It looks like there's a potential race in

    void trace_locking(Mutex* mutex, char *type, char *state) {

where if more than one thread calls it when TraceLocking_lock is
NULL, they will all try to allocate a Mutex and assign it to
TraceLocking_lock. That's race #1. Then whether any given
thread uses the same value for TraceLocking_lock->lock() and
TraceLocking_lock->unlock() is questionable, too.