Re: Network instability with jumbo frames on a G5
Ken Wellsch via freebsd-ppc <[email protected]>
| Newsgroups | gmane.os.freebsd.devel.ppc |
|---|---|
| Message-ID | <[email protected]> |
First off, thank you Julio for posting about your experience installing with a reduced amount of memory. That got me to dig out my 2.3 DC box to try again. Last year after initial success with 12-stable, things went down hill for me with later 12-stable images and I put the system aside. I found the current 12-stable installed fine for me a month back and it continues to work well. I tried 13-stable and 13-release but see the USB hang during boot consistently I saw last year with later 12 copies. I did not find that switching from 8G memory to 2G made any difference with my usb boot hangs. Also thank you for your bug posting about the 12-stable tool chain snafu. I was able to pull the necessary libraries and static binutils items from an older 12 release, to then build current 12-stable src using your suggested patch. FWIW it took just over 3 hours to buildworld && buildkernel. I saw your jumbo frames posting and thought I would give it a try. I was disappointed with the bge nics as they would not auto negotiate to 1000baseT (only 100baseT) on my modest GS308 switch. So after experimenting with several modern Intel nics I found one that works well for me. I tried an i340 and a VT I believe, but both would quickly come to a complete stop and hang. So I tried a PT and found that it works like a champ. After researching I see that the PT chip is PCI 1.0 which is likely a good thing on the G5. Anyway, here are the snippits: em0: <Intel(R) PRO/1000 Network Connection> port 0x20-0x3f mem 0x80160000-0x8017ffff,0x80140000-0x8015ffff irq 62 at device 0.0 on pci7 em0: Using 1024 TX descriptors and 1024 RX descriptors em0: Using an MSI interrupt # ifconfig em0 em0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 9000 I was able to shamelessly switch to an mtu of 9000 over an ssh session that survived without being dropped. I ran a couple gitup's on stable & ports without incident and also dumped a few large text files on the ssh connection without any glitches. While I was poking around on the web I did see a note saying "best not to use jumbo frames at 10 or 100 as they are likely to have problems..." ... I wonder if you are experiencing the same 100 auto negotiation as I did with the bge's? _______________________________________________ [email protected] mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-ppc To unsubscribe, send any mail to "[email protected]"