CAJ | 학술논문

操作系统内核故障往往集中分布在特定位置，其中动态内存分配和资源竞争相关代码段为典型的故障集中点，针对上述两类故障集中点，提出了一种新的基于内核动态追踪的故障监测技术，通过追踪导致内核全局数据状态迁移的方法调用，依据设计的规则对记录的调用序列和数据进行分析，实现对故障的监测和定位。监测技术在Linux操作系统中以可加载内核模块的形式实现，不需要额外硬件支持和对原系统代码进行修改。通过故障注入实验验证了监测技术的有效性，监测延时低于已有的基于时间和系统性能指标的故障监测技术。
조작계통내핵고장왕왕집중분포재특정위치，기중동태내존분배화자원경쟁상관대마단위전형적고장집중점，침대상술량류고장집중점，제출료일충신적기우내핵동태추종적고장감측기술，통과추종도치내핵전국수거상태천이적방법조용，의거설계적규칙대기록적조용서렬화수거진행분석，실현대고장적감측화정위。감측기술재Linux조작계통중이가가재내핵모괴적형식실현，불수요액외경건지지화대원계통대마진행수개。통과고장주입실험험증료감측기술적유효성，감측연시저우이유적기우시간화계통성능지표적고장감측기술。
Code segments about dynamic memory allocation and resource preemption operations are the major source of operating system faults. This paper provides a new fault detection technique based on dynamic kernel tracing, by collecting information about the call stack of kernel and global state transition, the fault source and type can be specified. The fault detection technique is implemented as loadable kernel module in Linux, which can collect kernel information effectively without additional hardware and modification of original system code. Results of fault injection experiments can prove that the proposed technique can detect faults effectively and the detection delay is smaller than former methods based on time-out and performance metrics.