Automatically classifying benign and harmful data races using replay analysis

Abstract

Many concurrency bugs in multi-threaded programs are due to data races. There have been many efforts to develop static and dynamic mechanisms to automatically find the data races. Most of the prior work has focused on finding the data races and eliminating the false positives.In this paper, we instead focus on a dynamic analysis technique to automatically classify the data races into two categories -the data races that are potentially benign and the data races that are potentially harmful. A harmful data race is a real bug that needs to be fixed. This classification is needed to focus the triaging effort on those data races that are potentially harmful. Without prioritizing the data races we have found that there are too many data races to triage. Our second focus is to automatically provide to the developer a reproducible scenario of the data race, which allows the developer to understand the different effects of a harmful data race on a program's execution.To achieve the above, we record a multi-threaded program's execution in a replay log. The replay log is used to replay the multithreaded program, and during replay we find the data races using a happens-before based algorithm. To automatically classify if a data race that we find is potentially benign or potentially harmful, we replay the execution twice for a given data race -one for each possible order between the conflicting memory operations. If the two replays for the two orders produce the same result, then we classify the data race to be potentially benign. We discuss our experiences in using our replay based dynamic data race checker on several Microsoft applications.

Software Analysis Techniques for Detecting Data Race

“…Furthermore, even true data races can be benign and would not do any harm to the system operation. It has been reported that the amount of such harmless data races can be so big as 76%-90% of the true data races reported by the modern detectors [25], which calls for a sophisticated detection mechanism for distinguishing harmful races over harmless ones [26].…”

Understanding and identifying latent data races cross-thread interleaving

“…This simple idea for classifying the characteristic of data races is used in Ref. [4]. We extend it for the distinction of benign and conflicting synchronization pair.…”

“…A data race often occurs in the scenario where more than two different threads, at least one of which is writing the value, access the same shared memory location without appropriate synchronization [1]. Despite the destructiveness of data races (e.g., Northeast blackout 1) and Therac-25 accidents 2) ), current thread specifications of C 3) and C++ 4) still permit the existence of data races intentionally for the sake of better program performance.…”

“…A volume of research [2][3][4] has been devoted on the solutions to diagnose data races, among which Lamport's happens-before (hb) relation [5] gradually functions as the fundamental dynamic approach. However, it is well-known that these existing dynamic Happens-Before based race detection techniques are interleaving-sensitive and may fail to detect some races.…”

Efficient Runtime Detection and Toleration of Asymmetric Races

“…In this situation, even though a race occurs by definition (the shared variable flag is written and read without explicit synchronization), it does not harm the program. Narayanasamy et al [40] show other types of benign symmetric races, e.g., redundant writes and disjoint bit manipulation.…”