Archive for February, 2009

USB pass-thru in RHEL 5 Xen VM doesn’t work; why do I buy support?

February 23, 2009

As part of my efforts to maintain ACE+TAO on LabVIEW RT (with Pharlap ETS kernel) I have a setup to run the test suite on a National Instruments chassis, driven by the build system on Windows. This arrangement is easily handled by ACE+TAO build environment, including a mechanism to reboot the NI box when things go wrong. The reboot is triggered by a USB-connected NI USB-6009 device that trips the reset signal on the NI box. It’s very slick and keeps from having to cycle power. The hitch is that it requires a USB 2.0 connection from the Windows machine.

In the past I’d used a VMware virtual machine hosted on Linux (RHEL 4) running a Windows guest OS to host this test environment. The VMware software passed the USB device through to the Windows VM without a hitch. However, over the winter I got a new machine set up with a great deal more capacity and decided to move the LabVIEW RT test environment to the new machine which runs RHEL 5 and Xen.

And that’s when the trouble started…

First I had to search quite a bit to find out how to configure the Xen VM to pass the USB device to the guest OS. After a bit of googling and reading, I found the magic configuration lines to add. I also found another blog entry (http://www.olivetalks.com/2008/02/03/usb-forwarding-on-xen-it-just-does-not-work/) saying it wouldn’t work right. But I forged on, confident that even if it didn’t work “out of the box” I had purchased support from Red Hat and could get any help I needed.

Well, long story short, the USB device didn’t pass through correctly from Xen. On December 9, 2008 I opened a support case with Red Hat to have them do whatever is needed to make it work. After twelve (12) exchanges over 22 days I requested escalation to someone who could do more to help than quote manual sections that were not applicable to what I needed.

After 11 more exchanges with 3 more support engineers over another 49 days I got the long-awaited answer: “It doesn’t work.”

Well, I wasn’t totally surprised since I had no success and had already seen a blog posting saying it won’t work. But I was still clinging to hope that my support contract would come through and Red Hat would make it work. Nope. Sorry. It don’t work. End of story.

So why do I buy support? Sure, I get all the updates, but I paid extra for someone to actually work on problems for me and all I get is “It doesn’t work.”? When my customers raise issues about ACE not working, they get fixes. Solutions. You know, like they paid for.

Apparently, solutions are optional for other providers.

So what happened in the end? I went back to running the Windows VM in a VMware environment, where it’s happily chugging along.