On Thu, 26 Jul 2001, Kenneth Stephen wrote: > I didnt know that it wasnt a problem to observe collisions. It > surprised me to see collisions on what is effectively a point-to-point > link. The only two devices on that wire are the card and the cablemodem - > so when I saw collisions, I thought maybe there was some mismatch. Thanks > for the clarification. > > I did not understand your statement about a duplex mismatch. Are > you saying that the autonegotiationwould have ensured a full duplex > connection... Autonegotiation doesn't require full duplex capability (10/100 bridged repeaters report only half duplex capabilities) but standard-conforming full duplex requires autonegotation. > ...and that forcing the card to full_duplex will cause a > problem? Forcing full duplex (which necessarily disables autonegotiation) is a bad practice that causes configuration problems, as you have demonstrated. > Or are you saying that the card may be operating at half-duplex, > but since that is what was negotiated, that is the best setting? If autoneogtiation occured (use 'mii-diag' to find out), half duplex was the best setting. If the link partner doesn't support autonegotation, the device should sense the link speed and use standard Ethernet (half duplex). If there is no link beat, the device should generate 10baseT link beat if it supports autonegotiation. > I am > unable to tell whether the card is operating at full or half duplex - is > there a way I can find this out? Use the device-independent 'mii-diag' program, with 'vortex-diag' providing confirmation by reading the hardware configuration from the chip registers. http://www.scyld.com/diag/index.html Donald Becker becker@scyld.com Scyld Computing Corporation http://www.scyld.com 410 Severn Ave. Suite 210 Second Generation Beowulf Clusters Annapolis MD 21403 410-990-9993 - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org
