Re: [PATCH 2/5] sched,ptrace: Fix ptrace_check_attach() vs PREEMPT_RT

[Peter Zijlstra <peterz@infradead.org>]

On Fri, Apr 15, 2022 at 12:57:56PM +0200, Oleg Nesterov wrote:
> On 04/15, Oleg Nesterov wrote:
> >
> > OK, so far it seems that this patch needs a couple of simple fixes you
> > pointed out, but before I send V2:
> >
> > 	- do you agree we can avoid JOBCTL_TRACED_FROZEN in 1-2 ?
> >
> > 	- will you agree if I change ptrace_freeze_traced() to rely
> > 	  on __state == TASK_TRACED rather than task_is_traced() ?
> >
> 
> Forgot to say, I think 1/5 needs some changes in any case...
> 
> ptrace_resume() does  wake_up_state(child, __TASK_TRACED) but doesn't
> clear JOBCTL_TRACED. The "else" branch in ptrace_stop() leaks this flag
> too. Perhaps I missed something, I'll reread 1/5 again, but the main
> question to me is whether 1-2 actually need the JOBCTL_TRACED_FROZEN flag.

Ok, getting back to this. So I did the change to ptrace_resume(), but
I'm not entirely sure I understand the issue with the else branch of
ptrace_stop().

My understanding is that if we hit that else branch, we've raced wth
__ptrace_unlink(), and that will have done:

  if (... || task_is_traced(child))
	ptrace_signal_wake_up(child, true);

Which will have done that wakeup and cleared both __state and jobctl.