Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 25 Apr 2022 19:47:20 +0200
From:   Oleg Nesterov <oleg@redhat.com>
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     rjw@rjwysocki.net, mingo@kernel.org, vincent.guittot@linaro.org,
        dietmar.eggemann@arm.com, rostedt@goodmis.org, mgorman@suse.de,
        ebiederm@xmission.com, bigeasy@linutronix.de,
        Will Deacon <will@kernel.org>, linux-kernel@vger.kernel.org,
        tj@kernel.org, linux-pm@vger.kernel.org
Subject: Re: [PATCH v2 2/5] sched,ptrace: Fix ptrace_check_attach() vs
 PREEMPT_RT
Message-ID: <20220425174719.GB12412@redhat.com>
References: <20220421150248.667412396@infradead.org>
 <20220421150654.817117821@infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220421150654.817117821@infradead.org>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk

On 04/21, Peter Zijlstra wrote:
>
> @@ -2225,7 +2238,7 @@ static int ptrace_stop(int exit_code, in
>  	 * schedule() will not sleep if there is a pending signal that
>  	 * can awaken the task.
>  	 */
> -	current->jobctl |= JOBCTL_TRACED;
> +	current->jobctl |= JOBCTL_TRACED | JOBCTL_TRACED_QUIESCE;
>  	set_special_state(TASK_TRACED);

OK, this looks wrong. I actually mean the previous patch which sets
JOBCTL_TRACED.

The problem is that the tracee can be already killed, so that
fatal_signal_pending(current) is true. In this case we can't rely on
signal_wake_up_state() which should clear JOBCTL_TRACED, or the
callers of ptrace_signal_wake_up/etc which clear this flag by hand.

In this case schedule() won't block and ptrace_stop() will leak
JOBCTL_TRACED. Unless I missed something.

We could check fatal_signal_pending() and damn! this is what I think
ptrace_stop() should have done from the very beginning. But for now
I'd suggest to simply clear this flag before return, along with
DELAY_WAKEKILL and LISTENING.

>  	current->jobctl &= ~JOBCTL_LISTENING;
> +	current->jobctl &= ~JOBCTL_DELAY_WAKEKILL;

	current->jobctl &=
		~(~JOBCTL_TRACED | JOBCTL_DELAY_WAKEKILL | JOBCTL_LISTENING);

Oleg.