Re: [PATCH v3 1/1] i2c: add ACPI support for I2C mux ports

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Wed, Oct 21, 2015 at 01:52:41AM -0700, Dustin Byford wrote:
> On Wed Oct 21 11:34, Mika Westerberg wrote:
> > On Wed, Oct 21, 2015 at 01:21:16AM -0700, Dustin Byford wrote:
> > > On Wed Oct 21 11:12, Mika Westerberg wrote:
> > > > On Tue, Oct 20, 2015 at 10:49:59AM -0700, Dustin Byford wrote:
> > > > > I considered it, but I thought a default that fairly closely matches the
> > > > > old behavior was more convenient.
> > > > > 
> > > > > On the other hand, leaving it up to the controllers makes it all very
> > > > > explicit and perhaps simpler to reason about.
> > > > > 
> > > > > 
> > > > > I could be convinced either way.  But, if we move it to the controller
> > > > > drivers, which ones need the change?
> > > > > 
> > > > > grep -i acpi drivers/i2c/busses/i2c*
> > > > > 
> > > > > shows 18 drivers that might care.
> > > > 
> > > > I'm quite confident the designware I2C is enough for now. Intel uses it
> > > > for all SoCs with LPSS and I think AMD has the same block for their I2C
> > > > solution.
> > > 
> > > I certainly care about i801, ismt, and isch.  Doesn't this affect any
> > > i2c controller with clients that may be enumerated through ACPI?
> > 
> > Yes, but so far I haven't seen any other devices being used for this
> > than the I2C designware.
> > 
> > Which hardware you are testing this patch on?
> 
> I'm working with a number of x86-based network switch platforms.  Mostly
> rangeley at the moment, but I'm sure others are in the works.  Have a
> look at:
> 
> http://www.opencompute.org/wiki/Networking/SpecsAndDesigns
> 
> for examples.

Cool :)

> My goal, hence the recent patches, is to help the network switch
> industry move a lot of platform description into ACPI.  That means lots
> of complicated I2C trees; switches are full of I2C devices.

I see.

I don't really have strong feelings whether it should be the I2C core or
individual drivers setting the ACPI companion. However, it would be nice
to match DT here and they assign their of_node per driver.