Deep learning (DL) and reinforcement learning (RL) based methods can efficiently generate offloading strategies for computational offloading problems in mobile edge computing (MEC) environments. However, the rapid movement of vehicles in the vehicular network causes dynamic changes in the network environment, and DL or RL methods require additional training samples and multiple gradient updates before the model converges, which is time-consuming. In this letter, we propose a meta reinforcement learning-based computation task offloading and resource allocation (MRLOA) algorithm for vehicular networks. Specifically, the MRLOA can converge quickly for a new task with a small number of experience and gradient updates based on a pre-trained meta policy. Simulation results show that our proposed algorithm can more quickly adapt to new computational offloading tasks in the vehicular network environment compared to traditional baseline algorithms.