Subject: Re: delay before cpu_reset()
To: Hubert Feyrer <feyrer@rfhs8012.fh-regensburg.de>
From: Chris G. Demetriou <cgd@netbsd.org>
List: tech-kern
Date: 10/10/1998 18:17:32
Hubert Feyrer <feyrer@rfhs8012.fh-regensburg.de> writes:
> Maybe making dmesg not restart from scratch on reboot for all machines
> would help here? On some systems (Solaris/sparc, i think even
> NetBSD/sparc), i can see old messages after reboot/panic. I think this is
> implemented by using the last page of memory for the dmesg buffer, and
> unless this memory is cleaned out on reboot, the information should be
> preserved (I don't know what die BIOS' memory test does there...). 

Last i heard, resetting most PCs (for whatever reason, probably BIOS)
clears memory.

I've seen only a few PCs that didn't have this behaviour, but at this
point, they're all so old as to be pretty much useless.  (e.g. one's a
386-20...)


Indeed, having the message buffer stay around between reboots is a
very desirable, and expected, feature, if the machine allows it.  If
the i386 port (or any other port that _may_ support message buffer
permanence) has bugs in its implementation that prevent this from
happening, they should be addressed.  (Last i heard/saw, on some
ports, it just isn't practical to try to keep the message buffer
around, because e.g. another OS is going to run before NetBSD.  those
can't necessarily be helped.)



cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.