This looks wrong and may overflow! Do it like shown below instead, and implement mul_u64_u64_div() at the same time:

static inline uint64_t
mul_u64_u64_div(uint64_t num, uint64_t mul, uint64_t div)
{
       const uint64_t rem = num % div;

       return (num / div) * mul + (rem * mul) / div;
}

In any case, BSD style is putting the return (value) in ().

x->num and y->rem