Linux Boot Failures Under Proton Irradiation

To meet the ever-growing computational demand of high-performance applications, and specifically to keep up with the trend of AI and increased autonomy, space-based systems must be equipped with and operate increasingly more powerful computing systems.
At the same time, costs demand using commercial off the shelf (COTS) systems and the functionally rich operating systems they run.
However, naturally, these systems have not been built for the harsh environment of space, and even in low-earth orbit, such systems are exposed to radiation levels, which may cause software to fail and, when left untreated, may crash the entire system permanently.
Rejuvenation is one method for addressing radiation faults. It involves restarting and power cycling the computer system and rebooting its software stack to return it to a state in which it can continue to operate correctly.
However, when applied to COTS systems, frequent reboots become necessary, and while the radiation effects of software running on radiation-hardened systems is relatively well understood, the same is not true for the boot process of legacy operating systems, such as Linux, on COTS platforms in radiative environments.
To bridge this gap, we performed an up to 58 MeV proton radiation test, observing two Linux-based operating systems booting on an iMX8 Plus compute module and report here on our results.
We found Linux to be significantly less resilient to radiation faults until it has properly configured the processor's fault handling capabilities. Moreover, while file system failures were the most frequent, failures in the Linux core and driver initialization code caused most of the crashes.