Date: Fri, 19 Oct 2007 17:00:38 -0700 From: Julian Elischer <julian@elischer.org> To: Sten Daniel Soersdal <netslists@gmail.com> Cc: Len Gross <sandiegobiker@gmail.com>, John-Mark Gurney <gurney_j@resnet.uoregon.edu>, freebsd-net@freebsd.org Subject: Re: Disable Exponential Backoff (retry) on Ethernet? Message-ID: <47194526.7080200@elischer.org> In-Reply-To: <47193112.4030701@gmail.com> References: <27cb3ada0710172051t536a4d11pfdfdb079ebd98932@mail.gmail.com> <20071018082056.GW39759@funkthat.com> <27cb3ada0710181842x4d214d31ob4f474ff790355b8@mail.gmail.com> <47193112.4030701@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Sten Daniel Soersdal wrote: > Len Gross wrote: >> Thanks so much for the response. Here is some additional information. >> >> I'm trying to emulate an RF network where there are colisions (e.g. >> "Aloha" >> type protocol) so I actually need collisions! I had forgotten that >> modern >> hardware >> essentially eliminated them. So, lets say I can find/use an "old hub", >> can I control the number of retries? Maybe I have to find some old >> NICs and >> old drivers? > > That would be very NIC specific. Retries are done in NIC hardware as far > as i know. > > All you need to do to get some collisions is to set the rates to > half-duplex on both sides (hubs were half-duplex). Hubs with lot's of > traffic between other hosts would definitely produce more collisions. > > Modern hardware didn't eliminate them. Full-Duplex medium did. > There is a broad range of full-duplex RF systems too :) one thing I'd like, that is related, is that if you disconnect a link for 3 hours it shouldn't sit there using 3000 Mbufs in its output queue.. how about discarding them.... currently they sit in the queue forwever as far as I can see. certainly in the em or bce drivers they do.. >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?47194526.7080200>