Loss of ethernet network capability on XP1000 Tru64 Unix v5+

[Darrin Wahl]

I'm a new member on this forum. I have inherited sys admin responsibility for an XP1000 running tru64 Unix v5+, although I really only have Unix user experience.

The machine has been up and running on our LAN for years, but suddenly became unable to see the LAN or have any other machine see it on the network. Ping, traceroute, netstat, etc. all simply hang with no feedback when run on the machine looking for an external IP address, or when run from another host looking for the machine's IP address.

I can only log on as root (user accounts/passwd are obtained via yp/NFS), and the boot process and login are *very* slow, but there are no error messages in /var/adm/messages. I've rebooted several times with no effect.

ifconfig indicates that the tu0 (ethernet) interface is "up" running 100baseT, full duplex. I tried taking tu0 down and up with ifconfig, restarting the inetd process, and tried to force various configurations using lan_config in the etc/inet.local file. Our local network admin even went to the LAN switch and was able to see a link to the machine via a networking diagnositc, but could not ping or traceroute the machine.

Does anyone have any ideas about what could be the problem? Do I have to procure another NIC card (the original connector is part of the motherboard)? If so, what device should I get and what drivers are needed?

Thanks for any help you can provide.

Darrin

[Ann Majeske]

I'll leave it to others to help with your main problem, but as regards to login being very slow that is expected if your system can't see the NIS server. Even for root, the system has to check for NIS groups and wait for the NIS timeout before completing the login. To work around this while you're debugging the other issue disable NIS temporarily until you get your network back.

Ann

[Steven Schweda]

> [...] what could be the problem?

Bad hardware? Could be the cable, or the box
at either end. Have you tried any of the
obvious things, like a different cable or a
different port on the hub/switch?

If the stuff on the main board is really
fried, and if you have a free PCI slot, then
adding a card would probably be easier (and
cheaper) than getting it replaced/repaired.

I assume that some official document would
list the supported cards, but I'd also assume
that something like a DE500-BA would be
roughly equivalent to the built-in gizmo,
which appears in the SRM console "show
device" report like this:

Slot Option Hose 1, Bus 0, PCI
3 DECchip 21143-AA ewa0.0.0.3.1 00-00-F8-71-D4-CD
[...]

(Yours does still appear there, right?)

Alternatively, and probably available for
less than $10 at your local PC junk store, a
card with an Intel 82558 chip should also be
suitable. For example:

Slot Option Hose 1, Bus 2, PCI
[...]
10 Intel 8255x Ethernet eia0.0.0.2010.1 00-50-8B-11-3A-D5

The Intel chip/card would appear as "ee0"
instead of "tu0", so some IP reconfiguration
would be needed, but that would be true for a
"tu1", too, I'd guess.

While you're talking to the SRM console, try:
show ewa0*

Are things like "ewa0_mode" reasonable?

With Tru64 running, what do you get from:
hwmgr get attribute -category network
?

(Like "link = Down", say?)

[Darrin Wahl]

Thanks for the replies. I was pulled off to work on a proposal, and I'm just getting back to this.

Summary: The Ethernet interface just started working again after several reboots, so the problem is gone but the reason is unresolved.

To follow up your advice/help:

> Have you tried any of the
> obvious things, like a different cable or > a different port on the hub/switch?

Yes, I tried switching cables, ports on the router, and even a different router to no avail.

> (Yours does still appear there, right?)
> While you're talking to the SRM console, > try:
> show ewa0*
> Are things like "ewa0_mode" reasonable?
> :
> With Tru64 running, what do you get from:
> hwmgr get attribute -category network
> (Like "link = Down", say?)

I'm not very familiar with this machine or tru64 unix, so I am not sure what to expect from SRM or hwmgr. However, ewa0 did show up as DECchip 21143 at MAC 00-00-F8-71-5C-9E. None of the IP addresses were populated (ie 0.0.0.0).

For hwmgr, there was not "link=XXX" output, but "mode=Fast Full Duplex" and "state=available".

[Darrin Wahl]

Problem disappeared on its own after several additional reboots. It appears to be a hardware problem with the Ethernet adapter, so I may resort to installing a second one on another PCI slot.