Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 27 Dec 1995 16:03:05 -0500
From:      "Garrett A. Wollman" <wollman@lcs.mit.edu>
To:        peter@haywire.dialix.com (Peter Wemm)
Cc:        freebsd-current@freebsd.org
Subject:   Re: Tick, tock, adjust the clock
Message-ID:  <9512272103.AA17279@halloran-eldar.lcs.mit.edu>
In-Reply-To: <4bqdc4$8h2$1@haywire.DIALix.COM>
References:  <199512261708.LAA14134@miller.cs.uwm.edu> <4bqdc4$8h2$1@haywire.DIALix.COM>

next in thread | previous in thread | raw e-mail | index | archive | help
<<On 27 Dec 1995 11:08:52 +0800, peter@haywire.dialix.com (Peter Wemm) said:

> I've had the same problem for quite some time..  On all the
> FreeBSD machines I have access to, the xntpd oscilates very very badly
> (like you've shown) and eventually logs "not logging any more time
> steps" or something like that.

I'm curious to know whether the clock speed is correctly diagnosed on
these machines...  This sounds like the sort of thing I would expect
to happen if the clock-speed-diagnosis code doesn't come up with an
answer that's reasonably close.  My 60-MHz machine is diagnosed as
59.99-MHz, and xntpd reports its frequency error as about 7.6 seconds
per day.

You might try making i586_ctr_rate tunable via sysctl(8), and fiddling
the value a little bit to see if you can come up with something more
correct.  That would be a very good feature to have, and I may get
around to adding it sometime soon since PHK's sysctl is able to deal
with variables that are only conditionally defined in a reasonable
way.  (It's a fixed-point number with I586_CTR_RATE_SHIFT bits of
fractional precision.)

> The really depressing part, is the SVR4 machines (you know, the ones
> with the horrible 10ms clock resolution) sitting right next to the
> FreeBSD machines with their super-high-res clocks are locking right in
> and staying very stable (xntpd getting to 1024 poll and very low
> dispersion), while the FreeBSD machines are constantly wobbling all
> over the place. :-(

My machine looks like this:

     remote           refid      st t when poll reach   delay   offset    disp
==============================================================================
+catfish.lcs.mit rackety.udel.ed  2 u  294 1024  357     2.24    5.002    1.25
+amsterdam.lcs.m ncar.ucar.edu    2 u  540 1024  377     2.09    5.686    4.17
 eiffel.lcs.mit. tock.usno.navy.  2 -  38m  512    0     2.93    3.993 16000.0
 pepper.lcs.mit. catfish.lcs.mit  3 u  654 1024  377     4.50    5.002    1.46
*pooch.osf.org   .WWV.            1 u   69 1024  377    10.24    4.069    4.06
 halloran-eldar. catfish.lcs.mit  3 u  815  512  324     3.97    6.322    4.65
 hergotha.lcs.mi catfish.lcs.mit  3 u  417 1024  252     0.87    4.868    7.98
 NTP.MCAST.NET   0.0.0.0         16 u    -   64    0     0.00    0.000 16000.0

-GAWollman

--
Garrett A. Wollman   | Shashish is simple, it's discreet, it's brief. ... 
wollman@lcs.mit.edu  | Shashish is the bonding of hearts in spite of distance.
Opinions not those of| It is a bond more powerful than absence.  We like people
MIT, LCS, ANA, or NSA| who like Shashish.  - Claude McKenzie + Florent Vollant



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9512272103.AA17279>