Re: [PATCH 2/2] iio: accel: bmc150: Check for a second ACPI device for BOSC0200

[Hans de Goede]

Hi,

On 04-12-17 10:58, Mika Westerberg wrote:
> On Sat, Dec 02, 2017 at 12:19:27PM +0000, Jonathan Cameron wrote:
> > On Wed, 29 Nov 2017 22:31:12 +0000
> > Jeremy Cline <jeremy@xxxxxxxxxx> wrote:
> >
> > > Some BOSC0200 acpi_device-s describe two accelerometers in a single ACPI
> > > device. Check for a companion device and handle a second i2c_client
> > > if it is present.
> >
> > + Mika and Wolfram - please cc them on anything odd and i2c / ACPI related.
> > (I like to share the pain)
> >
> > My usual question, just out of curiosity as we have to cope with this
> > fun anyway.  Are you actually allowed to do this under the ACPI spec
> > or not?  I would assume an acpi device is supposed to be just that A
> > device...  I fall asleep every time I try to read that spec ;)
>
> Yes, it is allowed. Typically you have an ACPI device and it can have
> multiple I2cSerialBus() connections.
>
> Linux ACPI/I2C core then picks the first one and creates i2c_client from
> that but the additional connections need to be created by the driver in
> question.
>
> BTW, there is a function i2c_new_secondary_device() that is supposed to
> be used for this but it does not have ACPI support yet (maybe it is good
> time to add it now, with this patch series?)

i2c_new_secondary_device() is for a different purpose, this is for when
a single i2c device listens on multiple addresses and the driver wants
separate i2c_client-s to use to talk to each address.

In this case there are 2 separate devices, not a single device listening
on multiple addresses.  Something like i2c_new_secondary_device() ACPI
support might be useful for i2c devices where a single device / "IC" listens
on multiple addresses and all these addresses are listed in the ACPI resource
table, but not for this specific case.

Regards,

Hans