Daniel Xu

ping @caizixian

@xh4n3 can you take a look at the CI failures on master? Looks like legit errors

Maybe you could set a kprobe on the kernel exception handler?

cc @olsajiri

I think we've had some brief discussion before. IIRC a DWARF -> BTF converter was considered but IMO that's pretty hacky and not likely to work well. In my opinion,...

> What about using libdw (from elfutils) instead of libdwarf? It's supposed to be faster, has a nicer API, and we're already using libelf so we have a dependency on...

I wrote up an initial design for DWARF based stack walks: https://dxuuu.xyz/stack-symbolize.html

@michaelgugino it may be https://github.com/iovisor/bpftrace/issues/1305 (I didn't take a close look at your report yet). Maybe try using the statically linked bpftrace from our CI and see if the issue...

Also note that kprobes and kfuncs (fentry/fexit) have different semantics in the kernel. kprobes can be attached to arbitrary offsets while kfunc cannot. And (I think) kprobes get access to...

I think this would make sense for AOT. AOT could ship both branches and patch out the irrelevant branches at load time. Verifier would be OK with this too b/c...