KVM issues with recent 2.6 kernels

KVM issues with recent 2.6 kernels

[Ken Moffat]

Hi,

 I'm seeing some oddities on two 2.6 boxes here.  I use a Belkin Omni
4-way PS/2 KVM switch.  It has a "hot key" to switch boxes (scroll-lock
twice, then number 1-4).  This has always been a little problematic on
2.4 (you have to get the timing of the key presses right, worst case
when you get to the session either the number has been treated as input,
or the display has scroll-lock turned on), but usable.

 On the two boxes running 2.6 (Celeron 1GHz with 2.6.4, Duron 1GHz with
2.6.5) I'm now seeing two strangenesses:

 The celeron is running console sessions, including a long-winded series
of shell scripts run from a Makefile.  I've started to notice that if I
start the make then switch to another machine, sometimes when I come
back the build is "stalled" between the scripts - the console looks as
if scroll-lock is on, but when I free it the next stage of the build
starts (confirmed from `top').  My expectation is that the whole build
would have continued, even if that particular console display was
somehow "locked".

 The duron is now being used to do things in xterms.  From time to time
the alt key stops being recognised (no alt-tab to switch windows, no
ctrl-alt-n to switch desktops).  And then after again switching machines
it suddenly starts working properly again.

 Any suggestions on where to look, or which parts of my .config would be
relevant ?

Ken
-- 
 das eine Mal als Tragödie, das andere Mal als Farce

Re: KVM issues with recent 2.6 kernels

[Jurriaan]

From: Ken Moffat <ken@kenmoffat.uklinux.net>
Date: Mon, Apr 19, 2004 at 12:36:30PM +0100
> Hi,
> 
>  I'm seeing some oddities on two 2.6 boxes here.  I use a Belkin Omni
> 4-way PS/2 KVM switch.  It has a "hot key" to switch boxes (scroll-lock
> twice, then number 1-4).  This has always been a little problematic on
> 2.4 (you have to get the timing of the key presses right, worst case
> when you get to the session either the number has been treated as input,
> or the display has scroll-lock turned on), but usable.

I have a 4 port Vista KVM from Rose Electronics, where hot key is a
single left control. That doesn't interfere as much.

I'm running 2.6.x on 2 machines, windows XP on a third.

> 
>  The duron is now being used to do things in xterms.  From time to time
> the alt key stops being recognised (no alt-tab to switch windows, no
> ctrl-alt-n to switch desktops).  And then after again switching machines
> it suddenly starts working properly again.
> 
>  Any suggestions on where to look, or which parts of my .config would be
> relevant ?
> 
Does your KVM also have keys on the kvm to switch? If so, does the
problem go away if you use those?

Otherwise, perhaps hack drivers/input/keyboard/atkbd.c (dunno if that is
the correct path) to log when a scroll-lock is received) and then throw
away that keypress?

Good luck,
Jurriaan
-- 
"Satire is great, but for Nazis you use baseball bats and broken bottles."
        Woody Allen
Debian (Unstable) GNU/Linux 2.6.5-mm6 2x6062 bogomips 2.86 1.98

Re: KVM issues with recent 2.6 kernels

[Ken Moffat]

On Mon, 19 Apr 2004, Jurriaan wrote:

> Does your KVM also have keys on the kvm to switch? If so, does the
> problem go away if you use those?
>

 I was about to say the problem doesn't appear like that (it just takes
an age to cycle through all 4 boxes), but I thought I'd better try it -
guess what, the alt key is no longer functioning on the duron/2.6.5 box
when I go to it like that.  I've now gone back to it using key presses
and it sorted itself out after I threw in an extra press of the scroll
lock (the xterm was responding, unlike console sessions, it was only the
alt key that seemed to be swallowed).  At least I've now found out a way
to fix the immediate "loss of alt key" problem - thanks !

> Otherwise, perhaps hack drivers/input/keyboard/atkbd.c (dunno if that is
> the correct path) to log when a scroll-lock is received) and then throw
> away that keypress?
>

 Interesting idea, but I've got enough trouble with userspace apps
without hacking the kernel, and anyway /some/ of the scroll-lock
keypresses are needed to get directly to the desired session.

> Good luck,
> Jurriaan
>

-- 
 das eine Mal als Tragödie, das andere Mal als Farce