On Wed, May 20, 2015 at 7:20 PM, Dmitry Torokhov
[off-list ref] wrote:

On Wed, May 20, 2015 at 07:30:50PM +0100, Russell King - ARM Linux wrote:

quoted

On Wed, May 20, 2015 at 07:25:41PM +0100, Russell King - ARM Linux wrote:

quoted

On Wed, May 20, 2015 at 09:58:51AM +0800, Shawn Guo wrote:

quoted

On Wed, May 20, 2015 at 01:04:35AM +0800, Frank.Li at freescale.com wrote:

quoted

Frank Li (6):
  arm: imx: power-off: change to syscon to access register
  rtc: arm: imx: snvs: change use syscon to access register
  Document: dt: fsl: snvs: change support syscon
  arm: dts: imx: update snvs to use syscon access register
  document: devicetree: input: imx: i.mx snvs power device tree bindings
  arm: dts: imx6sx: enable snvs power key

So, the series will break existing DTBs for RTC and poweroff, right?
If this is unavoidable, should we at least have some warning messages
telling users about his breakage?

What's the justification for breaking existing DTBs?  Really, this is
something we should strive to avoid, _and_ actually avoid by providing
backwards compatibility.

Looking at the RTC code, it would be pretty trivial to do.

Rather than passing around struct regmap, pass around the private data
structure.

Rename the regmap_* function calls to be private accessors.  Implement
a set of new accessors - if the private data has the regmap, use the
regmap API.  Otherwise, use the old method, and print a warning that an
old DT is being used.

Or maybe have drivers create regmap themselves if they see old DTS and
not bother with the 2 sets of accessors?

Does any dt node convert to regmap regardless if there are "syscon"
compatible string?

Thanks.

--
Dmitry