Clock stoppage

[glywood]

I'm working on installing gentoo on my computer, but I'm having some hardware difficulties. It seems that after running gentoo (either an installed version or the one on the live cd) for a long time (I can't give an exact number, but it seems like it's usually an hour or two) things start to behave a little weird. The first thing I noticed was that key repeat stopped working, that is, I could hold down a key for as long as I wanted and the character would only show up once. Also, some of the packages I was emerging started to give compilation warnings about having invalid timestamps. I ran the date command a few times, and found out that my clock had stopped running.

Well, to be fair, the clock didn't actually stop dead right away. It went through a period where it was usually at around the same time, give or take a couple of seconds, with the odd jump a couple of hours or days into the future before falling back.

hwclock always returns the correct time. I'm running an IBM NetVista with a 2.2GHz Celeron pretty well straight out of the box.

Has anyone else had these problems? Any help would be appreciated.

[setagllib]

Is it using the RTC or TSC or ACPI or what timesource? Technically only the RTC can be relied upon for the 'real' time. If you can force changing this, try it.

Besides that, see if similar things happen with BSDs, Windows, other Linux kernels, etc. It could be a bug or a BIOS 'feature' or anything. At least if it's a kernel issue you can work around it... by running something else. If it's hardware you're stuffed, if it's BIOS you could try updating.
_________________
My other computer is your Windows box.

[glywood]

How do I go about checking what the timesource is?

I have never seen this sort of thing happen in other operating systems (I've used windows xp and redhat 9 extensively on this machine).

I've recompiled the kernel without ACPI support, we'll see in the morning if that fixed things.