On Wed, 20 Jun 2001, Charles Tassell wrote: > I'm having some strange problems with two different systems running the > 2.4 kernel (2.4.3, 4, and 5 tested) with two VIA Rhine cards (D-Link 530TX > card I think, they are definately the VIA VT3043 Rhine chipset though). > > The problem is that every once in a while, the network will just drop, no > traffic will flow for a few seconds/minutes, nothing shows up in the logs > (*.* /var/log/everything in syslog) and then it will suddenly correct > itself. Sometimes doing a ping from the machine to an outside host will > bring everything back up immediately. Do you get a kernel log message about "Transmit timed out" when this happens? Oh, no, you don't ... (not even a tiny little something in 'dmesg'? :) The driver only reacts to interrupts, one per received packet. If one interrupt is missed I guess it would just sit there. A ping where the interrupt isn't lost makes it find the queued packets. My only problem with that idea is: does your machine sometimes have minutes between something being sent? What motherboards (w/ chipset if you know) do the systems have? > Now, I have another machine that has one VIA Rhine card (same D-Link) and > a tulip card (D-Link 528 or 538 I think, Digital DC21041 Tulip rev 33) > where this problem does not occur. On the same network, I assume. What happens if you switch one of the via-rhine cards for the tulip card? > Any ideas? Should I try using the original Donald Becker driver at > http://www.scyld.com/network/via-rhine.html or is Jeff's modified version > that is in the current source tree considered to be more current? They are relatively close (that's the idea anyway). But if you compiled the kernel tree version as a module, switching to Donald's version shouldn't be too difficult. It's certainly worth trying. /Urban - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org
