Date: Mon, 11 Aug 1997 23:21:13 -0700 (PDT) From: Vinay Bannai <vinay@agni.nuko.com> To: tom@sdf.com (Tom Samplonius) Cc: black@zen.cypher.net, dg@root.com, hackers@FreeBSD.ORG Subject: Re: fxp driver full duplex packet loss problem Message-ID: <199708120621.XAA06902@agni.nuko.com> In-Reply-To: <Pine.BSF.3.95q.970811173800.8572B-100000@misery.sdf.com> from "Tom Samplonius" at "Aug 11, 97 05:39:29 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
According to Tom Samplonius: > > On Mon, 11 Aug 1997, Ben Black wrote: > > > a good indication of this duplex mismatch is a sudden appearance of FCS > > errors and Runts on the cisco interface. > > When I switch full duplex on the FreeBSD server, the switch starts > recording late collision errors on that port (a 10BT port on a Cat 1900). > I'm not sure what this means. > > Tom Late collision errors typically when the ethernet collision domain distance limitations are violated. Typically the collision window for a transmitting node is 512 bit times. This also is the minimum ethernet packet size (64 bytes). This also imposes a limit on the maximum diamter of the ethernet collision domain. Thus the furthest distance between any two nodes in the LAN should not be more than 512 bit times. This basically means if you (transmitting node) don't see a collision in the first 512 bit times then you basically acquired the ethernet and can be assured that you won't see a collision. In case you see a collision after the 512 bit times than you got yourself a late collision. For a full duplex connection there are NO collisions and there should not be any late collisions. So if the switch says there are late collisions that means that the switch is basically in some sort of screwed up half-duplex mode for the port. Maybe the NIC card thinks it is in full duplex and the switch is in half duplex. This also means that the auto-negotiation is not working properly between the switch and the NIC. I would presume the NIC would be at fault. I say that because if the NIC gets any garbled response from the switch it should automatically fall back in the half duplex mode. Or the switch might be saying it supports full dup but does not do it. In that case I would say it would have to be a bug in the switch. Vinay -- Vinay Bannai E-mail: vinay@agni.nuko.com (408)-526-0280 x 275 (Work) http://agni.nuko.com/~vinay
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199708120621.XAA06902>