Linaro Connect Bangkok

Linaro has been building on KernelCI to handle continuous integration of the Linux kernel with multiple different compilers and compiler versions. This is used for catching regressions upstream in the Linux kernel and LLVM code bases. This helps ensure that Android and ChromeOS can reliably ship LTS branches of the kernel built with Clang. Come learn more about building Linux kernels with Clang, and how Linaro is helping enable this work via KernelCI.

"I could talk to you all day kernel debugging. Really! In fact I, along with my colleague Leo, have spent are large portion of our time recently doing exactly that. However I don't have all day... I have just 25 minutes... and no slides."

In this session Daniel will demonstrate live a some of the ways to exploit both static and dynamic tracepoints to study kernel behaviour. We'll start out using just the basic tools available in even tiny busybox distribution before expanding our toolkit very slightly by copying a couple of extra binaries onto the system under debug.

Continuation of discussion from YVR18. Qualcomm kernel upstream status, issues, roadblocks, planning.

While Virtual Machines have been around for many decades containers are a relatively new development. Their usage has grown rapidly as users have started designing solutions using swarms of micro services in on-demand clouds managed by orchestration systems. At the same time containers are really just a group of processes sharing a host kernel which has led to concerns about security and isolation if things go wrong. Efforts are now underway to bring the strong isolation of virtual machines into the free-wheeling world of rapidly updated
containerised applications.

This talk will give an overview of the two technologies and how they are being brought together to provide the best of both worlds. This includes topics such as making orchestration systems VM aware as well as projects to move container run times into specialised virtual machines. We will also discuss what else needs to be done to enable the ARM eco-system to take advantage of these two complimentary technologies.

As network interfaces keep getting faster drivers, CPUs and memory need to keep up. Linux kernel network drivers usually allocate memory on-the-fly, process the packets and then free that memory. This talk focuses on a proposed RFC about a native memory recycling strategy that applies on both the default network stack, XDP(eXpress Data Path) and the improvements it offers compared to the existing approach.

DescriptionThis presentation is a report on the last improvements on the BFQ I/O scheduler. These improvements benefit virtually any system, from embedded devices, to personal systems, to nodes in a data center.

This first set of changes concerns throughput. In the most complex scenarios for guaranteeing I/O bandwidths, BFQ delivers up to five-time higher throughput than existing solutions. But the same mechanisms that gained BFQ this primacy become a hindrance with some 'deceptive' workloads. These workloads trick BFQ mechanisms into wrongly believing that some I/O flows need to be privileged with respect to other flows, even at the expense of losing throughput dramatically. In contrast, total throughout is the only performance parameter that matters. We took countermeasures to offset this loss of throughput, countermeasures that fully succeed with some 
unfriendly workloads.

Then, as for I/O control, the combination of several new improvements and fixes let the worst-case start-up time of applications drop by an additional 35%. We show these results not only through graphs, but also through a new demo with a Chromebook.

The last contributions shown in this presentation are about
efficiency. In fact, even the execution overhead of an I/O scheduler may limit maximum throughput with very fast drives. So, to reduce BFQ overhead, we tried to turn some properties of these drives into BFQ's advantage: we looked for costly optimizations that are no longer necessary with these drives. We found some, and added controls that automatically turn them off when not needed.