bizarre networking issue... *SOLVED!*

[XARiUS]

Okay, challenge of the week(end).

Switch: Linksys SD2005 5 port gigabit unmanaged switch
XP Box: NForce4 based onboard Gigabit LAN, latest drivers.
Linux Box: (Debian, Ubuntu, any flavor really) Onboard 10/100
Cisco 7960G Phone: 10/100

All devices use DHCP from WRT54GS - 255.255.255.0 netmask, 192.168.0.1 gateway, 192.168.0.100 - 150 DHCP range.

WRT54GS --> SD2005 (Switch) --> Devices

Scenario:

Windows XP box can ping phone and WRT54GS router just fine. Cannot ping the linux box, ssh to it, telnet, notta. Windows ping returns request timeouts.

Linux box can ping phone and WRT54GS router just fine. Cannot ping XP box. Linux ping returns destination host unreachable.

After pinging all devices from both systems, and doing an arp -a to check out the existing arp cache, I notice that in both systems, the phone entry is correct however:

Linux reports my XP box as: xp-box (192.168.0.142) at <incomplete> on etho0

XP reports my Linux box as: 192.168.0.141 00-00-00-00-00-00 invalid

So, neither the XP box or Linux box seem to be able to read each other's hardware address. They both do infact both have valid hardware addresses.

Additionally, the routing tables for both systems seem correct. We're talking a dirt simple 192.168.0.0 255.255.255.0 network here.

I'm really stumped at this point. The only thing I have not tried is using a different switch, mainly because my checking account is not at liberty to make such a purchase until later in the week lol.

Anyone have any clues?

X

 

[petey117]

wow...so they can see the layer 3 address, but not the layer 2 address....
i feel like the answer is staring me right in the face....but i just can;t think of it
try a crossover between the linux and XP box (with statically assigned addresses) and see if that works
this problem seems charachteristic of an ip address conflict (devices sharing the same ip)

 

[RebateMonger]

Another NForce 4 chipset?
Maybe the same problem as in this post, with defective NForce NIC drivers for Windows. Apparently the "latest" NForce drivers don't work properly under XP SP2.

 

[petey117]

i wish nvidia would get their s*it together. some of the buggiest drivers i have seen to date.
great chipset. bad drivers
although my asus k8n-SLI premium has yet to crash on me - maybe the motherboard makers are to blame

 

[RebateMonger]

I guess the OP can't just change to NetBEUI protocol, huh? (this is intended to be humerous, folks.....!)

 

[XARiUS]

Originally posted by: RebateMonger
I guess the OP can't just change to NetBEUI protocol, huh?

After reading the thread referenced above, LOL @ that!

Okay so I just went and tried every NV driver revision back to like March of 2004.. neither made a bit of difference, sadly.

I suppose I'll try the crossover cable.

Thanks for the responses guys!

X

 

[RebateMonger]

Yeah, start simple, with a cable, and verify basic TCP/IP communication. It would be nice if you had a third computer, so you could see which of your two computers is actually malfunctioning (if it's a computer problem).

If they were both gigabit NICs, you probably wouldn't even need a crossover cable. Some of the gigabit NICs (at least the Intel ones) are auto-sensing and can use standard cables. Also, can't you just hook the PCs up to your router and dump the switch, temporarily?

Apparently gigabit connections use all eight lines of the connection. Be sure that you have good cables.

Edit:
I'm assuming you know that the default Windows XP SP2 firewall won't allow incoming pings?

Also, under the "Windows" convention, a message "Destination host unreachable" would mean that the computer has no route or default gateway to send packets to the destination network. Even if an IP address didn't actually exist, all you'd get would be timeouts. You WOULDN'T get a "Destination host unreachable" message. I just tried it with XP....it just times out, with no error message. I don't know how Linux is supposed to behave.

 

[JackMDS]

Being MDX is part of the Giga card standard No need for crossover.

:sun:

I like NetBEUI.:thumbsup:

Did you know that NetBEUI is GUI to NetBIOS?:shocked:

 

[Pabster]

Originally posted by: JackMDS
I like NetBEUI.:thumbsup:

Did you know that NetBEUI is GUI to NetBIOS?:shocked:

:thumbsup: NetBEUI rocks.

 

[nweaver]

iptables -F to flush the linux firewall?

 

[XARiUS]

Well I'm really sorta stumped on this one.

The only thing I can think of is there is some sort of extreme incompatibility with this nforce gigabit network adapter.

I've tried a couple different systems now, all of them can see my *nix box just fine. However this XP box (nForce chipset) can't ping that linux box and some of the OTHER PC's I tested can't see this XP box...

It's gotta be something with this onboard gigabit nic.

Thanks for the feedback guys, I did try a crossover (and straight) cable, but had no luck. Also tried a different switch which also provided no insight.

 

[XARiUS]

Okay, update.. FIXED!

802.1p

An IEEE standard for providing quality of service (QoS) in 802-based networks. 802.1p uses three bits (defined in 802.1q) to allow switches to reorder packets based on priority level. It also defines the Generic Attributes Registration Protocol (GARP) and the GARP VLAN Registration Protocol (GVRP). GARP lets client stations request membership in a multicast domain, and GVRP lets them register into a VLAN.

Having 802.1p enabled on the gigabit nic was basically causing all sorts of strange havoc, and explains why we were seeing the weird layer2/layer3 issues. 802.1p occurs at the hardware address level.

FWIW I did have VLAN'ing disabled on the NIC. I need to do some more reading as to how VLAN interoperates with 802.1p/GARP....

Everyone write 802.1p on a post-it for future reference, I know I will!

X