" to the client, everything works as expected. But after suspend-resume of the host the VB process begins to fill /var/log/kern.log, /var/log/syslog and /var/log/messages of the host with lines saying:

oliverhp kernel: [ 1232.854921] usb 3-2: usbfs: process 4918 (VirtualBox) did not claim interface 0 before

When I measured, there were about 23 lines per second in each of the 3 logfiles (about 500 kB/min). But the bug got detected after it filled about 300 MB in less than 2 hours until the root partition was full! So I dare to mark this bug as critical. But as not so many people are affected, I won't mind to lower its priority.

Affected USB device:

# lsusb
Bus 003 Device 002: ID 03f0:011d Hewlett-Packard Integrated Bluetooth Module

usbfs in /etc/fstab:

# cat /etc/fstab
none /proc/bus/usb usbfs defaults,devgid=46,devmode=666,nodev,nosuid,noexec 0 0

I have no other suspend-resume problems with my host when VB is not running. No wonder because VB is probably the only application which uses usbfs on my system.

[Oliver]

VBox.log of starting guest, connecting Bluetooth to it and suspend-resume host

[Michael Thayer]

Silly question, but was this a one off thing or is it reproducible. I tried a few times to get it here (admittedly using sysfs to access USB, as I don't have usbfs available on my host system, but the underlying kernel code should be the same) but without any luck.

[Oliver]

No, no. It happens every time under the named conditions.

Some version infos of my system:

> uname -a
Linux oliverhp 2.6.31-21-generic #59-Ubuntu SMP Wed Mar 24 07:28:56 UTC 2010 i686 GNU/Linux

If you want others to compare with your system, please just ask.

[Michael Thayer]

And if you disable usbfs does it make a difference? I don't actually expect that it will, as the kernel code which is used is still the same, but it is still worth testing. If you are interested in tracking this down I will try to get you a special build with more USB logging information sometime over the next week or so.

[Oliver]

I am not sure what you mean by disabling usbfs. I commented it out of /etc/fstab and did a "sudo mount -a". Then I was surprised that switching the BT adapter to guest still works this way! But syslog flooding after resume is still reproducible as you expected.

I'd be glad to help solving this bug. Just send me the files and instructions for testing.

[Oliver]

I did further tests. I booted:

Ubuntu Karmic from an iso, installed VBox 3.1.8 => this bug occurs
Ubuntu Lucid from an iso, installed VBox 3.1.8 => this bug occurs
Ubuntu Lucid from an iso, installed VBox 3.2.0 => this bug still occurs!

Ubuntu Lucid has kernel:

Linux ubuntu 2.6.32-21-generic #32-Ubuntu SMP Fri Apr 16 08:10:02 UTC 2010 i686 GNU/Linux

The syslog is filled forever if the (built-in) Bluetooth adapter is routed to the guest. With a Garmin GPS (non-mass-storage) no message is produced. With a Logitech mouse (with wireless adapter) only a few lines appear saying "process did not claim...".

Perhaps I found some interesting syslog lines right after resuming. They say:

ubuntu kernel: [ 2262.203093] usbfs 2-1:1.0: no reset_resume for driver usbfs?
 ...
ubuntu kernel: [ 2262.837069] btusb 3-2:1.0: no reset_resume for driver btusb?

(see ​syslog_after_resume)

[Oliver]

syslog just after resume only with Logitech mouse routed to guest (kernel 2.6.32-21, VBox 3.2.0)

[Michael Thayer]

Sorry to ask this without providing anything new myself, but is this still relevant with current VirtualBox and Ubuntu versions?

[Oliver]

@michael: I just tested it again and this bug still exists on my host which meanwhile is running Ubuntu 10.04 (kernel 2.6.32-26.48) and Virtualbox 3.2.10-66523 (non-OSE).

It only seems to occur with Linux as guest-OS: it's reproducible with Ubuntu 10.10 and Fedora 14 as guests. But it did not occur with Windows XP as guest.

[Oliver]

Unfortunately this bug is still reproducible with Virtualbox 4.1.4 r74291 on a host with Ubuntu Natty 11.04 (installed from scratch)! "Oracle VM Virtualbox Extension Pack" is installed and active. My hardware is still the same as in the description of this bug report.

[Michael Thayer]

Some googling brought up this link ​http://lists.us.dell.com/pipermail/linux-poweredge/2005-December/023700.html which looks relevant.

[Michael Thayer]

I'm sorry, I obviously didn't properly understand your bug description when you first wrote it (or perhaps it is because my laptop has been replaced in the mean-time). In any case, I am able to reproduce this on my current laptop. My guess is that the host driver gets re-loaded on resume and VirtualBox gets "thrown off" the device without realising.

[Oliver]

Glad to read that you can reproduce it now!

I read your link, and if VB does things like lsusb does then both issues could be related indeed! Your guess sounds reasonable to me too. It is still strange that VB triggers these messages endlessly (or at least until the BT adapter is disconnected from the guest).

[aeichner]

Please reopen if still relevant with a recent VirtualBox release.