[Intel-gfx] [PATCH][V2] drm/i915/gt: Fix a lockdep warning on RT kernel

[Tvrtko Ursulin]

On 15/04/2021 13:41, jun.miao wrote:
> 
> On 4/15/21 7:12 PM, Tvrtko Ursulin wrote:
>> [Please note: This e-mail is from an EXTERNAL e-mail address]
>>
>> Hi,
>>
>> On 14/04/2021 15:48, Jun Miao wrote:
>>> Don`t simple disable all the HD-irq, should race the region in the
>>> intel_breadcrumbs_disarm_irq() only.
>>>
>>
>> What is HD-irq, I am, not familiar with that term?
> 
> Disable local interrupt delivery from Hardware of cpu.:-)

HW then, not HD. ;)

[...]

>>>   static void add_signaling_context(struct intel_breadcrumbs *b,
>>> @@ -337,9 +339,7 @@ void __intel_breadcrumbs_park(struct 
>>> intel_breadcrumbs *b)
>>>       /* Kick the work once more to drain the signalers, and disarm 
>>> the irq */
>>>       irq_work_sync(&b->irq_work);
>>>       while (READ_ONCE(b->irq_armed) && !atomic_read(&b->active)) {
>>> -             local_irq_disable();
>>>               signal_irq_work(&b->irq_work);
>>> -             local_irq_enable();
>>
>> Unfortunately there is another lock inside signal_irq_work (rq->lock)
>> which needs to be taken irq safe.
>>
> Ok, i will change the left spin_lock -> spin_lock_irqsave.
> 
> In fact,  inside signal_irq_work,  intel_breadcrumbs_arm_irq 
> (&b->irq_lock)  which also needs to be taken irq safe.
> 
> Thanks,
> 
> Jun
> 
>> RT patches are in tree or out of the tree these days?
> 
> I base on the mainline kernel tree， and this BUG warning will not 
> happen.  But RT v5.10 will complain "BUG warning", so i want this patch 
> will solve RT WARNING without affecting mainline performance in mainline 
> tree.

So the problem is we did not typically do changes to cater for out of 
tree stuff, unless they are really minimal. And this one in my view does 
not quite qualify as such.

Regards,

Tvrtko