Re: yama: lockdep warning on yama_ptracer_del

[Kees Cook]

On Mon, Nov 19, 2012 at 10:59 AM, Sasha Levin <sasha.levin@xxxxxxxxxx> wrote:
> On 11/19/2012 11:23 AM, Kees Cook wrote:
>> On Sun, Nov 18, 2012 at 8:05 PM, Sasha Levin <levinsasha928@xxxxxxxxx> wrote:
>>> Hi Kees,
>>>
>>> On Thu, Oct 18, 2012 at 6:39 PM, Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>>>> On Wed, Oct 17, 2012 at 8:10 PM, Sasha Levin <sasha.levin@xxxxxxxxxx> wrote:
>>>>> Hi all,
>>>>>
>>>>> I was fuzzing with trinity within a KVM tools guest (lkvm) on a linux-next kernel, and got the
>>>>> following dump which I believe to be noise due to how the timers work - but I'm not 100% sure.
>>>>> ...
>>>>> [  954.674123]  Possible interrupt unsafe locking scenario:
>>>>> [  954.674123]
>>>>> [  954.674123]        CPU0                    CPU1
>>>>> [  954.674123]        ----                    ----
>>>>> [  954.674123]   lock(ptracer_relations_lock);
>>>>> [  954.674123]                                local_irq_disable();
>>>>> [  954.674123]                                lock(&(&new_timer->it_lock)->rlock);
>>>>> [  954.674123]                                lock(ptracer_relations_lock);
>>>>> [  954.674123]   <Interrupt>
>>>>> [  954.674123]     lock(&(&new_timer->it_lock)->rlock);
>>>>> [  954.674123]
>>>>> [  954.674123]  *** DEADLOCK ***
>>>>
>>>> I've been wanting to get rid of the Yama ptracer_relations_lock
>>>> anyway, so maybe I should do that now just to avoid this case at all?
>>>
>>> I still see this one in -rc6, is there anything to get rid of it
>>> before the release?
>>
>> I'm not sure about changes to the timer locks, but I haven't been able
>> to get rid of the locking on Yama's task_free path. I did send a patch
>> to get rid of locking during a read, though:
>>
>> https://lkml.org/lkml/2012/11/13/808
>
> Aw, alrighty. It didn't make it to -next yet though.
>
> I'll add the patch to my tree and test with it.

Unfortunately, I don't think it'll help since your example showed the
delete path on both sides, which is still locked. I've been trying to
think of ways to avoid the lock here, but haven't hit on anything
satisfying.

-Kees

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