A randomized scheduler with probabilistic guarantees of finding bugs

Abstract

This paper presents a randomized scheduler for finding concurrency bugs. Like current stress-testing methods, it repeatedly runs a given test program with supplied inputs. However, it improves on stress-testing by finding buggy schedules more effectively and by quantifying the probability of missing concurrency bugs. Key to its design is the characterization of the depth of a concurrency bug as the minimum number of scheduling constraints required to find it. In a single run of a program with n threads and k steps, our scheduler detects a concurrency bug of depth d with probability at least 1/ nk d-1 . We hypothesize that in practice, many concurrency bugs (including well-known types such as ordering errors, atomicity violations, and deadlocks) have small bug-depths, and we confirm the efficiency of our schedule randomization by detecting previously unknown and known concurrency bugs in several production-scale concurrent programs.

Delay-bounded scheduling

“…We show that delay-bounded depth-first state-reachability is an NPcomplete problem. 7 Though this exploration corresponds to an underapproximation of the program semantics, the complexity is lower than the precise EXPSPACE-complete explorations of Sen and Viswanathan [37] and Jhala and Majumdar [19] for non-preemptive asynchronous programs. Additionally, our underapproximation has a lower complexity than the PSPACE-hard underapproximation [5] used by Jhala and Majumdar [19]'s algorithm, which corresponds to our (non-preemptive) bounded bag semantics (see Scheduler 2).…”

“…Finally, runtime-schedule "fuzzers" such as CONTEST [12] and CALFUZZER [20] introduce delays by adding sleep statements; PCT [7] introduces delays by randomly perturbing task priorities. These techniques share with delay-bounding the ability to scale to many tasks.…”

Coverage guided systematic concurrency testing

“…Our method is also different from the various testing techniques based on randomization, e.g. IBM's Contest [4] and [1], although randomization can be used to diversify the input to our HaPSet learning.…”

The influence of environmental parameters on concurrency fault exposures

“…Therefore, it is expected that operating systems 1 The classification is not mean to be a complete list. 2 In the master-slave technology, a single core manages task assignments, whereas, in the dataflow-based core managements, the assignments are based on data dependencies. 3 A technology to hide the physical characteristic of computing platforms.…”

“…Eytani et al introduce a randomization algorithm which targets events causing interleaving defects. Burckhardt et al [2] state that interleaving bugs are usually caused by unanticipated interactions among a few instructions, which are shared by the threads. Burckhardt et al present a probabilistic concurrency testing scheduler that runs a test program several times with given inputs and computes the probability of finding bugs for each test run.…”