Re: lockdep: incorrect deadlock warning with two GPIO expanders

[Peter Zijlstra]

On Mon, Sep 12, 2016 at 05:16:14PM +0200, Bartosz Golaszewski wrote:

> >>                                                - - - - -
> >>  -------             ---------  Bus segment 1 |         |
> >> |       |           |         |---------------  Devices
> >> |       | SCL/SDA   |         |               |         |
> >> | Linux |-----------| I2C MUX |                - - - - -
> >> |       |    |      |         | Bus segment 2
> >> |       |    |      |         |-------------------
> >>  -------     |       ---------                    |
> >>              |           |                    - - - - -
> >>         ------------     | MUX GPIO          |         |
> >>        |            |    |                     Devices
> >>        |    GPIO    |    |                   |         |
> >>        | Expander 1 |----                     - - - - -
> >>        |            |                             |
> >>         ------------                              | SCL/SDA
> >>                                                   |
> >>                                              ------------
> >>                                             |            |
> >>                                             |    GPIO    |
> >>                                             | Expander 2 |
> >>                                             |            |
> >>                                              ------------

> > The tricky part, and here I have absolutely no clue what so ever, is
> > being able to tell at pca953x_probe() time that this is so.
> >
> 
> 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?

You cannot inspect the device tree while probing?

> It uses the fact that the two expanders we have are of different type
> (pca9534 and pca9535). The id pointer points to per-chip device info
> residing in .data which makes it suitable for mutex key.
> 
> I don't think such hack is suitable for mainline though.

Right, that works by accident rather than anything else :-)