Unsupported CONFIG_FPROBE and CONFIG_RETHOOK on ARM64
Masami Hiramatsu (Google)
mhiramat at kernel.org
Tue Sep 10 17:39:49 PDT 2024
On Tue, 10 Sep 2024 13:29:57 -0700
Andrii Nakryiko <andrii.nakryiko at gmail.com> wrote:
> You are probably talking about [0]. But I was asking about [1], i.e.,
> adding HAVE_RETHOOK support to ARM64. Despite all your emotions above,
> can I still get a meaningful answer as for why that wasn't landed and
> what prevents it from landing right now before Masami's 20-patch
> series lands?
As I replied to your last email, Mark discovered that [1] is incorrect.
From the bpf perspective, it may be fine that struct pt_regs is missing
some architecture-specific registers, but from an API perspective,
it is a problem.
Actually kretprobes on arm64 still does not do it correctly, but I also
know most of users does not care. So currently I keep it as it is. But
after fixing this issue on fprobe. I would like to update kretprobe so
that it will use sw-breakpoint to handle it. It will increase the overhead
of kretprobe, but it should be replaced by fprobe at that moment.
Thank you,
>
> [0] https://lore.kernel.org/linux-trace-kernel/172398527264.293426.2050093948411376857.stgit@devnote2/
> [1] https://lore.kernel.org/bpf/164338038439.2429999.17564843625400931820.stgit@devnote2/
>
> >
> > Again, just letting you know.
> >
> > -- Steve
--
Masami Hiramatsu (Google) <mhiramat at kernel.org>
More information about the linux-arm-kernel
mailing list