On Wed, Dec 19, 2007 at 06:00:00PM +0000, sborrill@precedence.co.uk wrote:
> >Number:         37569
> >Category:       kern
> >Synopsis:       On a machine with USB keyboard only, the kernel hangs on entry for a considerable time
> >Confidential:   no
> >Severity:       serious
> >Priority:       medium
> >Responsible:    kern-bug-people
> >State:          open
> >Class:          sw-bug
> >Submitter-Id:   net
> >Arrival-Date:   Wed Dec 19 18:00:00 +0000 2007
> >Originator:     Stephen Borrill
> >Release:        NetBSD 4.99.42
> >Organization:
> >Environment:
> 	
> 	
> System: NetBSD 4.99.42 (GENERIC) #4: Tue Dec 18 14:36:45 GMT 2007
> Architecture: i386
> Machine: i386
> >Description:
> On an HP ML110 boot a GENERIC kernel. There is a very long pause (5 mins?) before
> any kernel text is printed, a similar pause after pckbd0 as isa0 is probed and also when
> shutting down.
> 
> The ML110 is odd in that it appears to have  PS/2 ports, but in fact these are connected
> to an internal PS/2 to USB convertor:
> 
> uhidev0 at uhub3 port 1 configuration 1 interface 0
> uhidev0: ServerEngines SE USB Device, rev 1.10/0.01, addr 2, iclass 3/1
> ukbd0 at uhidev0
> wskbd1 at ukbd0 mux 1
> wskbd1: connecting to wsdisplay0
> uhidev1 at uhub3 port 1 configuration 1 interface 1
> uhidev1: ServerEngines SE USB Device, rev 1.10/0.01, addr 2, iclass 3/1
> ums0 at uhidev1: 8 buttons and Z dir.
> wsmouse0 at ums0 mux 0
> >How-To-Repeat:
> Boot GENERIC i386 or amd64 on a HP ML110.
> >Fix:
> Options:
> 1) put "no pckbd" in your kernel config
> 2) Set a shorter timeout in dev/ic/pckbc.c
> 
> --- sys/dev/ic/pckbc.c.orig	2007-12-19 17:56:16.000000000 +0000
> +++ sys/dev/ic/pckbc.c	2007-12-19 17:57:40.000000000 +0000
> @@ -134,6 +134,11 @@
>   * in some system configurations.
>   * This is not canonical way to handle polling input.
>   */
> + 
> +#ifndef PCKBC_POLLTIME
> +#define PCKBC_POLLTIME 100
> +#endif
> +
>  int
>  pckbc_poll_data1(pt, slot)
>  	void *pt;
> @@ -143,7 +148,7 @@
>  	struct pckbc_slotdata *q = t->t_slotdata[slot];
>  	int s;
>  	u_char stat, c;
> -	int i = 100000; /* if 1 port read takes 1us (?), this polls for 100ms */
> +	int i = PCKBC_POLLTIME;
>  	int checkaux = t->t_haveaux;
>  
>  	s = splhigh();

Won't this cause problem on other hardware ? It may be better to use delay()
here, if it's possible at this time.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--