Date: Wed, 29 May 2002 11:17:07 +0800 From: Eugene Grosbein <eugen@www.svzserv.kemerovo.su> To: stable@freebsd.org Subject: DEVICE_POLLING problem with 4.5-STABLE Message-ID: <20020529111706.A63231@svzserv.kemerovo.su>
next in thread | raw e-mail | index | archive | help
Hi! I tried to use DEVICE_POLLING for my router. So I rebuilt a kernel and added kern.polling.enable=1 to /etc/sysctl.conf My system has two fxp and two ed network interfaces acting as router/shaper, it heavily uses dummynet/WF2Q making lots of low-bandwidth pipes (100Kbit/s and less) with short queue lenght (5 slots). When I created 200Kbit/s flow from the segment served by ed1 to the segment served by fxp0, the system became practically unresponsible and I got the load of kernel messages written to console and syslog: May 28 21:27:50 gw2 /kernel: poll stalled [7] in phase 4 May 28 21:32:47 gw2 /kernel: poll stalled [8] in phase 4 May 28 21:32:47 gw2 /kernel: poll stalled [9] in phase 4 May 28 21:32:53 gw2 /kernel: poll stalled [10] in phase 4 May 28 21:32:53 gw2 /kernel: poll stalled [11] in phase 4 .... May 29 10:38:56 gw2 /kernel: poll stalled [1651] in phase 4 And so on. Sometimes there was 'phase 6'. It was not possible to manage it using ssh and console's respose time was unacceptable: it took about a minute to login. However, I could successfully (ehem) reboot router using C-A-D, booted it to the single mode, disabled polling and repeated the experiment. No lockups, no messages, 75% idle time, 190Kbit/s of FTP transfer. I know ed(4) doesn't currently support polling. Does it mean I cannot use polling here? My system is Intel P75, both fxp run in 100baseTX <fullduplex> mode. Eugene Grosbein To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020529111706.A63231>