AZMA: A Zynq-Based Monitoring and Fault Injection Framework for Processor Assessment

In the digital world, processors are crucial for executing operations, and any flaw can cause system deviations. In some applications, data or instructions in memory can be manipulated to alter the system's main goal. Therefore, cycleaccurate analysis and evaluation of processors are essential, particularly when fault injections inside memory. Consequently, it is necessary to enhance processors with features for debugging and assessment. Recent advancements in hardware have led to aathe integration of application-specific functions directly into processors. However, embedding new functionalities is resource-intensive and time-consuming. The AZMA framework addresses this by providing a flexible, hardware-assisted runtime monitor and fault injection. Additionally, the framework incorporates startup and control signals to manage the evaluation flow effectively. The processor is instantiated on the programmable logic (PL) segment of the Zynq, with design points accessible via the AXI bus, which serves as the interface between the PL and the processing system (PS) segment of the Zynq. The PS executes a piece of code that manipulates these design points for evaluation and monitoring purposes. Furthermore, a Python-based library has been developed to facilitate user-defined fault injection and monitoring, seamlessly integrating with the framework. Experimental validation was conducted by executing the Fibonacci sequence on the processor, during which faults were injected. The design's hotspots were monitored with cycle-level precision to verify the functionality and resilience of the tested design.