I've been fiddling with a new kernel, and have had several occasions
to reboot lately.
If I mounted /boot to cp the new kernel etc over, I have a problem on
reboot for sure.
Somehow the date of last fsck on /boot is seen as `in the future' so
fsck fails on /dev/had1 (/boot).
Which means nearly all other boot time services also fail. So I end
up logging into a system with no services running and only `/' mounted.
At that point, I run fsck /dev/hda1 which finds a date error, fixes it
and then reboot... this time everything works, and if I don't mount
/boot a reboot just works... but if I end up having to fiddle further
with kernel, mount /boot to copy over etc. On reboot the same problem
occurs.
I tried to get ahead of the game by umounting /boot after cp over
kernel and running fsck on it before reboot. fsck doesn't find a
problem. But at reboot... the same problem occurs.
What it means is every reboot requires 2 reboots (if I mounted /boot)
I'm guessing its some kind of timing problem with events during boot.
But not sure what to do about it.
The clock can't be getting that far off in a few seconds, and is reset
when ntp-client runs. So I don't understand the error saying `in the
future'.