Describe the feature you'd like supported

eBPF supports an arbitrary number of extensions, and at the moment there are no diagnostic events within eBPF to trace each time an extension calls batch_invoke(), batch_start(), etc. This would be useful to solve at the eBPF runtime layer, rather than each extension devising its own logging scheme.

Proposed solution

Add enter/exit traces for each eBPF entrypoint from an extension. Manifested ETW events have lower CPU and memory cost than the existing TraceLogging infrastructure within eBPF (but higher development costs) and since this is a hot and performance-sensitive code path, it may be worth setting up a manifest for hot code path diagnostics.

The diagnostics should include common ETW fields (timestamp, CPU, thread, etc.) along with unique identifiers for the eBPF program and extension.

Additional context

Helpers functions implemented within extensions should perform their own logging. It may also be useful to add logs for eBPF internal subroutines or eBPF helper functions.