Re: Getting empty callchain from perf_callchain_kernel()

[Peter Zijlstra]

On Thu, May 23, 2019 at 10:24:13AM -0500, Josh Poimboeuf wrote:

> Here's the latest version which should fix it in all cases (based on
> tip/master):
> 
>   https://git.kernel.org/pub/scm/linux/kernel/git/jpoimboe/linux.git/commit/?h=bpf-orc-fix

That patch suffers an inconsitency, the comment states:

  'if they have "jump_table" in the name'

while the actual code implements:

  'if the name starts with "jump_table"'

Other than that, I suppose that works just fine ;-)

> There's no need to put special cases in the FP unwinder when we can
> instead just fix the frame pointer usage in the JIT code.
> 
> For ORC, I'm thinking we may be able to just require that all generated
> code (BPF and others) always use frame pointers.  Then when ORC doesn't
> recognize a code address, it could try using the frame pointer as a
> fallback.

Yes, this seems like a sensible approach. We'd also have to audit the
ftrace and kprobe trampolines, IIRC they only do framepointer setup for
CONFIG_FRAME_POINTER currently, which should be easy to fix (after the
patches I have to fix the FP generation in the first place:

  https://git.kernel.org/pub/scm/linux/kernel/git/peterz/queue.git/log/?h=x86/wip
)