Linear Models & Regression — weighted least-squares estimate of a common mean

Independent observations $y_1, y_2, ..., y_n$ have been drawn from $N(\mu,\sigma^2/x_i^2)$; $i=1,2,...,n$; where $x_i$'s are not equal, uncorrelated with the errors in the model and $x_i\neq0$ for all i. The least squares estimate of $\mu$ would be