Put a number on omitted variable bias

The true model is $y = 2 x_1 + 3 x_2 + \varepsilon$ with $\varepsilon$ exogenous. You omit $x_2$ and regress $y$ on $x_1$ alone. The auxiliary regression of the omitted $x_2$ on the kept $x_1$ has slope $\delta = 0.5$.

What does your estimated coefficient on $x_1$ converge to, and what is the size and sign of the bias?