Re: [PATCH v1 1/3] i2c: designware-pci: Make bus number allocation robust

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Wed, 2016-06-15 at 16:42 +0300, Jarkko Nikula wrote:
> On 06/15/2016 01:55 AM, Andy Shevchenko wrote:
> > On some platforms, such as Intel Medfield, the I2C slave devices are
> > enumerated
> > through SFI tables where bus numbering is expected to be defined in
> > the OS.
> > Make the bus number allocation robust for such platforms.

> > ...
> > -	[medfield_3] = {
> > -		.bus_num     = 3,
> > -		.bus_cfg   = INTEL_MID_STD_CFG |
> > DW_IC_CON_SPEED_STD,

> > +	[medfield] = {
> > +		.bus_num = -1,
> >  		.bus_cfg   = INTEL_MID_STD_CFG |
> > DW_IC_CON_SPEED_FAST,

> Now bus 3 speed is pumped up to fast speed. Can it cause problems?

Oh, I have no idea, so, I would leave it at lower speed.

While we are here, what are your recommendations to put ->setup() call
in the ->probe()? Currently it is located somewhere in the middle of
assignments. I think the workflow has to be the following:
1. Take static controller data from platform based on ID's
2. Call controller->setup() if defined
3. Re-assign i2c device parameters with data in controller

-- 

Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Intel Finland Oy