Subject: port-i386/12407: timeout updating leds (pccons/wscons unusable)
To: None <gnats-bugs@gnats.netbsd.org>
From: None <phil@cs.wwu.edu>
List: netbsd-bugs
Date: 03/14/2001 15:17:47
>Number:         12407
>Category:       port-i386
>Synopsis:       timeout updating leds (pccons/wscons unusable)
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-i386-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Mar 13 23:20:01 PST 2001
>Closed-Date:
>Last-Modified:
>Originator:     Phil Nelson
>Release:        NetBSD-1.5
>Organization:
NONE
>Environment:
System: NetBSD grayling 1.5 NetBSD 1.5 (GRAYLING) #8: Wed Mar 7 18:27:50 PST 2001 phil@grayling:/home/phil/netbsd-1.5/src/sys/arch/i386/compile/GRAYLING i386

>Description:
	Both pccons and wscons have problems updating the keyboard
leds on a MSI K7T Pro 2-a motherboard which uses the VIA KT133 chipset
and using a 800Mhz Duron processor.  On pccons, it is just a bother to
not have the lights and see that message all the time. On wscons, the
results are much more severe.  Initial boot to multi-user appears OK
until one tries to login and after the first key is hit, it repeats
until the keyboard is unplugged.  After plugging it back in, it works
until the caps-lock key is pressed, at which time the keboard quits
working.  (IIRC, it also gives a message about the leds.)  The only
way I could get a usable system was to go back to pccons.

This was tried on several motherboards of the same kind and it appears
to be consistent.  Windows 98 works correctly so dealer insists it is
a software problem.  Keyboard diagnostics also say keyboard is working
completely.

>How-To-Repeat:
	Boot on one of those motherboards.
>Fix:
	No clue yet.  I tried the fix listed in an OLD e-mail related
to PR port-i386/372 saying to add (void) inb(0x84) just before
the final } in the for(i=100000...) loops in kbc_cmd.  It didn't work.
(Email Date 07/31/1994 04:58:18 by mycroft@gnu.ai.mit.edu.)

I'm wondering if there shouldn't be a better method of timing in 
kbc_cmd that doesn't depend on loops.


>Release-Note:
>Audit-Trail:
>Unformatted: