A Dynamic Memory Leak Detection Method for Linux Based on eBPF Uprobe and LKM

Existing dynamic memory leak detection methods typically require starting the target application in a monitored environment, which poses diagnostic challenges in scenarios where the program is already running and cannot be easily stopped. Additionally, current detection methods struggle to differentiate between temporarily unreleased memory allocations and those that genuinely cause memory leaks, increasing the difficulty for administrators to troubleshoot. To address these challenges, this paper proposes a dynamic memory leak detection method based on eBPF uprobe and LKM. This approach combines the flexible data collection capabilities of eBPF uprobe with the kernel-level monitoring capabilities of LKM, enabling real-time monitoring and analysis of memory allocation and usage, dynamically tracking memory leaks, and accurately identifying the source of the leaks. Compared to existing methods, this approach does not require special configuration of the application, allows for memory leak detection at any time during program execution, and distinguishes memory allocations at program exit, thereby enhancing the flexibility, accuracy, and practicality of the detection process.