Characterization of Interrupt Handling in Board Management Controllers

Outside the Systems Group at ETH Zurich, there are not many research possibilities in regards to Board Management Controllers (BMC). However, research into them is nevertheless significant as BMCs are responsible for many things, one of them being the emergency response on hardware faults. Even with such a critical responsibility, the response time of the BMC interrupt handler on the Enzian system, developed at ETH, was preliminarily measured in the order of hundreds of milliseconds. Further, this vital system could not be characterized in detail up to this point, as no tools to do so worked on the operating system of the BMC. This work shows how it is possible to get some tracing to work for the OpenBMC-based kernel in use on Enzian. Additionally, a test suite named the TestBench, was created to trace multiple Linux tools repeatedly and consistently. Inefficiencies in the current implementation were found utilizing these tools and the improvements to them reduce the total response time of the interrupt handling by 30-45%