Date: Mon, 7 Jun 1999 14:22:14 -0600 (MDT) From: "Kenneth D. Merry" <ken@plutotech.com> To: sthaug@nethelp.no Cc: alpha@FreeBSD.ORG Subject: Re: lousy ethernet performance w/21143 Message-ID: <199906072022.OAA24304@panzer.plutotech.com> In-Reply-To: <63479.928784881@verdi.nethelp.no> from "sthaug@nethelp.no" at "Jun 7, 1999 09:48:01 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
sthaug@nethelp.no wrote... > > The machine is on a half-duplex 100BaseT network (i.e. none of the full > > duplex problems with the de driver and the 21143). > > > > With the old kernel, from late April, I can get 7-8MB/sec throughput using > > tcpblast to a machine on the local network. With the new kernel, from > > Saturday, I'm only getting about 6-7 Kilobytes/sec. > ... > > Every outgoing packet seems to generate an error! > > Sounds a lot like duplex mismatch. Have you tried playing with half/full > duplex? Back to back with another machine and a different card (ie. Intel > Pro 100/B)? Well, the duplex settings haven't changed on the box. The only thing that has changed is the kernel. The old kernel works, the new one doesn't work very well. I think I've tracked down the problem, though. I backed out revision 1.104 of if_de.c, and my network throughput is back to normal. The thing I don't understand, though, is why. Theoretically that change was supposed to just read the speed setting out of the SRM. That's fine, I've had ewa0_mode in the SRM set to "Fast" ever since I did the netboot install of the machine last year. In both cases, with and without that change, the de driver prints out: de0: enabling 100baseTX port However, when I had revision 1.104 in there, that message would get printed out along with the main driver probe message: de0: <Digital 21143 Fast Ethernet> irq 0 at device 3.0 on pci0 de0: interrupting at CIA irq 0 de0: DEC 21143 [10-100Mb/s] pass 3.0 de0: address 00:00:f8:75:7c:14 de0: enabling 100baseTX port Without revision 1.104, it would get printed around the time interrupts got enabled. So it was probably autodetecting the speed. It seems to be getting the speed and duplex settings right in both cases, but why, when it reads the value out of the SRM, is the throughput so bad? Ken -- Kenneth Merry ken@plutotech.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199906072022.OAA24304>