Kernel Live Patching at Scale

Explore the intricacies of Kernel Live Patching (KLP) at scale in this comprehensive conference talk from the Linux Plumbers Conference. Delve into the challenges and solutions of applying fixes without system reboots, replacing ftrace-able kernel functions, and patching multiple functions atomically. Understand how KLP is utilized in enterprise distributions to fix bugs while maintaining system uptime. Learn about the potential conflicts between KLP and tracing tools, and examine a KLP transition failure example. Discover strategies to proactively identify and address corner cases, as well as new features being added to the tool chain to reduce KLP transition failures. Gain insights into the importance of tracing in data centers and how it intersects with live patching techniques.

Kernel Live Patching at Scale
Applying fix via a full reboot
Replace ftrace-able kernel functions
Patch multiple functions atomically
KLP is common among enterprise distros -Fix bugs without rebooting the system
How: Homogeneous configuration
insmod task was hogging CPU - starving ksoftirqd
Tracing is first class citizen in data centers
KLP may conflict with tracing tools
KLP transition failure example
Proactively identify and fix corner cases
Add new features to the tool chain
Reduce KLP transition failures