[linux-audio-user] Re: Realtime Kernel Slows My Clock

[Nigel Henry]

On Tuesday 07 February 2006 19:12, Chuck Martin wrote:
> On Tue, Feb 07, 2006 at 09:30:15PM +1100, Loki Davison 
loki.davison-at-gmail.com |LAU| wrote:
> > On 2/7/06, Chuck Martin <dsdhpnw02 at sneakemail.com> wrote:
> > > I've been trying to get a realtime kernel running, but every kernel
> > > I've tried with a realtime patch slows my clock to a crawl, which
> > > causes my time (as shown with the date command) to be off.  It also
> > > causes any program that has a delay to wait far longer than it should. 
> > > "sleep 1" can take anywhere from 10 to 30 seconds.
> > >
> > > My most recent try was with patch-2.6.15-rt16, which I've tried
> > > applying to both the 2.6.15 and 2.6.15.2 kernels, with the same result.
> > >  Does anyone have any idea what I could be doing wrong, since no one
> > > else on this list seems to be having this problem.
> >
> > Dual core amd 64 or not?
>
> No and no.  600MHz Pentium III.
>
> Chuck

Hi Chuck. My old Gateway 500 P111 Katmai using 
kernel-2.6.10-0.4.rdt.rhfc2.ccrma, and realtime modules "rtirq" and "rtload", 
is showing similar problems. I had got NTP up and running a while back, and 
initially had problems with ntpd dying when the machine had booted up. With 
suggestions from Fernando at Planetccrma, I moved the load point of rtload to 
before ntpd loaded, which solved that problem. I was then getting time slips, 
even with ntpd running, and problems contacting the ntp Internet servers. 
This was severly exaggerated during the night when losing the dial-up 
connection, resulting in the system clock stopping. Next day, without 
re-connecting to the Internet, using the time and date facility, I'd reset 
the time as near as possible, leave the time and date window open, and not 
move the mouse pointer or do anything on the machine. After a few minutes the 
clock would stop again. Move the mouse pointer, and Bingo, the clock starts 
again. I havn't solved the problem on this machine, but it's clearly 
connected to realtime, and reverting to using the original kernel-2.6.5-1.358 
(no realtime), the problem is resolved, and the clock is rock steady, with no 
problems with ntp contacting the Internet servers.

Strangely, the other machine on my network. An i-Friend 1.3Ghz celeron, also, 
in this example running FC2, with the same planetccrma kernel with the 2 
realtime modules, has no clock problems. This machine, again with ntpd 
enabled, retrieves it's time from the machine with the problems. Even with 
the terrible time slips (clock stopped) on the Gateway, the i-Friend machines 
clock stayed on time.

Incidentally. The main reason that I enabled NTP was because I'd been getting 
time slips on the Gateway machine, but obviously this is a realtime problem, 
specific to this machine. Processor problem? I notice that your P111 is only 
a bit faster than mine, and strange that my other machine with the same setup 
shows no time slips.

I have posted this problem to questions at lists.ntp.org, but with no replies.

Nigel.