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;
From:   "Eric W. Biederman" <ebiederm@xmission.com>
To:     Oleg Nesterov <oleg@redhat.com>
Cc:     linux-kernel@vger.kernel.org, rjw@rjwysocki.net, mingo@kernel.org,
        vincent.guittot@linaro.org, dietmar.eggemann@arm.com,
        rostedt@goodmis.org, mgorman@suse.de, bigeasy@linutronix.de,
        Will Deacon <will@kernel.org>, tj@kernel.org,
        linux-pm@vger.kernel.org, Peter Zijlstra <peterz@infradead.org>,
        Richard Weinberger <richard@nod.at>,
        Anton Ivanov <anton.ivanov@cambridgegreys.com>,
        Johannes Berg <johannes@sipsolutions.net>,
        linux-um@lists.infradead.org, Chris Zankel <chris@zankel.net>,
        Max Filippov <jcmvbkbc@gmail.com>,
        inux-xtensa@linux-xtensa.org, Kees Cook <keescook@chromium.org>,
        Jann Horn <jannh@google.com>
References: <878rrrh32q.fsf_-_@email.froward.int.ebiederm.org>
        <20220426225211.308418-9-ebiederm@xmission.com>
        <20220427154158.GG17421@redhat.com>
Date:   Wed, 27 Apr 2022 17:35:05 -0500
In-Reply-To: <20220427154158.GG17421@redhat.com> (Oleg Nesterov's message of
        "Wed, 27 Apr 2022 17:41:59 +0200")
Message-ID: <87mtg62m06.fsf@email.froward.int.ebiederm.org>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Subject: Re: [PATCH 9/9] ptrace: Don't change __state
Precedence: bulk

Oleg Nesterov <oleg@redhat.com> writes:

2> On 04/26, Eric W. Biederman wrote:
>>
>>  static void ptrace_unfreeze_traced(struct task_struct *task)
>>  {
>> -	if (READ_ONCE(task->__state) != __TASK_TRACED)
>> +	if (!(READ_ONCE(task->jobctl) & JOBCTL_DELAY_WAKEKILL))
>>  		return;
>>
>>  	WARN_ON(!task->ptrace || task->parent != current);
>> @@ -213,11 +213,10 @@ static void ptrace_unfreeze_traced(struct task_struct *task)
>>  	 * Recheck state under the lock to close this race.
>>  	 */
>>  	spin_lock_irq(&task->sighand->siglock);
>
> Now that we do not check __state = __TASK_TRACED, we need lock_task_sighand().
> The tracee can be already woken up by ptrace_resume(), but it is possible that
> it didn't clear DELAY_WAKEKILL yet.

Yes.  The subtle differences in when __TASK_TRACED and
JOBCTL_DELAY_WAKEKILL are cleared are causing me some minor issues.

This "WARN_ON(!task->ptrace || task->parent != current);" also now
needs to be inside siglock, because the __TASK_TRACED is insufficient.


> Now, before we take ->siglock, the tracee can exit and another thread can do
> wait() and reap this task.
>
> Also, I think the comment above should be updated. I agree, it makes sense to
> re-check JOBCTL_DELAY_WAKEKILL under siglock just for clarity, but we no longer
> need to do this to close the race; jobctl &= ~JOBCTL_DELAY_WAKEKILL and
> wake_up_state() are safe even if JOBCTL_DELAY_WAKEKILL was already
> cleared.

I think you are right about it being safe, but I am having a hard time
convincing myself that is true.  I want to be very careful sending
__TASK_TRACED wake_ups as ptrace_stop fundamentally can't handle
spurious wake_ups.

So I think adding task_is_traced to the test to verify the task
is still frozen.

static void ptrace_unfreeze_traced(struct task_struct *task)
{
	unsigned long flags;

	/*
	 * Verify the task is still frozen before unfreezing it,
	 * ptrace_resume could have unfrozen us.
	 */
	if (lock_task_sighand(task, &flags)) {
		if ((task->jobctl & JOBCTL_DELAY_WAKEKILL) &&
		    task_is_traced(task)) {
			task->jobctl &= ~JOBCTL_DELAY_WAKEKILL;
			if (__fatal_signal_pending(task))
				wake_up_state(task, __TASK_TRACED);
		}
		unlock_task_sighand(task, &flags);
	}
}

>> @@ -2307,6 +2307,7 @@ static int ptrace_stop(int exit_code, int why, int clear_code,
>>
>>  	/* LISTENING can be set only during STOP traps, clear it */
>>  	current->jobctl &= ~JOBCTL_LISTENING;
>> +	current->jobctl &= ~JOBCTL_DELAY_WAKEKILL;
>
> minor, but
>
> 	current->jobctl &= ~(JOBCTL_LISTENING | JOBCTL_DELAY_WAKEKILL);
>
> looks better.

Yes.


Eric