|
Hi!
On Mon, 23 Jul 2007, Jesse Cantara wrote:
> The device on my machine is the Broadcom BCM5721, and the reason why I
> decided that the driver was at fault is because I found somebody else
> online with the same problem and that particular model of Broadcom NIC.
My fault, I neglected to take into consideration that the tg3
driver supports more than one particular chip(set).
Our machines have BCM5703X GE chips. We also have machines that
have BCM5708 chips, but those are served by another driver
(bnx2).
> Do you have any advice as to what to look for more in-depth?
The question is on which layer the error originates (as opposed
to where you *see* it). This can be as simple as using wireshark
to see the connection break down (can be a hassle if it takes
long to trigger) or be as full-fledged as bringing in a hardware
network analyzer. The order should be obvious :)
> I have replaced some of the switching hardware, but not all of it. The
> reason why I don't think it is the switch or cables is because direct
> system-to-system communication works just fine, it's only when I'm doing
> any sort of packet-forwarding (with lvs-nat or just simple iptables
> port-forwarding).
That surely points to software rather than hardware. Do you have
any funky iptable setups that might interfere? Also you might
want to try to use add-in GE boards but keep everything else the
same. Intel's EEPro1000 might be worth a try - it uses an
entirely different driver, yet it's readily available. That way,
you could rule out both the NIC hw and the driver.
Regards & HTH,
Tobias
--
In the future, everyone will be anonymous for 15 minutes.
|
