Re: clock jump in 0.2.0.23-rc, too

[Scott Bennett <bennett@xxxxxxxxxx>]

    On Wed, 2 Apr 2008 09:21:44 -0400 phobos@xxxxxxxxxx wrote:
>On Tue, Apr 01, 2008 at 04:58:22AM -0500, bennett@xxxxxxxxxx wrote 4.5K bytes in 68 lines about:
>: 	hellas # sysctl machdep.hyperthreading_allowed=0
>: 	machdep.hyperthreading_allowed: 1 -> 0
>
>Just for a test, have you tried disabling HT in the bios?  I'll setup a

     No, I haven't.  Now that I think of it, I don't remember *ever* doing it
that way, probably because I had no reason to play around with it, and a reboot
would be required to change it and then again to change it back.  I suppose I
could also build a non-SMP kernel to try, as well.  The GENERIC kernel is a
non-SMP kernel, and it recognizes the two logical CPUs at boot time, but then
ignores cpu1 from that point on.

>fbsd-7-stable sytem with 0.2.0.23-rc on an HT system to see if I can
>replicate.  
>
     Okay, but that sounds like an inordinate amount of hassle.  Also, I'm
running 6.3-STABLE, not 7.0-STABLE.
     Shortly after midnight, my Internet connection went down, and it didn't
come back up till about 10 a.m.  While the link was down and more than 21 hours
after I had disabled hyperthreading in the kernel, tor put out the following
messages:

Apr 02 02:01:22.178 [warn] Your system clock just jumped 115 seconds forward; assuming established circuits no longer work.
Apr 02 02:03:14.235 [warn] Your system clock just jumped 112 seconds forward; assuming established circuits no longer work.

Notice that, once again, the second message is timestamped the same number of
seconds later than the first message as the number of seconds complained of in
the second message.  I.e., no time actually elapsed between the two messages.
(Note also that these messages *always* say the clock has jumped *forward*,
never backward, which also argues against the idea of being whipsawed between
two out-of-sync clocks on a SMP system.)  It did this at a time that was about
14 - 16 minutes before ntpdate would next have been run by cron, and while the
network connection was down, meaning that ntpdate would not have changed the
time anyway, and while hyperthreading was disabled.
     So I reenabled hyperthreading because it clearly is not the problem.  All
the evidence suggests, but does not prove, that some change was introduced in
0.2.0.21-rc that is causing the problem.  The change might possibly have been
made earlier than that, but with the symptoms not appearing for some reason
until 0.2.0.21-rc.


                                  Scott Bennett, Comm. ASMELG, CFIAG
**********************************************************************
* Internet:       bennett at cs.niu.edu                              *
*--------------------------------------------------------------------*
* "A well regulated and disciplined militia, is at all times a good  *
* objection to the introduction of that bane of all free governments *
* -- a standing army."                                               *
*    -- Gov. John Hancock, New York Journal, 28 January 1790         *
**********************************************************************