T. Arai, Y. Sugita, M. Hiramatsu, and K. Serizawa (Japan)
Software testing, OS debugging, kernel trace, Linux
Debugging an operating system (OS) is a difficult job because a bug in the OS is usually both timing dependent and hardware dependent. The bug can rarely be detected by simply dumping memory when the system crashes because the cause of the crash usually precedes the crash itself. We therefore investigated the information required for OS debugging and enumerated the events generating that information. We then developed a Linux kernel status tracer (LKST) that traces the sequence of events that occur in a Linux kernel and provides enough information for OS debugging. LKST hooks the events about which information that will be useful for kernel debugging needs to be collected and stores it in event buffers. It also allows kernel programmers and debuggers to consider the overhead and data efficiency trade-offs and select the data they need for debugging accordingly.
Important Links:
Go Back
