[TriLUG] Good report on Linux, Stay-Online, and IOGear KVMs

[John Matthews]

It is often remarked that one hears only about negative experiences
(e.g. from customers) and too rarely hears the good experiences. I
wanted to report on a good experience I had recently, both with Linux
and a local reseller. Although this is a story about difficulties I had
both with hardware and with Linux, the outcome is positive.

When I asked about a USB KVM, the Belkin one at Best Buy was mentioned
and I tried it. It is not a USB KVM in the sense that I wanted one, so
it went back to the store. (It requires the use of a PS/2 keyboard and
mouse, but can send the keyboard/mouse signals to computers along a USB
cable. Pretty silly setup, really.)

The next day I went to Stay-Online, a reseller often mentioned here and
one for which TriLUG users get a discount. They had a very new USB KVM
that I had never seen before from IOGear. It was similar to the ones
that IOGear already had, but the model number was different. The price?
$140 for TriLUGgers, and they even gave me the discount even though I
didn't have my card with me (d'oh!). When I realized it was a true USB
KVM and could use hotkeys on the keyboard and had an on-screen display
(OSD) and came with two good-sized VGA/USB KVM cables I was ready to buy
it! (For reference, the Belkin KVM at Best Buy sells for $150 and
doesn't include cables, OSD, or hotkeys but does handle audio. And it's
not true USB.) So I plonked down my $140 and went home a happy guy.

Except that when I got home, the KVM refused to play nice with Linux!
Sure, it worked like a charm in Windows (2000), and the video switching
and hotkeys and OSD worked well. But Linux refused to take any USB
signals from it whatsoever. I tried a newer kernel. No luck. IOGear's
site had a press release about this KVM but no support documentation
yet. I looked again: yes, it said "Linux support" on the box.
Crestfallen, I figured I was about to return my second KVM in as many
days.

I decided to be stubborn and dig in and find out more before I gave up.
On groups.google.com I found a thread going on that very day about a
similar USB KVM from IOGear. On the linux-usb-users mailing list I found
one message about yet another IOGear USB KVM but the developers had
never responded. I posted my story, and links to the other two reports,
to linux-usb-users and waited. Eventually, Greg K-H (Kroah-Hartman)
responded and asked for a list of modules the kernel was loading and any
relevant kernel messages. I collected information like a madman and sent
it all in to the list. After a few more exchanges, the response was
"looks like the hub in the KVM isn't being seen by the kernel; if the
device is invisible to the kernel, we can't do anything."

As it turns out, the day I got that message I was leaving town for a 10
day trip for work. I would not have my KVM with me and could not do
anything with it during this time. David Tilley, also on the
linux-users-usb mailing list came to the rescue: he found out that Red
Hat's 2.4.2 kernel worked and that starting from that point, standard
kernel.org kernels worked up to 2.4.9 but broke with 2.4.10. This break
would lead to the ultimate fix, as Greg pointed out that a big HID code
modification had taken place between 2.4.9 and 2.4.10. This change had
indeed broken compatibility with these KVMs, and Vojtech Pavlik, the
author of that change, created a patch to fix the problem in the 2.5
kernel tree.

By the time I got home, a similar patch for the 2.4 series had been
submitted and it took me about 10 minutes to patch, recompile, and
reboot into a working USB KVM.

While this doesn't look great for IOGear, the patch really seems to be
the addition of their KVM hub (by ATEN) to a blacklist of devices that
don't play nice, they are at least putting Linux prominently on their
hardware's boxes. Further, when that device's packaging was designed and
the testing done it was probably well in advance of the code shift in
2.4.10 and I don't think that makes their claim unreasonable. This does
look good for Linux, since it clearly shows some of the strengths of the
development model: user/developer interaction, freely available code
versions for testing (as David Tilley did), and the quick distribution
of patches to those that need or want them.

Anyway, I would buy another IOGear KVM if I ever needed it and I would
surely consider Stay-Online more often that I would have before, since a
price like the one they gave me just can't be beat. And the USB for
Linux developers seem to be doing an excellent job, developing code that
is already quite good and being ready to help fix it when problem
devices show up on the market.

Regards,
matt

-- 
Matt Matthews     \ ph: 919.660.2811        \ Use GNU/Linux  _o) w00t
Duke Univ., Postdoc\ jvmatthe at math.duke.edu  \____________   /\\
Dept. of Mathematics\ http://www.math.duke.edu/~jvmatthe/ \ _\_V