Path: utzoo!utgpu!attcan!uunet!tektronix!tekgen!tekigm2!neals
From: neals@tekigm2.TEK.COM (Neal Sedell)
Newsgroups: comp.sys.m6809
Subject: Re: ARTICLE
Message-ID: <3783@tekigm2.TEK.COM>
Date: 11 Nov 88 01:34:28 GMT
References: <8811071544.AA29817@decwrl.dec.com> <16634@agate.BERKELEY.EDU>
Reply-To: neals@tekigm2.TEK.COM (Neal Sedell)
Distribution: na
Organization: Tektronix, Inc., Beaverton,  OR.
Lines: 38

In article <16634@agate.BERKELEY.EDU> koonce@brahms.berkeley.edu (tim koonce) writes:
>You've encountered the now-famous Boot List Order Bug (BLOB).  It has
>been partly diagnosed as hardware timing problems.  Basically, most
>coco peripherals weren't really well designed for the 1.8 meg clock
>speed, so there's marginal bus timing for many things.

Excuse me if I'm wrong, but hasn't the BLOB been around as long as OS9???
I've had BLOB problems with my trusty old COCO II w/Hard Drive Spec. floppy
controller (unless is was all a bad dream :-) years ago.  I don't eat and
sleep OS9, but I have messed with several floppy and SASI (that's no typo,
unfortunately) interfaces on several different systems.

Having said that, I find it EXTREMELY unlikely that it is a hardware
problem....  The odds against reading a sector from a floppy (or hard)
disk without a CRC error are so low it virtually isn't possible.  And why
oh why would the order the modules are read in matter????  They still
contain the exact same data.  I can imagine no possible "memory" mechanism
that would be sensitive to the order of the boot file, unless it is sensitive
to the destination address in RAM the data is written to, and since the 6809
doesn't MUX the address and data together it seems not too likely.  Consider
also that for a disk read operation the sloppy E timing would contribute
to an over-long read of the controller data register.  That would only help the
data hold timing, and not contribute to an error between the controller and
CPU.  Now, maybe it's the write operations to the controller.  Perhaps the
wrong register gets a short write glitch on the way to addressing the right
register - and maybe the wrong sector gets read....  But then the OS9
CRC would be off.  What does OS9 do when it tries to boot and the necessary
system modules are corrupt???  Surely it doesn't try to execute them anyway.

So we have a situation where we can't get bad data off of a disk sector
to start with, and even if we did OS9 won't try to run it unless it passes
another even more rigorous validity check.  Sure sounds like a software bug
to me.

Then again, until the BLOB is unmasked there is always the chance that
I am wrong.
-- 
Neal Sedell