Index of /pub/FreeBSD/FreeBSD-current/xperimnt/hp100vg

 Name                    Last modified       Size  Description

 Parent Directory        19-Nov-2004 09:58      -  
 hp100vg-sysdiff.gz      15-Oct-1997 12:44    18k  

This is a snapshot of a low-priority `work-in-progress' project of mine.
Albeit HP's 100VG (100 Mbit/s `voice-grade' Ethernet) technology didn't
win in the end, there are still a number of cards around for it, some
of them 10 MBit/s only, some both 10 Mbit/s and 100 MBit/s VG.

This is a very early snapshot, not meant as any form of a completed
source.  However, it's simply not sure when i'll find more time to
complete it, so if anybody feels compelled by this source, feel
free to contact me, and complete it.  Alas, HP itself is not helpful
at all in providing documentation, so they are certainly one of the
reasons for why this project didn't get into better shape yet.  The
author of the Linux driver, Jaroslav Kysela <perex@pf.jcu.cz>, allowed
us to redistribute our driver under a BSD copyright, but it was one of
his early works so he couldn't help me much either.  The co-author of
the Linux driver (who implemented the busmaster DMA code there), Sieg-
fried `Frieder' Loeffler <floeff@tunix.mathematik.uni-stuttgart.de>,
was very helpful with hints and even a loaner EISA board for testing.
Thanks to both these guys here.

The PCI driver basically works; i'm currently typing this on a machine
connected to the Internet with a J2970 (10 Mbit/s) card.  However,
there's a still unexplained phenomenon where for about half of the
received Ethernet frames, the card doesn't prepend the status and
length information word to the packet as it ought to do.  The current
implementation of the driver has a gross hack to detect this situation,
and obtain the packet length from another register.  This seems to work
well enough to allow for a telnet connection, and some light FTP traffic.
I've also observed a number of panics in ether_output() when attempting
to do heavy NFS traffic though.  All these tests were done solely on
FreeBSD 3.0-current systems of several vintages.  Anyway, as of finishing
this writing, i'm typing it using the driver on a 2.2.5-BETA system, as
a proof of function.

There's something wrong with the EISA interface still.  The driver
doesn't send anything useful, but i can already run tcpdump on it,
and see the packets on the wire.

The ISA interface hasn't been written at all, but it should not be
too difficult to derive it from the PCI and EISA wrappers.

In order to use the driver (on PCI), just do

	cd /usr/src
	mkdir sys/dev/hp100vg
	zcat /cdrom/xperimnt/hp100vg/hp100vg-sysdiff.gz | patch

...and add

	device	hp0

to your kernel config file.

Feedback is welcome, as well as volunteers to complete the task.


Dresden, October 15, 1997		Jörg Wunsch
					<joerg@FreeBSD.org>