Re: lockdep: incorrect deadlock warning with two GPIO expanders

[Bartosz Golaszewski]

2016-09-15 14:41 GMT+02:00 Linus Walleij <linus.walleij@xxxxxxxxxx>:
> On Thu, Sep 15, 2016 at 9:51 AM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>> On Tue, Sep 13, 2016 at 02:29:24PM +0200, Linus Walleij wrote:
>>> On Mon, Sep 12, 2016 at 5:33 PM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>>> > On Mon, Sep 12, 2016 at 05:16:14PM +0200, Bartosz Golaszewski wrote:
>>>
>>> >> AFAIK there is no clean way to tell that a GPIO is used by an I2C
>>> >> multiplexer at probe time. Linus, Alexandre could you confirm?
>>>
>>> Nominally, the GPIO descriptors are just abstract resources such
>>> as regulators or clocks, they can be used for a lot but just like
>>> a clock, regulator, dma channel etc does not know who is using
>>> it and for what, it does not know this, no.
>>>
>>> > You cannot inspect the device tree while probing?
>>>
>>> Of course it *can* but we would end up encoding a special
>>> case every time something like this happens, tied to just
>>> device tree, then another bolt-on for ACPI etc.
>>>
>>> I have a hard time following the problem really, I'm
>>> afraid I'm simply just not smart enough :(
>>
>> Why would this be DT or ACPI specific? Linux itself has a tree/graph of
>> all busses and devices right? That's what all this drivers/base/ stuff
>> is on about.
>>
>> So can't you walk up that and see if you encounter the exact same driver
>> again?
>>
>> Something like:
>>
>>         for (nr = 0, parent = dev->parent; parent; parent = parent->parent) {
>>                 if (parent->device_driver == &pca953x_driver.driver)
>>                         nr++;
>>         }
>
> Oh clever. Of course.
>
> Bartosz can you try out this approach?
>

I think it may be more complicated than that, depending on the hw
topology, but the general idea seems reasonable. I'll try this.

Thanks,
Bartosz Golaszewski