mbox series

[RFC,v2,0/6] bpf: fprobe: rethook: Use ftrace_regs instead of pt_regs

Message ID 169139090386.324433.6412259486776991296.stgit@devnote2 (mailing list archive)
Headers show
Series bpf: fprobe: rethook: Use ftrace_regs instead of pt_regs | expand

Message

Masami Hiramatsu (Google) Aug. 7, 2023, 6:48 a.m. UTC
Hi,

Here is the 2nd version of RFC series to use ftrace_regs instead of pt_regs.
But this includes the generic part and minimum modifications of arch
dependent code. (e.g. not including rethook for arm64.) This series is based
on the discussion at

https://lore.kernel.org/all/20230801112036.0d4ee60d@gandalf.local.home/

This version includes 1 patch to expose ftrace_regs. so

 - Simply replace pt_regs in fprobe_entry_handler with ftrace_regs.
 - Expose ftrace_regs even if CONFIG_FUNCTION_TRACER=n.
 - Replace pt_regs in rethook and fprobe_exit_handler with ftrace_regs. This
   introduce a new HAVE_PT_REGS_COMPAT_FTRACE_REGS which means ftrace_regs is
   just a wrapper of pt_regs (except for arm64, other architectures do this)
 - Update fprobe-events to use ftrace_regs natively.
 - Introduce ftrace_partial_regs(). (This changes ARM64 which needs a custom
   implementation)
 - Update bpf multi-kprobe handler use ftrace_partial_regs().

Florent, feel free to add your rethook for arm64, but please do not remove
kretprobe trampoline yet. It is another discussion point. We may be possible
to use ftrace_regs for kretprobe by ftrace_partial_regs() but kretprobe
allows nest probe. (maybe we can skip that case?)

This series can also be found below branch.

https://git.kernel.org/pub/scm/linux/kernel/git/mhiramat/linux.git/log/?h=topic/fprobe-ftrace-regs

Thank you,

---

Masami Hiramatsu (Google) (6):
      fprobe: Use fprobe_regs in fprobe entry handler
      tracing: Expose ftrace_regs regardless of CONFIG_FUNCTION_TRACER
      fprobe: rethook: Use fprobe_regs in fprobe exit handler and rethook
      tracing/fprobe: Enable fprobe events with CONFIG_DYNAMIC_FTRACE_WITH_ARGS
      ftrace: Add ftrace_partial_regs() for converting ftrace_regs to pt_regs
      bpf: Enable kprobe_multi feature if CONFIG_FPROBE is enabled


 arch/Kconfig                    |    1 +
 arch/arm64/include/asm/ftrace.h |   11 ++++++
 arch/loongarch/Kconfig          |    1 +
 arch/s390/Kconfig               |    1 +
 arch/x86/Kconfig                |    1 +
 arch/x86/kernel/rethook.c       |    9 +++--
 include/linux/fprobe.h          |    4 +-
 include/linux/ftrace.h          |   56 ++++++++++++++++++-----------
 include/linux/rethook.h         |   11 +++---
 kernel/kprobes.c                |    9 ++++-
 kernel/trace/Kconfig            |    9 ++++-
 kernel/trace/bpf_trace.c        |   14 +++++--
 kernel/trace/fprobe.c           |    8 ++--
 kernel/trace/rethook.c          |   16 ++++----
 kernel/trace/trace_fprobe.c     |   76 ++++++++++++++++++++++++---------------
 kernel/trace/trace_probe_tmpl.h |    2 +
 lib/test_fprobe.c               |   10 +++--
 samples/fprobe/fprobe_example.c |    4 +-
 18 files changed, 154 insertions(+), 89 deletions(-)

--
Masami Hiramatsu (Google) <mhiramat@kernel.org>

Comments

Florent Revest Aug. 8, 2023, 2:29 p.m. UTC | #1
On Mon, Aug 7, 2023 at 8:48 AM Masami Hiramatsu (Google)
<mhiramat@kernel.org> wrote:
>
> Florent, feel free to add your rethook for arm64, but please do not remove
> kretprobe trampoline yet. It is another discussion point. We may be possible
> to use ftrace_regs for kretprobe by ftrace_partial_regs() but kretprobe
> allows nest probe. (maybe we can skip that case?)

Ack :)

>  arch/Kconfig                    |    1 +
>  arch/arm64/include/asm/ftrace.h |   11 ++++++
>  arch/loongarch/Kconfig          |    1 +
>  arch/s390/Kconfig               |    1 +
>  arch/x86/Kconfig                |    1 +
>  arch/x86/kernel/rethook.c       |    9 +++--
>  include/linux/fprobe.h          |    4 +-
>  include/linux/ftrace.h          |   56 ++++++++++++++++++-----------
>  include/linux/rethook.h         |   11 +++---
>  kernel/kprobes.c                |    9 ++++-
>  kernel/trace/Kconfig            |    9 ++++-
>  kernel/trace/bpf_trace.c        |   14 +++++--
>  kernel/trace/fprobe.c           |    8 ++--
>  kernel/trace/rethook.c          |   16 ++++----
>  kernel/trace/trace_fprobe.c     |   76 ++++++++++++++++++++++++---------------
>  kernel/trace/trace_probe_tmpl.h |    2 +
>  lib/test_fprobe.c               |   10 +++--
>  samples/fprobe/fprobe_example.c |    4 +-

I believe that Documentation/trace/fprobe.rst should also be modified
following the API change
Masami Hiramatsu (Google) Aug. 8, 2023, 2:53 p.m. UTC | #2
On Tue, 8 Aug 2023 16:29:27 +0200
Florent Revest <revest@chromium.org> wrote:

> On Mon, Aug 7, 2023 at 8:48 AM Masami Hiramatsu (Google)
> <mhiramat@kernel.org> wrote:
> >
> > Florent, feel free to add your rethook for arm64, but please do not remove
> > kretprobe trampoline yet. It is another discussion point. We may be possible
> > to use ftrace_regs for kretprobe by ftrace_partial_regs() but kretprobe
> > allows nest probe. (maybe we can skip that case?)
> 
> Ack :)
> 
> >  arch/Kconfig                    |    1 +
> >  arch/arm64/include/asm/ftrace.h |   11 ++++++
> >  arch/loongarch/Kconfig          |    1 +
> >  arch/s390/Kconfig               |    1 +
> >  arch/x86/Kconfig                |    1 +
> >  arch/x86/kernel/rethook.c       |    9 +++--
> >  include/linux/fprobe.h          |    4 +-
> >  include/linux/ftrace.h          |   56 ++++++++++++++++++-----------
> >  include/linux/rethook.h         |   11 +++---
> >  kernel/kprobes.c                |    9 ++++-
> >  kernel/trace/Kconfig            |    9 ++++-
> >  kernel/trace/bpf_trace.c        |   14 +++++--
> >  kernel/trace/fprobe.c           |    8 ++--
> >  kernel/trace/rethook.c          |   16 ++++----
> >  kernel/trace/trace_fprobe.c     |   76 ++++++++++++++++++++++++---------------
> >  kernel/trace/trace_probe_tmpl.h |    2 +
> >  lib/test_fprobe.c               |   10 +++--
> >  samples/fprobe/fprobe_example.c |    4 +-
> 
> I believe that Documentation/trace/fprobe.rst should also be modified
> following the API change

Indeed. Let me update it.

Thanks!