Atom-Aid: Detecting and Surviving Atomicity Violations

Writing shared-memory parallel programs is an error-prone process. Atomicity violations are especially challenging concurrency errors. They happen when programmers make incorrect atomicity assumptions and fail to enclose memory accesses that should occur atomically inside the same critical section. If these accesses happen to be interleaved with conflicting accesses from different threads, the program might behave incorrectly. Recent architectural proposals arbitrarily group consecutive dynamic memory operations into atomic blocks. This provides what we call implicit atomicity, as the atomic blocks are not derived from explicit program annotations. In this paper, we make the fundamental observation that implicit atomicity probabilistically hides atom-icity violations by reducing the number of interleaving opportunities between memory operations. We then propose Atom-Aid, a new approach that creates implicit atomic blocks intelligently instead of arbitrarily, dramatically reducing the probability that atomicity violations will manifest themselves. Atom-Aid can infer when an atomicity violation is likely to happen before it actually happens and avoid dangerous interleavings. Atom-Aid does not change the memory semantics observed by the software and also applies to transactional memory systems. Atom-Aid is also able to report where likely atomicity violations are in the code, providing resilience and debugability. We evaluate Atom-Aid using buggy code from applications including Apache, MySQL and XMMS. Our evaluation shows that Atom-Aid hides from 98.7 % to 100 % of the atomicity viola-tions, virtually eliminating them.