Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 2 Jan 2018 16:45:26 +0000
From:      Gary Palmer <gpalmer@freebsd.org>
To:        Konstantin Belousov <kostikbel@gmail.com>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: Odd behaviour on recent boot of 11.1 with timecounters
Message-ID:  <20180102164526.GA63396@in-addr.com>
In-Reply-To: <20171231164738.GY1684@kib.kiev.ua>
References:  <20171231141708.GA2629@in-addr.com> <20171231145147.GW1684@kib.kiev.ua> <20171231154913.GA63310@in-addr.com> <20171231164738.GY1684@kib.kiev.ua>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Dec 31, 2017 at 06:47:38PM +0200, Konstantin Belousov wrote:
> On Sun, Dec 31, 2017 at 03:49:13PM +0000, Gary Palmer wrote:
> > On Sun, Dec 31, 2017 at 04:51:47PM +0200, Konstantin Belousov wrote:
> > > On Sun, Dec 31, 2017 at 02:17:08PM +0000, Gary Palmer wrote:
> > > > 
> > > > Hi,
> > > > 
> > > > I recently updated to 11.1-RELEASE-p6 and on the most recent reboot 
> > > > (after rebuilding all the necessary packages) the clock was running 
> > > > slow and NTP wouldn't sync.  I looked in /var/log/messages and I found
> > > > that for some reason, on this latest boot, it got the frequency of
> > > > TSC-low wrong.
> > > > 
> > > > Aug 24 04:55:35 my kernel: Timecounter "TSC-low" frequency 1746073190 Hz quality 1000
> > > > Aug 26 03:11:38 my kernel: Timecounter "TSC-low" frequency 1746070760 Hz quality 1000
> > > > Aug 26 14:12:46 my kernel: Timecounter "TSC-low" frequency 1746075204 Hz quality 1000
> > > > Nov 19 16:01:09 my kernel: Timecounter "TSC-low" frequency 1746070746 Hz quality 1000
> > > > Dec 27 22:28:00 my kernel: Timecounter "TSC-low" frequency 1746074808 Hz quality 1000
> > > > Dec 27 22:51:12 my kernel: Timecounter "TSC-low" frequency 1746071892 Hz quality 1000
> > > > Dec 28 12:50:46 my kernel: Timecounter "TSC-low" frequency 1746069704 Hz quality 1000
> > > > Dec 28 14:03:52 my kernel: Timecounter "TSC-low" frequency 1937876448 Hz quality 1000
> > > > 
> > > > Until the December reboots the machine was running 10.x.  Dec 27 and later
> > > > are part of the process to get up to 11.x.
> > > > 
> > > > Any idea why the TSC-low frequency jumped 191,806,744Hz on the last
> > > > measurement?
> > > > 
> > > > I switched to HPET temporarily via sysctl and ntp seems happy.  I'm just
> > > > concerned that the problem might recur on later reboots as TSC-low seems
> > > > to be the preferred timecounter.
> > > 
> > > Show first 100 lines of the dmesg from a verbose boot.
> > > Also check BIOS settings related to overclocking and powersaving.
> > > 
> > 
> > Hi Konstantin,
> > 
> > BIOS settings haven't been changed in 4+ years.  No overclocking, and
> > all powersaving options are at "auto" or "disabled".
> > 
> > The first 100 lines of verbose dmesg didn't seem that interesting so
> > I've included up to the end of "Device configuration finished"
> > 
> > Note that this boot didn't have the TSC-low problem, and the boot
> > that had it wasn't verbose unfortunately.
> 
> It is really the CPU identification which I wanted to see.  You have
> IvyBridge, which is known to have good TSC.

Ah

> Try to obtain verbose dmesg with mis-identified frequency.

Tried, and failed after 20+ reboots.  I've left

boot_verbose=" -v"

in /boot/loader.conf to catch any boot-time wonkiness and undone it at
runtime with 

debug.bootverbose=0

in /etc/sysctl.conf as I found that the snd_hda driver is ... chatty
at runtime.

Thanks,

Gary



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180102164526.GA63396>