Re: [Bug 10489] Kprobe smoke test lockdep warning

[Masami Hiramatsu <mhiramat@redhat.com>]

Hi Peter,

Thank you for the patch,

Peter Zijlstra wrote:
> So the below makes it work - but I'm not getting this code; at the time
> I wrote that stuff I CC'ed each and every kprobe maintainer listed in
> the usual places but got no reposonse - can some please explain this
> stuff to me?

Sure, I came to the same decision. It is enough to remove unnecessary
trace_hardirqs_fixup_flags() from post_kprobe_handler. pt_regs->flags is
just a variable on stack -- not effective flags in that time. So we don't
need to change thread's hardirq flag.

> Are the saved flags only for the TF bit or are they made in full effect
> later (and if so, where) ?

The saved flags is for TF and IF bits.
Actually, kprobes does following:

---
do_int3()
 kprobe_exceptions_notify()
  kprobe_handler()
   prepare_singlestep() : save regs->flags, set TF and clear IF on regs->flags
"iret" : pop regs->flags image to EFLAGS reg.

<single step an instruction>

do_debug()
 kprobe_exceptions_notify()
  post_kprobe_handler() : restore regs->flags (clearing TF and restoring IF)
"iret" : pop regs->flags image to EFLAGS reg.
---

While executing a single step, EFLAGS is different from the original EFLAGS.
However, it will be recovered soon by post_kprobe_handler.
NOTE: if the single-stepping instruction is a pushf, kprobes writes saved
     EFLAGS to the top of stack.

> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>

I tested it on x86-64/i386.

Acked-by: Masami Hiramatsu <mhiramat@redhat.com>

> ---
> diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c
> index 34a5912..fe71e64 100644
> --- a/arch/x86/kernel/kprobes.c
> +++ b/arch/x86/kernel/kprobes.c
> @@ -865,7 +865,6 @@ static int __kprobes post_kprobe_handler(struct pt_regs *regs)
>  
>         resume_execution(cur, regs, kcb);
>         regs->flags |= kcb->kprobe_saved_flags;
> -       trace_hardirqs_fixup_flags(regs->flags);
>  
>         /* Restore back the original saved kprobes variables and continue. */
>         if (kcb->kprobe_status == KPROBE_REENTER) {
> 
> 
> 

-- 
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America) Inc.
Software Solutions Division

e-mail: mhiramat@redhat.com