Implementing Kdump for PPC64 - the nuts and bolts
Kexec is a technology present in recent Linux kernels, which allows the kernel to boot another kernel on top of itself without the help of firmware. The kernel acts as a bootloader for itself. Kdump, built on top of Kexec, will be the last word in crash dump solutions. The idea is that if the kernel crashes, it boots another kernel using Kexec, and the second kernel then performs a dump of the first kernel's memory to disk or network etc. The second kernel may even start gdb on the crashed kernel for realtime inspection. In this talk I will be covering the nitty-gritty details required to get Kdump happening on PPC64. Anyone who thought there weren't enough technical talks at LCA 2005 should pencil this one in. There'll be linker tricks, virtual and real mode fun, firmware interactions, IPIs, we're even gonna stare at some PPC64 assembler. I am currently part of the team actively developing the PPC64 Kdump code (http://patchwork.ozlabs.org/linuxppc64/patch?id=2155), so some of the details are still sketchy - by LCA 2006 the code will be complete but the details should still be fresh in my mind.
First started programming C sometime during highschool, using a borrowed copy of Borland Turbo C. Completed the four year Bachelor of Software Engineering, with first class Honours at ANU. Worked on software for Fedsat (http://www.auspace.com.au/projects/fedsat.htm) as part of my final year project. Spent two years at HIC working for IBM in a UNIX support role. Managed 80+ AIX & Linux servers with ~4-5 staff and a lot of scripts. Moved to IBM OzLabs in early 2004 to work on PPC64 kernel development.