Automated Classification of Data Races Under Both Strong and Weak Memory Models

Abstract

de Lausanne (EPFL), Switzerland Data races are one of the main causes of concurrency problems in multithreaded programs. Whether all data races are bad, or some are harmful and others are harmless, is still the subject of vigorous scientific debate [Narayanasamy et al. 2007;. What is clear, however, is that today's code has many data races [Kasikci et al. 2012;Jin et al. 2012;Erickson et al. 2010], and fixing data races without introducing bugs is time consuming [Godefroid and Nagappan 2008]. Therefore, it is important to efficiently identify data races in code and understand their consequences to prioritize their resolution.We present Portend + , a tool that not only detects races but also automatically classifies them based on their potential consequences: Could they lead to crashes or hangs? Could their effects be visible outside the program? Do they appear to be harmless? How do their effects change under weak memory models? Our proposed technique achieves high accuracy by efficiently analyzing multiple paths and multiple thread schedules in combination, and by performing symbolic comparison between program outputs.We ran Portend + on seven real-world applications: it detected 93 true data races and correctly classified 92 of them, with no human effort. Six of them were harmful races. Portend + 's classification accuracy is up to 89% higher than that of existing tools, and it produces easy-to-understand evidence of the consequences of "harmful" races, thus both proving their harmfulness and making debugging easier. We envision Portend + being used for testing and debugging, as well as for automatically triaging bug reports.

Dependence-aware, unbounded sound predictive race detection

“…Sampling-based analysis trades coverage for performance (opposite of predictive analysis) in order to detect data races in production [Biswas et al 2017;Bond et al 2010;Erickson et al 2010;Kasikci et al 2013;Marino et al 2009;Sheng et al 2011;. Custom hardware support can detect data races with low performance cost but has not been implemented [Devietti et al 2012;Lucia et al 2010;Marino et al 2010;Peng et al 2017;Segulja and Abdelrahman 2015;Wood et al 2014;Zhou et al 2007].…”

“…Dynamic analysis can estimate the likely harm of a data race [Boehm 2011;Burnim et al 2011;Cao et al 2016;Flanagan and Freund 2010a;Kasikci et al 2015;Narayanasamy et al 2007], which is orthogonal to detection. All data races are erroneous under language memory models that ascribe them undefined semantics [Adve and Boehm 2010;Adve 2008, 2012;Boehm and Demsky 2014;Manson et al 2005;Ševčík and Aspinall 2008].…”

“…Programs with data races written in shared-memory languages including Java and C++ have weak or undefined semantics, as a result of assuming data race freedom for performance reasons [Adve and Boehm 2010;Boehm and Adve 2008;Manson et al 2005]. Data races are culprits in real software failures, resulting in substantial financial losses and even harm to humans [Boehm 2011;Burnim et al 2011;Cao et al 2016;Flanagan and Freund 2010a;Kasikci et al 2012Kasikci et al , 2015Leveson and Turner 1993;Lu et al 2008;Narayanasamy et al 2007;PCWorld 2012;Sen 2008;U.S.śCanada Power System Outage Task Force 2004;Zhivich and Cunningham 2009].…”

Prescient memory: exposing weak memory model behavior by looking into the future

“…Data races are difficult to avoid, find, fix, and eliminate [1, 7, 13, 18-23, 37, 38, 41, 43, 44, 49-51]. Programmers often introduce data races intentionally for performance [9,10,28,29]. Data races and their erroneous effects are thus ubiquitous, even in mature software systems.…”

“…Since non-SC behaviors tend to manifest infrequently and unexpectedly, researchers have introduced dynamic analyses that intentionally expose non-SC behaviors allowed under weak memory models [17,24,29]. However, these dynamic analyses are limited in the kinds of behaviors they can expose.…”