SMP and NTPD (Eric Mader)

[David G. Miller]

Eric Mader <emader at icu-project.org> wrote:

> I also noticed the following in /var/log/messages: Apr 11 09:21:19 
> localhost kernel: warning: many lost ticks. Apr 11 09:21:19 localhost 
> kernel: Your time source seems to be instable or some driver is 
> hogging interupts I'll bet this has something to do w/ the problem. 
> (:-) Does anybody know how to figure out what this is about? (I have a 
> vague memory of having seen the same problem w/ the SMP kernel on a P4 
> w/ hyperthreading. On this system it was so bad that the whole system 
> would lock up. My AMD keeps chugging along, more or less.) Regards, 
> Eric Eric Mader wrote:
>
>>> Hello,
>>> 
>>> I'm running kernel 2.6.11-1.1369_FC4smp on an AMD64 x2 system. A couple 
>>> of days ago I noticed that NTPD wasn't keep time very well. I did some 
>>> investigating and found out that the jitter from the time servers kept 
>>> going up and up - after an hour the jitter will be several seconds.
>>> 
>>> As an experiment, I tried running the non-smp kernel, and NTPD was able 
>>> to keep rock-solid sync. After more than an hour, the jitter was only a 
>>> few milliseconds.
>>> 
>>> I'm guessing that this points to some problem w/ the SMP kernel. Can 
>>> anybody confirm this, or suggest what else the problem might be?
>>> 
>>> My system:
>>> ASUS A8V Delux
>>> AMD Athlon 64 X2 4200+
>>> 1 GM memory (512 MB x 2)
>>> WD1200JJ (w/ Windows XP x64)
>>> WD400BB (w/ FC4)
>>> NEC DVD RW 3520A
>>> TDK CDRW 5200B
>>> NVidia GeForce 6600 (ASUS brand)
>>> Samsung SyncMaster 204T (connected through a DVI KVM)
>>> USB Keyboard and mouse (connected through KVM)
>>> 
>>> Regards,
>>> Eric Mader
>>> 
>>    
>>
Might be specific to the dual core Athlon.  I'm running a traditional 
dual Athlon (Tyan Tiger MPX with 2x 32bit Athlon 2400+s) and not seeing 
that problem at all:

[dave at bend ~]# uname -a
Linux bend.local.davenjudy.org 2.6.15-1.1833_FC4smp #1 SMP Wed Mar 1 
23:56:51 EST 2006 i686 athlon i386 GNU/Linux
[dave at bend ~]# ntptime
ntp_gettime() returns code 0 (OK)
  time c7e82823.24468000  Wed, Apr 12 2006 19:30:11.141, (.141701),
  maximum error 606915 us, estimated error 5751 us
ntp_adjtime() returns code 0 (OK)
  modes 0x0 (),
  offset 847.000 us, frequency 152.009 ppm, interval 4 s,
  maximum error 606915 us, estimated error 5751 us,
  status 0x1 (PLL),
  time constant 6, precision 1.000 us, tolerance 512 ppm,
  pps frequency 0.000 ppm, stability 512.000 ppm, jitter 200.000 us,
  intervals 0, jitter exceeded 0, stability exceeded 0, errors 0.
[dave at bend ~]# uptime
 19:30:26 up 8 days, 22:50,  8 users,  load average: 0.31, 0.16, 0.18

I have seen lots of clock weirdnesses with my AMD64 (single core) laptop 
(ATI chipset).  Lots of stuff about this under bug 
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=152170.  Any idea 
what chipset your ASUS board uses?

Cheers,
Dave

-- 
Politics, n. Strife of interests masquerading as a contest of principles.
-- Ambrose Bierce