On 04/07/2013 05:14 PM, Jan Kiszka wrote:
> On 2013-04-07 17:10, Gilles Chanteperdrix wrote:
>> On 04/06/2013 08:32 PM, Jan Kiszka wrote:
>>
>>> On 2013-04-06 20:27, Gilles Chanteperdrix wrote:
>>>>> And it has to map directly to might_fault
>>>>> in case CONFIG_IPIPE_DEBUG_CONTEXT is disabled (thus zero overhead).
>>>>
>>>> That would defeat the purpose of the patch. might_fault can not be
>>>> called over non root domains.
>>>
>>> True. But we should avoid leaving the costly irq-disable/enable and
>>> domain checks in production code behind when might_fault actually
>>> resolves to a nop. That should be the case for
>>> !CONFIG_DEBUG_ATOMIC_SLEEP && !CONFIG_PROVE_LOCKING.
>>
>>
>> I am almost sure I did not have either of these two options enabled when
>> I observed the bug.
>
> Hmm. And CONFIG_PREEMPT_VOLUNTARY? In that case, might_sleep seems to
> serve as a voluntary preemption point.
>From what I see, ipipe_root_only is called systematically, even when
CONFIG_PREEMPT_VOLUNTARY is disabled.
--
Gilles.
_______________________________________________
Xenomai mailing list
[email protected]
http://www.xenomai.org/mailman/listinfo/xenomai
