Subject: Re: Shark flashing (oops!)
To: Hans Ridder <hans.ridder@veritas.com>
From: Chris G. Demetriou <cgd@netbsd.org>
List: port-arm32
Date: 06/11/1999 17:31:26
Hans Ridder <hans.ridder@veritas.com> writes:
> I think I just blew away the brains in a Shark. I got the new firmware
> from Chris D. and before I knew what I was doing I typed "net-flash" and
> it loaded and flashed the netbsd kernel. I guess it asked my dhcp server
> what to load and was told to load the kernel. A brief look at the
> net-flash command suggested it was going to ask for a rom image file,
> but apparently my Forth understanding needs work.

*sigh*  no, your reading of the Forth was correct.  This is a 'known
issue' that I probably should have warned you about.  Sorry.  (In my
defense, for some reason I thought that the DHCP server that was known
to have this problem was fixed...)

What's happening is that the firmware is asking for the ROM image file
name, and instead of replying with that filename (as is reasonable
and, IIRC, specified by the RFC), your DHCP server is returning the
filename that it likes.

In a nutshell, the DHCP server is broken.  If it's the ISC DHCP server
shipped with 1.3.x, yes, this is a known problem.  (I thought it was
fixed in the one in 1.4, but obviously i could be mistaken.)  A
workaround is to remove the 'filename' specifier from the Shark's
entry in the dhcpd.conf, while upgrading flash images.

There's a quick hack (which might have shown up in the latest FW
image) to sanity check the image before it's burned into the EEPROM.
If the firmware doesn't have it, it can be placed into the nvramrc...


> Anyway, I'm posting to help other avoid the same mistake, and in the
> hope that someone knows of a way to fix this. My fear is that it has to
> be done with a JTAG probe, or similar specialized hardware that only the
> "factory" has. But my hope is that there is a "fail safe" flash loader
> built in there that I can somehow access (serial port?)

There's no failsafe loader built in, but the flash EEPROM is a
standard and socketed part, so you can yank it out and pop it into an
EEPROM burner if you have or have access to one.

If you neither have nor have access to one, well, i'm sure somebody on
this list might be able to help you out.  (Unfortunately, I don't have
or currently have access to one.)


> Any help is appreciated. I'd hate to see a nice box like this be
> relegated to the trash heap because of stupid mistake.

It's not nearly this bad.  You'd be surprised to know how many times
we made this same mistake ourselves.  8-)



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