Speaker
Description
One pain point that is emerging more and more as live patching becomes more common on systems is that although there is coordination between ftrace and live patching (for live patching approaches that use it), the same cannot be said for kprobe and kretprobe. Sudden lack of trace data being generated is often the only indication that the function being traced is being bypassed due to live patching.
It would be nice if live patching of a function could trigger "something" to allow tracing tools the ability to detect that certain probes are no longer going to be useful. The presenter has a few ideas on the topic to spark further discussion in this area and to hopefully come out of this session with some concrete ideas on how to proceed.
It is understood that no solution will be perfect, but we can try to improve the current situation.