It's been a while since I used MySQL (Thanks Redis!) but mysqld_safe should just be the control script for managing the MySQL daemon, not much to do with the actual running MySQL server itself, I think?
In which case if it's using any meaningful amount of CPU at all it's probably just broken. Try shutting down the MySQL server using init or service or whatever, then just force kill the mysqld_safe and reboot when you're sure it's gone, sounds like it'd be the safest quick, easy fix that might work.
[ Edit: After thinking this couldn't be that rare,
a quick Google turns up the interesting issue that this happens quite often... There are all sorts of ways of attempting to get around it, too, by the look of things, but what I've suggested is likely going to be quicker than any of them. Low disk space would probably be the common suggestion that I'd look into next. ]