| 摘要 |
Detecting and/or tolerating races. Races occur due to malicious threads not respecting software locks. A method of detecting and/or correcting races includes making local copie(s) and reference copie(s) of shared data. Any read and write operations performed by a safe thread are caused to be performed on the local copie(s) during a critical section. The critical section defines a time frame which a variable lock is placed on shared data. Any read and write operations performed by malicious threads are allowed to be performed on the shared data during the critical section. The shared data, the local copie(s), and the reference copie(s) are compared to determine that a race has been detected. An indication can be output that a race has occurred or the race can be corrected.
|
