There are currently a few different implementations on how to do live kernel patching in the kernel. That is, the ability to apply a patch to the kernel while it is still running, requiring no reboot. The problem is that the kernel should only have one implementation. Although userspace may be different, the more the various implementations can cooperate the better.

Not only is this an issue of cooperation, but there are some serious issues with live patching itself. What is considered safe? Can anything except for trivial buffer overflows and off by one errors be corrected. How long should a patched kernel be running before it should be rebooted.

• Brief 10 minute presentations
  • criu+kexec
  • kGraft
  • kpatch
• What is needed of ftrace for kernel patching
• How much state can be saved via checkpoint/restart and kexec
• What features are needed from a live patching solution
• kpatch vs kGraft
  • Design comparison
  • Feature comparison (kernel)
  • Is there any way to combine the kernel implementations?
  • Feature comparison (user space)
  • Is there any way to combine the user space implementations?
• kpatch/kGraft vs criu+kexec
  • Feature comparison
  • Do we need both?

There may be brief 10 minute presentations to describe each implementation. Then the rest of the time should be going over the topics.

• Josh Poimboeuf
• Seth Jennings
• Steven Rostedt
• Jiri Kosina
• Jiri Slaby
• Masami Hiramatsu

• James Bottomely
• Vojtech Pavlik
• Thomas Gleixner
• Petr Mladek [ will be having new borns to deal with ]
• Praneeth Bajjuri

• Josh Poimboeuf
• Seth Jennings
• Steven Rostedt
• James Bottomely
• Jiri Kosina
• Jiri Slaby
• Vojtech Pavlik
• Thomas Gleixner to tell everyone how crappy their ideas are