Research On Reliability Evaluation Method For Linux Operating System Kernel

With the wide usage of computer systems in social life, the demand of those systems'ability of reliable and stable operations improves more and more. The kernel is the most important part in computer operating system. It includes the basic functions of operating system such as memory management, file system, task scheduling, equipment management and others. In a monolithic operating system kernel, these functions run in the form of subsystems. During runtime these subsystems share a common memory address space, making fault propagation a serious issue. Once an error occurs in the kernel, it will probably affect the whole system, so the research of the reliability of the operating system kernel is very important. Fault injection is an effective experimental method to accelerate system failure. It can provide data support to reliability evaluation of target systems by studying the failure behaviors of the target system in an environment which simulates the real situation, and collecting the information of system responses.This paper studies the method based fault injection for reliability evaluation of the operating system kernel. Firstly the architecture of the Linux operating systems is introduced, and the factors that may affect the reliability are analyzed. By introducing the development of the technology and tools of fault injection, the methods of reliability evaluation and the fault injection are analyzed in detail. With the direction of operating system kernel architecture and the characteristics of reliability property, this paper presents a fault injection tool to test the operating system kernel under memory corruption of statically allocated kernel data. By matching the source code definition of a data structure to the low-level memory address where the data is stored, it monitors the errors propagation to other subsystems, which improves the completeness and accuracy of the system evaluation. Then based on fault injection a tool for the reliability evaluation of the operating system kernel is designed and implemented.By analyzing the principle and process of fault injection, the architecture of fault injection tool for reliability evaluation of the operating system kernel is given. And the overall structures of each function are designed and developed, such as the system control module, the heartbeat detection module, the failure load generation module, the kernel memory fault injection module and the results of recovery and data analysis module, supporting different fault types and different fault injection interval. Finally a series of practical fault-injection experiments take place, verifying the effectiveness of our fault injection tool for operating system kernel and the preliminary results are also presented.