FAST AND ACCURATE DATA RACE DETECTION FOR CONCURRENT PROGRAMS WITH ASYNCHRONOUS CALLS

• US 20100235817A1
• Filed: 02/08/2010
• Published: 09/16/2010
• Est. Priority Date: 03/11/2009
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Laboratories America Inc (NEC Corporation)
• Inventors: Zhang, Yun, Kruus, Eric J., Sinha, Nishant, KAHLON, VINEET
• Granted Patent: US 8,539,450 B2
• Time in Patent Office: Days
• US Class Current: 717/126
• CPC Class Codes: G06F 11/3604 Software analysis for verifying properties of programs byte-code verification G06F9/44589
• CPC Class Codes: G06F 8/458 Synchronisation, e.g. post-wait, barriers, locks synchronisation among tasks G06F9/52

Abstract

A system and method for analyzing a concurrent program employ asynchronous function calls for communication and recursion. A control flow graph is constructed based on a context-sensitive pointer analysis, whereupon encountering a function pointer, a points-to set of the function pointer is computed in a context-sensitive fashion to determine a set of potential function calls. The context-sensitive pointer analysis is terminated when no new potential function calls are encountered and where the potential function calls may contribute new data races other than those that exist in the contexts traversed thus far. To decide this, a characterization of pointer aliasing based upon complete update sequences is employed. A set of contexts that may contribute to different data races are enumerated by tracking update sequences for function and lock pointers and pointers that are shared or point to shared memory locations. Data race detection is carried out on the control flow graph.

Claims

1. A context-sensitive method for analyzing a concurrent program stored in memory, which employs asynchronous function calls for communication and/or recursion, comprising:

constructing a control flow graph, using a processor, based on a context-sensitive pointer analysis, wherein upon encountering a function pointer, a points-to set of the function pointer is resolved on-the-fly in a context-sensitive fashion to accurately determine a set of potential function calls;

terminating the context-sensitive control flow graph construction when no new potential function calls are encountered that potentially contribute new data races that are not already discovered in existing contexts; and

detecting data races in the concurrent program using the context-sensitive control flow graph which is constructed to not miss any potential data race.

2. The method as recited in claim 1, wherein function pointers encountered are resolved on-the-fly using a characterization of pointer aliasing based upon complete update sequences to decide whether new aliases are discoverable

3. The method as recited in claim 1, wherein the context-sensitive construction of the control flow graph is terminated based on whether new shared variables accesses with different locksets other than those encountered in existing contexts can be discovered.

4. The method as recited in claim 3, wherein a fixpoint computation is carried out via traversal of the control flow graph of the program by tracking aliases of lock pointers, aliases of shared variables and aliases of function pointers to decide whether new shared variable accesses with different locksets than those already encountered are discoverable.

5. A computer readable storage medium comprising a computer readable program for analyzing a concurrent program stored in memory, which employs asynchronous function calls for communication and/or recursion, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:

constructing a control flow graph, using a processor, based on a context-sensitive pointer analysis, wherein upon encountering a function pointer, a points-to set of the function pointer is resolved on-the-fly in a context-sensitive fashion to accurately determine a set of potential function calls;

terminating the context-sensitive control flow graph construction when no new potential function calls are encountered that potentially contribute new data races that are not already discovered in existing contexts; and

detecting data races in the concurrent program using the context-sensitive control flow graph which is constructed to not miss any potential data race.

6. The computer readable storage medium as recited in claim 5, wherein function pointers encountered are resolved on-the-fly using a characterization of pointer aliasing based upon complete update sequences to decide whether new aliases are discoverable

7. The computer readable storage medium as recited in claim 5, wherein the context-sensitive construction of the control flow graph is terminated based on whether new shared variables accesses with different locksets other than those encountered in existing contexts can be discovered.

8. The computer readable storage medium as recited in claim 7, wherein a fixpoint computation is carried out via traversal of the control flow graph of the program by tracking aliases of lock pointers, aliases of shared variables and aliases of function pointers to decide whether new shared variable accesses with different locksets than those already encountered are discoverable.

9. A system for analyzing a concurrent program stored in memory, which employs asynchronous function calls for communication and/or recursion, comprising:

a program application stored on program storage media and configured to construct a control flow graph for a concurrent program being analyzed, the control flow graph being constructed based on a context-sensitive pointer analysis executed by the program application, wherein upon encountering a function pointer in the concurrent program, a points-to set of the function pointer is computed in a context-sensitive fashion to accurately determine a set of potential function calls, the context-sensitive pointer analysis being terminated when no new potential function calls are encountered such that no new potential function calls are determined using a characterization of pointer aliasing encountered based upon complete update sequences to decide whether new aliases are discoverable; and

a processor configured to detect data races in the concurrent program using the control flow graph which includes resolved context sensitive pointers.

10. The system as recited in claim 9, wherein the resolved context sensitive pointers of the control flow graph include one or more of asynchronous calls, function pointers, thread pointers and recursion.

11. The system as recited in claim 9, wherein the control flow graph includes a thread pool where each thread iteratively processes tasks from its work queue, wherein the thread pool sends task requests to other threads to issue an asynchronous function call with the task as an argument.

12. The system as recited in claim 11, wherein asynchronous function calls include fork and join events.