A reliable booting system for Zynq Ultrascale+ MPSoC devices

CERN is working on the High-Luminosity LHC upgrade which will be installed in 2025. As a result, the CMS-experiment and its data acquisition (DAQ) system will also be upgraded. The upgrade of the CMS DAQ system involves the installation of new electronics that will also host the Zynq Ultrascale+ MPSoC from Xilinx (Multiprocessor Systems on a Chip). The Zynq Ultrascale+ MPSoC will run control and monitoring software on a Linux operating system (OS). Booting a Linux OS on the Zynq MPSoC involves a complex multi-stage booting process. The complexity of the booting process introduces possible failures that can prevent the Zynq MPSoC from booting correctly. This thesis presents the research, design, implementation, and testing of a reliable booting system that recovers the Zynq MPSoC from boot failures, upgrade failures, and running failures. The reliable booting system consists of five fallbacks in different parts of the Zynq MPSoC booting process, to account for a wide range of failures. The fallbacks have been designed to bring the Zynq MPSoC to a well-known booted state after a failure. The booting system can also boot through the network and perform automatic firmware upgrades with a rollback on failure. Users of the hardware are automatically notified after a failure was detected and a fallback was triggered in the system. The booting system is automatically built and packaged by a continuous integration build system. It has been made portable for new hardware by integrating the system in an easy-to-use board support package. Research on the possible failures in the Zynq MPSoC has been carried out. The test results have concluded that the fallbacks are able to successfully recover the Zynq MPSoC from all the researched failures. The results also highlighted a few areas that can be researched in a follow-up project to further improve the reliable booting system.