The Design And Implementation Of Linux Kernel Data Structure Identification System

In the development of the Linux operating system,developers may encounter kernel crash and memory leak.It will cost a lot of person-months to analysis or locate the crash point,which may lead to the waste of resources and low efficiency.Also,it request the developers the experience of kernel development to resolve the memory leak or crash problems,which increase the difficulty of the kernel debugging and resolving the kernel crash problem.For the memory leak of the k ernel data structures,this thesis proposes a solution that based on deep learning and image identification.it can be used to identify kernel data structures quickly,which can be useful to locate the crash point.Linux kernel data structure identification system uses LTP to trigger the creation and free of the kernel data structures.This thesis uses the kernel patch to collect data structures as training set,then transfers the data into image type,then uses the CNN to train the model.Finally this thesis uses the model to identify the data structures,in the purpose of imporving the efficiency of locating the crash point.Linux kernel data structure identification system can identify the unknown data structures quickly,which can improve the efficiency of repairing and locating the kernel crash problems.The system have trained ten categories of data structures which contain 5000 data each category.The accuracy can be 99%more and it can be used to identify the data structures in the ten categories.