linux-kernel - Re: [PATCH 2/3] futex: avoid double wake up in futex

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.DEB.2.11.1504072131400.3845@nanos>
Date:	Tue, 7 Apr 2015 21:47:13 +0200 (CEST)
From:	Thomas Gleixner <tglx@...utronix.de>
To:	Sebastian Andrzej Siewior <bigeasy@...utronix.de>
cc:	linux-kernel@...r.kernel.org,
	Peter Zijlstra <peterz@...radead.org>,
	Ingo Molnar <mingo@...hat.com>,
	Darren Hart <darren@...art.com>,
	Steven Rostedt <rostedt@...dmis.org>,
	fredrik.markstrom@...driver.com,
	Davidlohr Bueso <davidlohr@...com>,
	Manfred Spraul <manfred@...orfullife.com>
Subject: Re: [PATCH 2/3] futex: avoid double wake up in futex_wake() on -RT

On Tue, 7 Apr 2015, Sebastian Andrzej Siewior wrote:
> futex_wake() wakes the waiter while holding the hb->lock. This leads to
> a similar double wake up on -RT if the waiter has a higher priority than
> the process perfroming the wake up.

Well, the non pi wakeup is designed in a way that the waiter side does
not take the hash bucket lock.

So if you observe that hb->lock contention and the resulting PI
boosting dance on RT, then it's not the wakeup/waiter exit path. That
must be something like this:

T1   		  T2
wakeup()

  -->preemption

		  sys_exit(futex_wait)
		  ...
		  sys_enter(futex_wake);
		    lock(hb->lock);
 
> -/*
> - * The hash bucket lock must be held when this is called.
> - * Afterwards, the futex_q must not be accessed.
> - */

The comment should stay here, because you are not allowed to access
the futex_q after q->lock_ptr has been set to NULL and that happens in
this function.

> -static void wake_futex(struct futex_q *q)
> +static struct task_struct *__wake_futex(struct futex_q *q)
>  {
>  	struct task_struct *p = q->task;
>  
>  	if (WARN(q->pi_state || q->rt_waiter, "refusing to wake PI futex\n"))
> -		return;
> +		return NULL;
>  
>  	/*
>  	 * We set q->lock_ptr = NULL _before_ we wake up the task. If
> @@ -1117,6 +1113,19 @@ static void wake_futex(struct futex_q *q)
>  	 */
>  	smp_wmb();
>  	q->lock_ptr = NULL;
> +	return p;

...

> @@ -1256,14 +1266,23 @@ futex_wake(u32 __user *uaddr, unsigned int flags, int nr_wake, u32 bitset)
>  			if (!(this->bitset & bitset))
>  				continue;
>  
> -			wake_futex(this);
> +			if (nr_wake == 1)
> +				waiter = __wake_futex(this);
> +			else
> +				wake_futex(this);
>  			if (++ret >= nr_wake)
>  				break;
>  		}
>  	}
>  
>  	spin_unlock(&hb->lock);
> +
>  out_put_key:
> +	if (waiter) {
> +		wake_up_state(waiter, TASK_NORMAL);
> +		put_task_struct(waiter);
> +	}

This should go before out_put_key, because none of the other code
pathes which jump there can set waiter.

Thanks,

	tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/