Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Fri, 19 Jul 2019 18:27:28 +0200
From:   Christian Brauner <christian@brauner.io>
To:     Oleg Nesterov <oleg@redhat.com>
Cc:     "Joel Fernandes (Google)" <joel@joelfernandes.org>,
        linux-kernel@vger.kernel.org,
        Suren Baghdasaryan <surenb@google.com>,
        kernel-team@android.com, Andrea Arcangeli <aarcange@redhat.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        "Eric W. Biederman" <ebiederm@xmission.com>,
        Tejun Heo <tj@kernel.org>
Subject: Re: [PATCH RFC v1] pidfd: fix a race in setting exit_state for pidfd
 polling
Message-ID: <20190719162726.u5fi5k3tqove6hgn@brauner.io>
References: <20190717172100.261204-1-joel@joelfernandes.org>
 <20190719161404.GA24170@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <20190719161404.GA24170@redhat.com>
User-Agent: NeoMutt/20180716
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Jul 19, 2019 at 06:14:05PM +0200, Oleg Nesterov wrote:
> it seems that I missed something else...
> 
> On 07/17, Joel Fernandes (Google) wrote:
> >
> > @@ -1156,10 +1157,11 @@ static int wait_task_zombie(struct wait_opts *wo, struct task_struct *p)
> >  		ptrace_unlink(p);
> >  
> >  		/* If parent wants a zombie, don't release it now */
> > -		state = EXIT_ZOMBIE;
> > +		p->exit_state = EXIT_ZOMBIE;
> >  		if (do_notify_parent(p, p->exit_signal))
> > -			state = EXIT_DEAD;
> > -		p->exit_state = state;
> > +			p->exit_state = EXIT_DEAD;
> > +
> > +		state = p->exit_state;
> >  		write_unlock_irq(&tasklist_lock);
> 
> why do you think we also need to change wait_task_zombie() ?
> 
> pidfd_poll() only needs the exit_state != 0 check, we know that it
> is not zero at this point. Why do we need to change exit_state before
> do_notify_parent() ?

Oh, because of?:

	/*
	 * Move the task's state to DEAD/TRACE, only one thread can do this.
	 */
	state = (ptrace_reparented(p) && thread_group_leader(p)) ?
		EXIT_TRACE : EXIT_DEAD;
	if (cmpxchg(&p->exit_state, EXIT_ZOMBIE, state) != EXIT_ZOMBIE)
		return 0;

So exit_state will definitely be set in this scenario. Good point.

Christian