lkml.org 
[lkml]   [2021]   [Mar]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH -tip 3/5] kprobes: treewide: Remove trampoline_address from kretprobe_trampoline_handler()
Hi Miroslav,

On Thu, 11 Mar 2021 00:42:25 +0900
Masami Hiramatsu <mhiramat@kernel.org> wrote:

> > > + */
> > > +static nokprobe_inline void *kretprobe_trampoline_addr(void)
> > > +{
> > > + return dereference_function_descriptor(kretprobe_trampoline);
> > > +}
> > > +
> >
> > Would it make sense to use this in s390 and powerpc reliable unwinders?
> >
> > Both
> >
> > arch/s390/kernel/stacktrace.c:arch_stack_walk_reliable()
> > arch/powerpc/kernel/stacktrace.c:__save_stack_trace_tsk_reliable()
> >
> > have
> >
> > if (state.ip == (unsigned long)kretprobe_trampoline)
> > return -EINVAL;
> >
> > which you wanted to hide previously if I am not mistaken.
>
> I think, if "ip" means "instruction pointer", it should point
> the real instruction address, which is dereferenced from function
> descriptor. So using kretprobe_trampoline_addr() is good.

Ah, sorry I misunderstood the question.

Yes, the per-arch stacktrace implementation must be fixed afterwards.
It is reliable or not depends on the actual unwinder implementation,
for example, on x86, frame-pointer based unwinder can unwind kretprobe,
but ORC based one doesn't (and discussing with Josh how to solve it)

Anyway since it strongly depends on the architecture, I would like to
leave those for each architecture stacktrace maitainer in this series.

Thank you,

--
Masami Hiramatsu <mhiramat@kernel.org>

\
 
 \ /
  Last update: 2021-03-11 14:39    [W:0.074 / U:0.064 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site