Monitoring of I/O for safety-critical systems using PCI express advanced error reporting

Certification is the process in which a manufacturer has to prove to authorities that an aircraft and systems like safety-critical avionic systems work safely as intended without unacceptable hazards. Means to achieve this are redundancy, separation/protection, monitoring, and recovery mechanisms. The objective of this paper is to provide a cost-efficient solution for monitoring of Commercial Off-The-Shelf (COTS) Input/Output (I/O) devices using COTS components and standards. The first contribution of this paper is an evaluation-proved single-event monitoring concept to permit PCI Express (PCIe) debugging without an expensive PCIe protocol analyzer and to facilitate build-in self-tests for COTS PCIe I/O devices. This is proved with an evaluation. The second contribution is a run-time monitoring concept to monitor, detect, prevent errors and to recover from errors for spatial separation of COTS PCIe I/O devices during operation. Both monitoring types use only available PCIe-compatible COTS components. This makes them portable to various computing architectures (Intel, ARM, PowerPC) and applicable for existing I/O virtualization and I/O sharing solutions. A practical evaluation reveals that the performance overhead (transfer time, transfer rate) inserted by the run-time monitoring is with 0.88% small and insignificant.