the clock stopped in F7 ?!

[Mike]

On Sun, 26 Aug 2007, Karl Larsen wrote:

> Lonni J Friedman wrote:
>> On 8/26/07, Karl Larsen <k5di at zianet.com> wrote:
>> 
>>> Lonni J Friedman wrote:
>>> 
>>>> I've got a Fedora 7 (x86) system that started exhibiting truly bizarre
>>>> behavior about a week ago.  Basically, the clock stopped working.  If
>>>> I run 'date' it shows the date/time from a few days earlier, and it
>>>> *never* changes.  If I touch a file, it has the date/timestamp from
>>>> the time/date in date output.  The odd thing is that this behavior
>>>> only happens when the system sits relatively idle for a long chunk of
>>>> time (at least 24 hours).  If i'm actively using it every day, then
>>>> its fine.  If I reboot, then the problem goes away (and the system has
>>>> the correct time after rebooting).
>>>> 
>>>> The first time that this happened was last weekend (Aug 18), and I had
>>>> to reboot it last Monday (Aug 20) to fix the problem.  Its now
>>>> happened again.  At this moment in time, date claims that its Sat Aug
>>>> 25, even though its actually Sun Aug 26 right now.
>>>> 
>>>> To make matters worse, the system behaves oddly when this problem
>>>> occurs.  I suspect its because anything that relies on getting an
>>>> accurate (or changing) clock is failing.  If I attempt to reboot
>>>> cleanly, it just never happens.  The system acts frozen in time.
>>>> 
>>>> I've checked dmesg & messages, and there's nothing there.  messages
>>>> just stops logging anything around the time that the clock appears to
>>>> have frozen.
>>>> 
>>>> Anyone ever seen this bizarre behavior, or have any ideas what might
>>>> be going on?
>>>> 
>>>>
>>>>
>>>     There is a battery on your motherboard and it has a clock that needs
>>> the battery. Linux checks the computer battery every so often so check
>>> that battery and replace if needed. I can cause all your problems.
>>> 
>> 
>> If it was the CMOS battery,  why would it be working fine for days,
>> stop working, then start working again after a reboot?
>> 
>> Also, I've never heard of Linux being capable of checking the CMOS
>> battery.  What specifically is doing this check?
>> 
>> Additionally, the CMOS battery is only needed when the system is
>> powered down and/or doesn't have external power.  It certainly isn't
>> used to keep the system clock running while the system is running on
>> external power.
>> 
>> I appreciate your feedback, but what you're saying really doesn't make
>> any sense.
>>
>>
>   Sorry but I wanted to say the battery runs a clock on your motherboard 
> that Linux reads from time to time. If the battery is dead or weak it will 
> not run the clock with accuracy.
>
> Better?

Nope, the battery backed CMOS clock is only read by the kernel at boot 
time but never again.  After that the "software" clock in the kernel is 
updated by a hardware interrupt.

To the OP it almost sounds like a hardware problem.

Try 'cat /proc/interrupts | grep timer' two times or more.

Look at the number just to the right of '0:'

This should have incremented a bunch in between cat's.  If it didn't then 
you likely have a hardware problem or the timer is failing to get 
initialized.  If it is incrementing then I'm stumped...

-- Mike