Every year, we see a few new architecture ports getting merged into the Linux kernel, but unfortunately each one copies the same bugs from the existing architectures, creating a huge mess in system call interface differences and implementation bugs. In an attempt to simplify the work for future architecture ports, we will try to provide generic implementations for almost all architecture specific functions, and describe the minimally necessary implementation for a new port.

Linus originally wrote Linux as a terminal emulator on his 386, but in the 17 years since, it has grown to be the most portable operating system every written. This session examines the implications of this widespread porting, and the issues faced by hackers writing new code, whether it be new drivers, adding a new system call, or some new piece of user-space code.

A variety of specific examples will be investigated, ranging from how issues such as PCI write queues or IOMMUs affect driver authors, to how the various calling conventions of some of the more esoteric architectures impact the design of new syscalls, to how seemingly reasonable user-space code can SIGBUS when run on a new platform.

The overarching theme of this talk will be how these ports have improved the code quality of Linux through abstracting architecture dependent parts out of common code. Also explored will be how the RISC architectures of yesterday have enabled the mainstream (read: x86) architectures to stay ahead of the curve with new hardware.

In the quest to reduce costs there is an ever increasing trend towards virtualisation within corporate IT organisations. In this presentation I'll discuss how Transitive are working with the Linux KVM project to build a hypervisor aware system level translator, detail some of the technology behind our QuickTransit translator and how this all fits together to create cross-platform virtual machines. Followed by a short demonstration.

You hear the names of the most common web security problems frequently: cross-site scripting, SQL injection, cross-site request forgery, but what do those terms mean? This presentation will provide an introduction to those vulnerabilities along with examples and ways to avoid them. The presentation will be language agnostic as the problems can occur in any language used to develop web applications.

Developers of web applications sometimes get caught up in the excitement of developing the application and forget to consider the security implications. This presentation will help them get a handle on what to avoid so that the excitement doesn't get squashed by an attacker. Others who are curious about the kinds of attacks made against web applications will also find much of interest.

This talk tells the story of a filesystem with a hard childhood growing leaner and better over the years.

Born in 1993 XFS first had a protected childhood in the arms of a proprietary Unix environment, but got re-parented into the crazy creative chaos of the Linux community around 2000. It became a productive member of the society soon after.

After 15 years it's now time to heal the wounds the sudden change of environment caused, and time for some massive muscle buildup to keep the news kids of it's patch in high-end file serving and HPC. It's also slowly growing self-healing capabilities to withstand the toxic environment of modern storage subsystems.

Suspend and resume may be the most obvious components of power management, but in many ways they are the least important. The overall power consumption of a system is related to battery life, electricity bills and heat production. Reducing the power draw of a system at runtime improves all of these. The trick is in doing it in such a way that the user doesn't notice.

This presentation covers the runtime power management functionality available in Linux, ranging from CPU voltage scaling through LVDS reclocking to USB autosuspend. How can you act to reduce your bills now, and how are we going to make things even better in the future?