Autonomous driving has rapidly developed and shown promising performance with recent advances in hardware and deep learning methods. High-quality datasets are fundamental for developing reliable autonomous driving algorithms. Previous dataset surveys tried to review the datasets but either focused on a limited number or lacked detailed investigation of the characters of datasets. To this end, we present an exhaustive study of over 200 autonomous driving datasets from multiple perspectives, including sensor modalities, data size, tasks, and contextual conditions. We introduce a novel metric to evaluate the impact of each dataset, which can also be a guide for establishing new datasets. We further analyze the annotation process and quality of datasets. Additionally, we conduct an in-depth analysis of the data distribution of several vital datasets. Finally, we discuss the development trend of the future autonomous driving datasets.