Subject: kern/20988: gsip gigabit ethernet driver fails with Netgear GA621
To: None <>
From: Dave Tyson <>
List: netbsd-bugs
Date: 04/03/2003 16:11:08
>Number:         20988
>Category:       kern
>Synopsis:       gsip gigabit ethernet driver fails with Netgear GA621
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Apr 03 07:12:00 PST 2003
>Originator:     Dave Tyson
>Release:        NetBSD current 1.6Q from snapshot 25th March 2003
The University of Liverpool
System:NetBSD 1.6Q (GENERIC) #5: Tue Mar 25 11:14:20 GMT 2003 root@splunge:/usr/src/sys/arch/i386/compile/GENERIC
Architecture: i386
Machine: i386
	The GA621 is a fibre gigabit card using the National Semiconductor
chipset supported by the gsip driver. The driver recognises the card and
configures it thus:

gsip0 at pci2 dev 1 function 0: NatSemi DP83820 Gigabit Ethernet, rev 00
gsip0: interrupting at irq 11
gsip0: Ethernet address 00:40:f4:29:5e:d1
gsip0: 64-bit PCI slot detected, using 64-bit data transfers
gsip0: using ten-bit interface
ukphy0 at gsip0 phy 0: Generic IEEE 802.3u media interface
ukphy0: OUI 0x000000, model 0x0000, rev. 0
ukphy0: 1000baseSX, 1000baseSX-FDX, auto

With the card connected to a cisco gigbit switch and the link configured
to be up the card reports no carrier. The link light on the card appears
to be on implying it can see a carrier from the cisco box. The cisco link
light remains off. This state of affairs remains the same for all the
possible media/mediaopt combinations. e.g

	address: 00:40:f4:29:5e:d1
	media: Ethernet 1000baseSX full-duplex
	status: no carrier
	inet netmask 0xffffff00 broadcast
	inet6 fe80::240:f4ff:fe29:5ed1%gsip0 prefixlen 64 scopeid 0x1
Insert a Netgear GA621 in a system. load NetBSD-current. Observe you
cannot get the link to pass packets
none supplied (sorry).

The card works fine with FreeBSD 5.0 - but we want to use NetBSD!

The problem looks to be with the handling of the PHY - the OUI etc being
reported as all zeros looks suspicious. The interface chip between the NS
controller is made by Aligent (sp?) and appears from the data sheet to be a
fairly dumb convertor which drives the fibre laser - it is not programmable.

Initially I thought I might be able to fix it, but long hours reading the
code and the NS data sheet show I have a bit more to learn on the relationship
between the driver layer and phy layer :-) Help!

Computing Services Dept         Phone/Fax: 0151-794-3731/3759
The University of Liverpool     Email:
Chadwick Tower, Peach Street    WWW:
Liverpool L69 7ZF               Open Source O/S: