A Networking-basee Linux Crash Dump

Thedebuggingofprogramisalwaysoneofbusiestthingssincethecomputerhasbeenmade. Asthe technology of computer develops rapidly, the more strict requirements to the technology ofdebuggingisraised. Insomeimportanceenvironments,itisnecessarytofigureouttherootcausein first fault analysis. This is a huge challenge to traditional technology of debugging, which isconsist of reproducing bug, continuing to track and debugging step by step. An effective wayis to dump to a file when system crashes and analysis it whenever you want. As a new kernel,main stream of Linux is always lack of a crash dump. That leaded to there are many crashdump were made, but no compatibility each other. As far as, there are many crash dump thatyou can find, such as LKCD, Netdump, Diskdump, and so on, but there are also many restrictson some architectures. In this paper, we will introduce a new crash dump called VMDump,which is network-based and constructed to C/S model. Program of capturing memory runson client and the server provides a service to receive data. Client captures memory and sendsit to server when the system of client crashes. Kernel RPC is used for the communication waybetweenclientandserver,itmaketheprogrammingmoresimpleandmoreeffectivelyinnetworktraffic. VMDump is used for special platform, Egenra's BladeFrame, it let Linux is possible toprovide a reliability service on special platform. At the same time, the dump file of VMDump iscompatible to LKCD, that means the traditional tools, such as lcrash and crash might be used toanalysis it, while no longer developing other tools. VMDump not only provides a crash dumpfor Egenera'BladeFrame, but also had a try on a new crash dump. Thus VMDump has hugeworth on practicality and theory researching.