Cross-over cable and USB-Ethernet adapters

[atreyu]

Howdy,

I've got a head-scratcher here...this post could also go under Networking, but I think it is a hardware issue.

The setup
2 identical computers:

Distro: CentOS 5.7
Kernel: 2.6.18-271.18.1
MoBo: Via Epia CX700
NIC 1: (eth0) on-board Via Rhine III (driver "via-rhine")
NIC 2: (eth1) USB-Ethernet Adapter: ASIX AX88772 (driver "asix")

The goal
I am trying to network the two boxes together using a CAT6 Ethernet cross-over cable. It is a store-bought (QVS.com) cable, marked as a cross-over cable right on it. I can't verify by looking at the wires b/c the RJ45 hood covers them, but I've used it successfully before.

I have disabled eth0 on both machines, and eth1 is configured as 10.1.1.x/24 on each (where x is 1 or 2). there are no routes, nor is iptables configured (or even installed) on either machine.

The problem
As you could guess by now, networking does not work! That is; pinging the one machine from the other fails (Destination Host Unreachable), the same with nmap (Host seems down), ssh times out, etc.

I verify that both machines do detect a link on eth1, via ethtool. also i have link-lights on the NIC LEDs. Also, I verify that:

• both eth1 devices support auto-negotiation
• both eth1 devices have auto-negotiation on
• both eth1 devices are set to the same speed (100Mb/s)
• both eth1 devices are set to the same duplex (Full)

There are no messages in dmesg, /var/log/messages, etc. The eth1 devices on both machines otherwise perform perfectly (if plugged into a hub/switch with a regular ethernet cable, etc.).

If I take down eth1 and use eth0 (on-board NIC) on both machines, and configure the ip the same way, and network them together w/the same x/o cable - networking is fine (ping/ssh/nmap, etc.).

i'm not doing anything tricky here. what the dilly-o? anyone else have issues using cross-over cables w/USB-to-Ethernet adapters?

btw, have not tried mixing one on-board NIC (eth0) with one USB-Eth adapter (eth1) - will try that ... but even if it works, the issue still stands.

also, i do not have any other x/o cables to test with, nor do i have a cable tester. i might have other USB-Ethernet adapters, but not sure about that.

[arespi]

Just a wild guess here ... eth1 being a usb network adapter , i dont know, I have had some bad experiences with them not working rigth in linux. Have you tested eth1 in one server, conected to eth0 in the other?, maybe you can get more info in the logs that way.

[atreyu]

Just a wild guess here ... eth1 being a usb network adapter , i dont know, I have had some bad experiences with them not working rigth in linux.

Do you recall if you were using an adapter based on the ASIX chip? Or do you remember the adapter vendor?

Have you tested eth1 in one server, conected to eth0 in the other?

Yeah, finally got to that test...yes, going from on-board eth0 to usb eth1 does work. i presumed it would, but this does not help my mystery, and no additional info was generated in dmesg/messages.

Thinking that MDI-X might have something to do with it, I've learned that more recent versions of ethtool can report the status of MDI-X on an interface, so I figured that might help me troubleshoot. So I rebuilt the ethtool source RPM from Fedora 16 for CentOS 5 and installed it on one of my CentOS boxes...but it does not report MDI-X status at all. I am guessing it is b/c the driver itself is not supporting it. So perhaps my troubles are stemming from the USB ethernet drivers ("asix" module and related)...I will try these USB adapters on Linux boxes w/more recent kernels and see if that turns up anything different.

[arespi]

Do you recall if you were using an adapter based on the ASIX chip? Or do you remember the adapter vendor?

Not really, but I have tried a bunch of "consumer grade" (read: cheap) usb ethernet and wifi adapters, some didn't work at all, others kind of worked but had this kind of weird issues. My best experience was with broadcoms and some ralinks

Good luck