Date: Fri, 9 Jul 2010 10:46:58 -0700 From: Jack Vogel <jfvogel@gmail.com> To: Shtorm <admin@shtorm.com> Cc: freebsd-net@freebsd.org, Ian Smith <smithi@nimnet.asn.au> Subject: Re: Intel 82574L Gigabit Ethernet Controller Message-ID: <AANLkTilX7OjU07LWzjGuEVw8XS7V0tKww-H1wgBI0wfT@mail.gmail.com> In-Reply-To: <1278696249.20612.96.camel@stormi-desktop> References: <1278330234.10826.18.camel@stormi-desktop> <AANLkTinePfVRKHdd5FHXEA1YXW9NvDo8mz48hoBgkZ9c@mail.gmail.com> <1278356796.10826.35.camel@stormi-desktop> <AANLkTilVYjuRGtE3gtcRhxIS4FBEAkiBqEPcxCgBh2Hk@mail.gmail.com> <1278404933.20433.26.camel@stormi-desktop> <AANLkTildoiVeA47P47wo027BI5NlppOCNgrV56O-cGaW@mail.gmail.com> <1278496982.21743.50.camel@stormi-desktop> <20100709233505.J54166@sola.nimnet.asn.au> <1278688046.20612.45.camel@stormi-desktop> <AANLkTil-Y-Mozv6q2hOmkl7flsjqSOUkB0f4IYaehV2i@mail.gmail.com> <1278696249.20612.96.camel@stormi-desktop>
next in thread | previous in thread | raw e-mail | index | archive | help
Well, you can always try increasing it, but its unlikely that the problem is the constant, in the older style watchdog code this was set to 5 so I am already being lenient :) Jack On Fri, Jul 9, 2010 at 10:24 AM, Shtorm <admin@shtorm.com> wrote: > Yes, I'm not a kernel hacker, but cat understand what C code says, I see > your point. It will take few days until I can get to server room, will > try without ntpd. > > Also, I have kern.hz=4000 in loader.conf, as far as I understand tick > length will be 25 ms and timeout for watchdog will be 10*25 ms = 250ms > Is it enough for adapter or I need to increase > > #define EM_WATCHDOG (10 * hz) > > I can be wrong with this, just do not have appropriate knowledge. > > Thanks for your help. > > On Fri, 2010-07-09 at 09:26 -0700, Jack Vogel wrote: > > LOL, the way the watchdog code works these days it records the clock > > at key TX points and then compares that in the timer code, so if your > > system is dinking around with the time that could be the cause of this. > > > > Can you change the environment to guarantee a continuous time > > stream and then see what happens?? > > > > Jack > > > > > > On Fri, Jul 9, 2010 at 8:07 AM, Shtorm <admin@shtorm.com> wrote: > > > > > On Fri, 2010-07-09 at 23:58 +1000, Ian Smith wrote: > > > > On Wed, 7 Jul 2010, Shtorm wrote: > > > > > > > > > > > Yow, 30 vlans, but only em1 is using vlans not em0? > > > > > > > > > > > > Is only em1 having watchdogs? I noticed you appear to > > > > > > have flow control off, maybe turning it on would help. > > > > > > > > > > > > I would like to see the log messages from the watchdogs. > > > > > > Jack > > > > > > > > > > Yes, em0 - plain untagged traffic to border router, em1 - tagged - > one > > > > > vlan per 200-300 pppoe clients. Anyway, I saw watchdogs on em0 > too, > > > > > there is no logs for it because remote syslog server connected via > em0 > > > > > and it looses messages during card reset, will enable local logs > to > > > get > > > > > some info. > > > > > > > > > > Log files are almost empty, is there any driver-specific debugging > > > > > options other than TUNABLE_INT("hw.em.sbp", &em_debug_sbp)? Anyway > > > will > > > > > try to set it to 1 and wait for watchdog. > > > > > > > > > > Here is a part from log file I have now > > > > > > > > Deleting the stuff you're most interested in :) > > > > > > > > > Jul 6 10:32:34 ntp info hostname x.x.x.8 ntpd adjusting local > clock > > > by 5.083720s > > > > > Jul 6 10:33:07 ntp info hostname x.x.x.8 ntpd adjusting local > clock > > > by 4.915903s > > > > > Jul 6 10:35:01 auth info hostname x.x.x.8 login login on ttyv2 > as > > > root > > > > > Jul 6 10:35:01 auth notice hostname x.x.x.8 login ROOT LOGIN > (root) > > > ON ttyv2 > > > > > Jul 6 10:35:24 kern crit hostname x.x.x.8 kernel em1: Watchdog > > > timeout -- resetting > > > > [..] > > > > > Jul 6 10:37:21 ntp info hostname x.x.x.8 ntpd adjusting local > clock > > > by 3.641940s > > > > > Jul 6 10:37:46 kern crit hostname x.x.x.8 kernel em1: Watchdog > > > timeout -- resetting > > > > [..] > > > > > Jul 6 10:38:40 kern crit hostname x.x.x.8 kernel Limiting icmp > > > unreach > > > > > response from 237 to 200 packets/sec > > > > > Jul 6 10:39:10 kern crit hostname x.x.x.8 kernel em1: Watchdog > > > timeout > -- resetting > > > > > > > > Probably completely unrelated, but I can't help noticing those big > clock > > > > shifts by ntp over a short period amidst all this. I don't know if > that > > > > could affect watchdogs, but is it a regular occurrence during these? > > > > > > > > >From your latest, a bit more noise from ntp: > > > > > > > > > Jul 8 07:23:40 <kern.crit> server kernel: em0: Watchdog timeout > -- > > > resetting > > > > [..] > > > > > Jul 8 07:23:56 <ntp.crit> server ntpd[3687]: 2 out of 3 peers > valid > > > > > Jul 8 07:23:56 <ntp.crit> server ntpd[3687]: bad peer from pool > > > pool.ntp.org (195.214.215.17) > > > > > Jul 8 07:27:15 <kern.crit> server kernel: em0: Watchdog timeout > -- > > > resetting > > > > > > > > Ignore if not relevant. > > > > > > > > cheers, Ian > > > > _______________________________________________ > > > > freebsd-net@freebsd.org mailing list > > > > http://lists.freebsd.org/mailman/listinfo/freebsd-net > > > > To unsubscribe, send any mail to " > freebsd-net-unsubscribe@freebsd.org" > > > > > > Yeah, saw this too, it was first boot for this install and I forgot to > > > run tzsetup during flash image build. > > > > > > As for the latest log, this box connected to internet via em0, ntpd > just > > > says it have some peers to sync with after interface flap. > > > > > > Thanks. > > > > > > > > > > > _______________________________________________ > > freebsd-net@freebsd.org mailing list > > http://lists.freebsd.org/mailman/listinfo/freebsd-net > > To unsubscribe, send any mail to "freebsd-net-unsubscribe@freebsd.org" > > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTilX7OjU07LWzjGuEVw8XS7V0tKww-H1wgBI0wfT>