I have built a 2.2.17 kernel for a Dell Dimension xps h266 and
as I stated in earlier messages to the group, it basically works. The
system comes up, does dhcp, and talks on the network like it should.
The first hint of trouble happens at almost exactly 10 minutes after
boot when the date suddenly shoots ahead some random amount of time.
I have had it gain about 16 hours before or it may gain as much as 42
or 43 hours. I don't think the actual amount tells us anything except
that something is writing junk to a counter.
I was telling this tale of woe to some of our local Linux
group and somebody said that there is a jiffy clock which is
deliberately set to overflow about 10 minutes after initial start so
that one can know if one's software can handle the overflow. It
sounds like something is not handling it, all right. I don't know for
sure that this is the problem, but the glitch seems to hit right at
the ten minute mark. I waited around another 10 minutes until the
uptime output said that the system was up for 20 minutes and there was
no further sudden increase. Oddly enough, the uptime counter must not
derive its value from the same place as the counter for the time and
date, or the conversion program for the local time zone must be
corrupted for the uptime value does not reflect the glitch. It keeps
indicating a value appropriate for the amount of time the system has
been what passes for up.
When I posted this message last week, I mainly asked where I
should be looking to solve a problem like this, and my question still
remains although I have found out a few more things since then.
One of the local Linux group folks I talked to said that he
had had similar problems and had gone back to an earlier kernel but
that 2.2.13 had a security problem of some kind. I certainly hope
that I am not perpetually bound to an old kernel for this otherwise
How often does the jiffy clock overflow under normal
conditions? The person who told me about it wasn't sure but thought
it was several weeks and that the kernel boot procedure simply set it
to a value that was about 10 minutes worth of ticks below the maximum
value so that one would find out what happened sooner rather than
later.:-) What is the granularity of this clock or how many times per
second does it get updated?
Martin McCormick WB5AGZ Stillwater, OK
OSU Center for Computing and Information Services Data Communications Group