A misconception in blocking time analyses under multiprocessor synchronization protocols

摘要

In multiprocessor systems, synchronization protocols can result in non-trivial (remote) blocking, which can cause timing impacts in real-time systems. To analyze the schedulability or the worst-case response time, it is essential to calculate a safe upper bound on the maximum remote blocking time due to the synchronization protocols. The derivation of the upper bound is sometimes unsafe in the literature when the analyses improperly adopt a misconception that assumes the well-known critical instant theorem. In this paper, we show that the original analyses for the distributed priority ceiling protocol and the multiprocessor priority ceiling protocol are unsafe in the calculation of the blocking time due to this misconception. This results in repetition of unsafe timing analyses in the literature. This paper also provides a simple remedy for such a flaw.