As the number of IoT devices grows at an exponential rate, the security issues of these devices are having a huge impact on people's lives. Fuzzing, a dynamic testing method that can be automated at scale, is becoming more and more extensively utilized on I0 devices in order to find the vulnerabilities in these devices. In this work, we present an efficient feedback-enhanced fuzzing scheme for Linux-based IoT firmwares. It uses a two-level scheduler and a feedback-enhanced monitor to collect operating data for seed selection and ranking. We develop a prototype system and evaluate it by emulating and testing five different firmwares. The result shows that our system is able to effectively discover more known vulnerabilities than the state-of-the-art IoT fuzzer and eventually discovered 5 unknown vulnerabilities.