Re: [RFC PATCH -tip 0/9] kprobes: Cleanup jprobe implementation

[Masami Hiramatsu]

On Sat, 10 Mar 2018 09:24:44 +0100
Ingo Molnar <mingo@xxxxxxxxxx> wrote:

> 
> * Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:
> 
> > Hello,
> > 
> > Since we decided to remove jprobe from kernel last year,
> > its APIs are disabled and we worked on moving in-kernel
> > jprobe users to kprobes or trace-events. And now no jprobe
> > users are here anymore.
> > 
> > I think it is good time to get rid of jprobe implementation
> > from the kernel. However, I need other arch developers help
> > to complete it, since jprobe is implemented multi arch wide.
> > I can remove those code, but can not test all of those.
> > 
> > Here is the series of patches to show how to do that.
> > I tried to remove it from x86 tree. Basically we need to
> > do 3 things;
> > 
> >  - Remove jprobe functions (register/unregister,
> >    setjump/longjump) from generic/arch-dependent code.
> >    [1/9][2/9][3/9]
> >  - Remove break_handler related code.
> >    [4/9][5/9][6/9]
> >  - Do not disable preemption on exception handler
> >    [7/9][8/9][9/9]
> > 
> > The [3/9] and [6/9] are destractive changes except for x86
> > (means causes build errors) since those arch still have some
> > references of those functions. So we need to write patches
> > similar to [2/9] and [5/9] for each arch before applying those.
> > In this series I sorted it as this order just for review,
> > [3/9] and [6/9] should be applied after all archs have
> > been fixed.
> > 
> > Also, [7/9] is a kind of destractive, which changes required
> > behavior for the pre_handlers which changes regs->ip.
> > So we also need a patch similar to [7/9] for each arch too.
> > Fortunately, current in-tree such user is very limited, both
> > works only on x86. So it is not hurry, but we need to change
> > arch dependent code.
> >
> > Thank you,
> > 
> > ---
> > 
> > Masami Hiramatsu (9):
> >       kprobes: Remove jprobe API implementation
> >       x86: kprobes: Remove jprobe implementation
> >       kprobes: Remove jprobe data structure and interfaces
> >       kprobes: Ignore break_handler
> >       x86: kprobes: Ignore break_handler
> >       kprobes: Remove break_handler from struct kprobe
> >       x86: kprobes: Do not disable preempt on int3 path
> >       error-injection: Fix to not enabling preemption in pre_handler
> >       tracing: kprobes: Fix to not enabling preemption
> > 
> > 
> >  Documentation/kprobes.txt        |   13 ++--
> >  arch/x86/include/asm/kprobes.h   |    3 -
> >  arch/x86/kernel/kprobes/common.h |   10 ---
> >  arch/x86/kernel/kprobes/core.c   |  114 ++------------------------------------
> >  arch/x86/kernel/kprobes/ftrace.c |   21 +------
> >  arch/x86/kernel/kprobes/opt.c    |    1 
> >  include/linux/kprobes.h          |   54 ------------------
> >  kernel/fail_function.c           |    1 
> >  kernel/kprobes.c                 |  115 ++------------------------------------
> >  kernel/trace/trace_kprobe.c      |    3 -
> >  10 files changed, 20 insertions(+), 315 deletions(-)
> 
> Nice work, and I love the diffstat, but please do a series that works (builds and 
> boots and has working kprobes) at every interim step.

Thank you! Yes, I'm thinking just keep header files no change (no removing
API and structure) until all archs apply C file changes in next version,
so that we can avoid any build error/test error.

Thanks,



> 
> Thanks,
> 
> 	Ingo


-- 
Masami Hiramatsu <mhiramat@xxxxxxxxxx>