[PATCH v3] irqchip/tango: Don't use incorrect irq_mask_ack callback

[Marc Gonzalez]

On 27/07/2017 20:17, Florian Fainelli wrote:

> On 07/26/2017 12:13 PM, Måns Rullgård wrote:
>
>> Florian Fainelli writes:
>>
>>> On 07/25/2017 06:29 AM, Måns Rullgård wrote:
>>>
>>>> Marc Gonzalez <marc_gonzalez at sigmadesigns.com> writes:
>>>>
>>>>> On 25/07/2017 15:16, Måns Rullgård wrote:
>>>>>
>>>>>> What happened to the patch adding the proper combined function?
>>>>>
>>>>> It appears you're not CCed on v2.
>>>>>
>>>>> https://patchwork.kernel.org/patch/9859799/
>>>>>
>>>>> Doug wrote:
>>>>>> Yes, you understand correctly.  The irq_mask_ack method is entirely
>>>>>> optional and I assume that is why this issue went undetected for so
>>>>>> long; however, it is slightly more efficient to combine the functions
>>>>>> (even if the ack is unnecessary) which is why I chose to do so for my
>>>>>> changes to the irqchip-brcmstb-l2 driver where I first discovered this
>>>>>> issue.  How much value the improved efficiency has is certainly
>>>>>> debatable, but interrupt handling is one area where people might care
>>>>>> about such a small difference.  As the irqchip-tango driver maintainer
>>>>>> you are welcome to decide whether or not the irq_mask_ack method makes
>>>>>> sense to you.
>>>>>
>>>>> My preference goes to leaving the irq_mask_ack callback undefined,
>>>>> and let the irqchip framework use irq_mask and irq_ack instead.
>>>>
>>>> Why would you prefer the less efficient way?
>>>>
>>>
>>> Same question here, that does not really make sense to me.
>>>
>>> The whole point of this patch series is to have a set of efficient and
>>> bugfree (or nearly) helper functions that drivers can rely on, are you
>>> saying that somehow using irq_mask_and_ack is exposing a bug in the
>>> tango irqchip driver and using the separate functions does not expose
>>> this bug?
>>
>> There is currently a bug in that the function used doesn't do what its
>> name implies which can't be good.  Using the separate mask and ack
>> functions obviously works, but combining them saves a lock/unlock
>> sequence.  The correct combined function has already been written, so I
>> see no reason not to use it.
> 
> Marc/Mason, are you intending to get this patch accepted in order to
> provide a quick bugfix targeting earlier kernels with the tango irqchip
> driver or is this how you think the correct fix for the tango irqchip
> driver is as opposed to using Doug's fix?

Hello Florian,

I am extremely grateful for you and Doug bringing the defect to
my attention, as it was indeed causing an issue which I had not
found the time to investigate.

The reason I proposed an alternate patch is that
1) Doug didn't seem to mind, 2) simpler code leads to fewer bugs
and less maintenance IME, and 3) I didn't see many drivers using
the irq_mask_ack() callback (9 out of 86) with a few misusing it,
by defining irq_mask = irq_mask_ack.

As you point out, my patch might be slightly easier to backport
than Doug's (TBH, I hadn't considered that aspect until you
mentioned it).

Has anyone ever quantified the performance improvement of
mask_ack over mask + ack?

Regards.