Glenn English wrote:
The only thing I can think of is that something got bent in the power failure -- something that the Debian boot process doesn't look at and set, but BSD does. But I don't quite believe it, and I have no suspects for the "something."
One thing a power failure (or possibly running another OS) can do is temporarily defeat the hwclock init scripts by throwing off the value of /etc/adtime. These scripts assume they are the only programs touching the hardware clock (see docs). NTP also might throw off these scripts, especially if its not working correctly. (I don't know if any of this is related to your problem or not.) The bottom line here is that you've got some basic debugging to do. :-) What I did to sort out my clock probems was to first disable NTP, then install adjtimex and experiment and watch the system and hardware clocks closely. I also recorded and analyzed all changes to /etc/adjtime and /etc/default/adjtimex. The man pages of hwlock and adjtimex are very informative and full of useful tips. If you spend a little time I think you will easily locate the problem, if not the solution. This is how I found out that I was losing timer ticks with the 2.6.8 kernel.