linux-kernel - Re: [RFC PATCH 1/2] Fix: sched/membarrier: p->mm->membarrier

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 for Android: free password hash cracker in your pocket

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

Message-ID: <88198910.1581.1567610670849.JavaMail.zimbra@efficios.com>
Date:   Wed, 4 Sep 2019 11:24:30 -0400 (EDT)
From:   Mathieu Desnoyers <mathieu.desnoyers@...icios.com>
To:     Oleg Nesterov <oleg@...hat.com>
Cc:     paulmck <paulmck@...ux.ibm.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        linux-kernel <linux-kernel@...r.kernel.org>,
        "Eric W. Biederman" <ebiederm@...ssion.com>,
        "Russell King, ARM Linux" <linux@...linux.org.uk>,
        Chris Metcalf <cmetcalf@...hip.com>,
        Chris Lameter <cl@...ux.com>, Kirill Tkhai <tkhai@...dex.ru>,
        Mike Galbraith <efault@....de>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...nel.org>
Subject: Re: [RFC PATCH 1/2] Fix: sched/membarrier: p->mm->membarrier_state
 racy load

----- On Sep 4, 2019, at 6:53 AM, Oleg Nesterov oleg@...hat.com wrote:

> On 09/03, Mathieu Desnoyers wrote:
>>
>> @@ -1130,6 +1130,10 @@ struct task_struct {
>>  	unsigned long			numa_pages_migrated;
>>  #endif /* CONFIG_NUMA_BALANCING */
>>
>> +#ifdef CONFIG_MEMBARRIER
>> +	atomic_t membarrier_state;
>> +#endif
> 
> ...
> 
>> +static inline void membarrier_prepare_task_switch(struct task_struct *t)
>> +{
>> +	if (!t->mm)
>> +		return;
>> +	atomic_set(&t->membarrier_state,
>> +		   atomic_read(&t->mm->membarrier_state));
>> +}
> 
> Why not
> 
>	rq->membarrier_state = next->mm ? t->mm->membarrier_state : 0;
> 
> and
> 
>	if (cpu_rq(cpu)->membarrier_state & MEMBARRIER_STATE_GLOBAL_EXPEDITED) {
>		...
>	}
> 
> in membarrier_global_expedited() ? (I removed atomic_ to simplify)
> 
> IOW, why this new member has to live in task_struct, not in rq?

As replied to Linus, if we copy the membarrier_state into the rq, we'd need
to ensure we have full memory barriers between:

prior user-space memory accesses  /  setting the runqueue membarrier state

and

setting the runqueue membarrier state / following user-space memory accesses

Because membarrier does not take any runqueue lock when it iterates over
runqueues.

I try to avoid putting too much memory barrier constraints on the scheduler
for membarrier, but if it's really the way forward it could be done.

And the basic question remains: it is acceptable performance-wise to load
mm->membarrier_state from sched switch prepare ?

Thanks,

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com