Quote:

I was also, at the same time, seeing clock drift. I fixed the clock drift with a kernel command-line option (which I can't remember -- I'll dig it out if necessary), which also fixed the keyboard repeat problem.


Ahh.. the infamous "unsynced TSC" problem, perhaps. Where one CPU grabs a timestamp from its TSC, representing the key "press".. and the other CPU eventually compares against its current TSC to see if it's time to generate "repeats".

If the two TSCs are not perfectly synchronized, then the keyboard repeat can go all to he**

This is apparently only a problem on AMD processors at present..

Disclaimer: I don't actually know if the keyboard driver uses the TSC for its timestamps; this example was just to illustrate the potential problems