Subject: Re: suggested patch to console(4) manpage
To: None <M.Drochner@fz-juelich.de>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: port-i386
Date: 01/19/2000 15:03:14
In message <200001192043.VAA0000016580@zel459.zel.kfa-juelich.de>
Matthias Drochner writes:

>
>david@fundy.ca said:
>> Disable your Intel BIOS console->serial feature. Enable the first
>> serial port via the BIOS. 

Yep.  But that's not a negotiable item.  Thhat setup is *broken*.
It's bad engineering.  That's not just my opinion, its a fact.  
More on that to David.


>So far so good - you get the same functionality as with standard PCs.
>You don't get the early BIOS startup messages, what that Intel
>BIOS could provide you.

No. Wrong.  That is not all the serial BIOS consoles give you.  They
let you you do pretty much *ANYTHING* you could do from a VGA and
101-key console.  See http://www.phoenix.com/platform/serverbios.html.
Poof, no need for KVM switches to manage a rackful of servers.  Just
use a Cyclades, or Rocketport, or what-have-you, with rtty to log
console messages (in case they syslog doesn't get them to disk during
a crash).  I care about this, btw, because I just got burnt by what
turned out to be uncorrectable ECC errors which werent getting logged
before they triggered a panic.


What I would do (for now, until we can find out from Phoenix how to
get at BIOS serial-console state) is turn on DIRECT_SERIAL.  Except
that DIRECT_SERIAL is broken by design, and you're telling me I can't
fix it.

Or, you can keep saying that NetBSD doesn't really give a damn about
this segment of potential users; that intermittently-connected consoles,
(like your irradiated environment), are all NetBSD cares about.

That's is the message I've gotten so far.