Blockchained Federated Learning (FL) has been gaining traction for ensuring the integrity and traceability of FL processes. Blockchained FL involves participants training models locally with their data and subsequently publishing the models on the blockchain, forming a Directed Acyclic Graph (DAG)-like inheritance structure that represents the model relationship. However, this particular DAG-based structure presents challenges in updating models with sensitive data, due to the complexity and overhead involved. To address this, we propose Blockchained Federated Unlearning (BlockFUL), a generic framework that redesigns the blockchain structure using Chameleon Hash (CH) technology to mitigate the complexity of model updating, thereby reducing the computational and consensus costs of unlearning tasks.Furthermore, BlockFUL supports various federated unlearning methods, ensuring the integrity and traceability of model updates, whether conducted in parallel or serial. We conduct a comprehensive study of two typical unlearning methods, gradient ascent and re-training, demonstrating the efficient unlearning workflow in these two categories with minimal CH and block update operations. Additionally, we compare the computation and communication costs of these methods.