Re: [PATCH v2 1/3] mfd: core: Make a best effort attempt to match devices with the correct of_nodes

[Michael Walle]

Hi,

Am 2020-06-24 08:41, schrieb Lee Jones:
> On Tue, 23 Jun 2020, Frank Rowand wrote:
>
> > On 2020-06-11 14:10, Lee Jones wrote:
> > > Currently, when a child platform device (sometimes referred to as a
> > > sub-device) is registered via the Multi-Functional Device (MFD) API,
> > > the framework attempts to match the newly registered platform device
> > > with its associated Device Tree (OF) node.  Until now, the device has
> > > been allocated the first node found with an identical OF compatible
> > > string.  Unfortunately, if there are, say for example '3' devices
> > > which are to be handled by the same driver and therefore have the same
> > > compatible string, each of them will be allocated a pointer to the
> > > *first* node.
> >
> > As you mentioned elsewhere in this thread, this series "fixes" the
> > problem related to the "stericsson,ab8500-pwm" compatible.
> >
> > I know, I said I would drop discussion of that compatible, but bear
> > with me for a second.  :-)
> >
> > The "problem" is that the devices for multiple mfd child nodes with
> > the same compatible value of "stericsson,ab8500-pwm" will all have
> > a pointer to the first child node.  At the moment the same child
> > of_node being used by more than one device does not cause any
> > incorrect behavior.
> >
> > Just in case the driver for "stericsson,ab8500-pwm" is modified
> > in a way that the child of_node needs to be distinct for each
> > device, and that changes gets back ported, it would be useful
> > to have Fixes: tags for this patch series.
> >
> > So, at your discretion (and I'll let you worry about the correct
> > Fixes: tag format), this series fixes:
> >
> > bad76991d7847b7877ae797cc79745d82ffd9120 mfd: Register ab8500 devices 
> > using the newly DT:ed MFD API
>
> This patch isn't actually broken.
>
> The issue is the DTB, which [0] addresses.
>
> [0]
> https://lkml.kernel.org/lkml/20200622083432.1491715-1-lee.jones@xxxxxxxxxx/

Now, I'm confused; because this patch doesn't use the reg property
but a different node name. I'd actually prefer this for any MFD
driver which has multiple nodes of the same compatible string. See
my reasoning here [1]. But until now, no one has responded. Thus,
I'd rather see a OF_MFD_CELL_NAME() which matches the node name
instead of the OF_MFD_CELL_REG() macro.

This would also circumvent the fact that the unit-address has one
number space. Eg. it is not possible to have:

mfd {
  compatible = "mfd,compatible";

  gpio@0 {
    reg = <0>;
  };
  gpio@1 {
    reg = <1>;
  };
  pwm@0 {
    reg = <0>;
  };
};

Although Rob mentioned to maybe relax that, but I sill fail to see
the advantage to have an arbitrary reg property instead of a unique
node name.

[1] 
https://lore.kernel.org/linux-devicetree/0709f20bc61afb6656bc57312eb69f56@xxxxxxxx/

--
-michael