Whilst diagnosing an unrelated problem by looking through
a users dmesg, I noticed this..
CPU#0 had 0 usecs TSC skew, fixed it up.
CPU#1 had 0 usecs TSC skew, fixed it up.
Which looks very strange.
Taking a look at the code in arch/i386/kernel/smpboot.c:synchronize_tsc_bp()
I'm puzzled at to how this happened, as we should only hit that
printk with a skew of more than 2 usecs.
Dave