Re: keyboard locks in 1.3.9x

Andries.Brouwer@cwi.nl
Wed, 24 Apr 1996 10:25:54 +0200


herf+:
: I have a Quickcam (which takes keyboard power) and a Microsoft Natural
: Keyboard -- no problems on the 1.3.7x kernel I've been running.
: However, upgrading to 1.3.9x shows "Keyboard error" in the logs, at
: which point the machine refuses to respond to any keypresses until a
: reboot.
aeb:
A rare but well-known phenomenon. There is one particular kernel
that gives me this effect. There is no obvious relation to kernel
source, and different people complain about different kernels.
Some hardware matter involving the timing of interrupts, I suppose.
tytso:
It happens to me 50% of the time when I switch CPU's --- I'm sharing one
keyboard across multiple CPU's, using a VGA/Keyboard 4->1 switch box.

Sending an arbitrary command to the keyboard solves the problem.

Is there a way to send an arbitrary command to the keyboard by logging
in over the network and running some user-mode program?

Commands that say something to the keyboard are for example setleds
and kbdrate. So, "setleds -L +num; setleds -L -num; setleds -L",
perhaps given twice, sets the keyboard going again. (Make sure that
stdin refers to the keyboard.)

Or better yet, is there some way we can make the kernel automatically
figure send some sort of reset command when it gets a "Keyboard error"?

Unfortunately, the most common phenomenon is just that the keyboard
dies without giving any message. And conversely, "Keyboard error"
usually just means that you pressed several keys at once.

About this "keyboard death", I have now also heard people complain
about it under OS/2 and under Windows 95, making it more likely
that it is a hardware problem. Still it would be nice if we could
avoid it. (Maybe by having a timer invoke a dummy keyboard command
every 5 seconds? Ugly.)

Andries