[PATCH v18 1/3] dt-bindings: i2c: aspeed,i2c.yaml: add transfer-mode and global-regs properties and update example

Rob Herring robh at kernel.org
Sat Sep 20 06:53:29 AEST 2025


On Mon, Sep 15, 2025 at 11:41:31AM +0800, Jeremy Kerr wrote:
> Hi Ryan,
> 
> > > OK, but the ast2400 and ast2500 I2C peripherals - which this binding also
> > > describes - do not have that facility. Given the 2600 is a distinct peripheral (as
> > > discussed on the v16 series), this would seem to warrant a distinct binding.
> > > 
> > > Should this be split out into an ast2600-specific binding, to reflect that it is
> > > different hardware? The reference to the global registers and transfer modes
> > > would then be added only to the ast2600-i2c-bus binding.
> > 
> > I agree it would be cleaner to split out a new binding file specifically for AST2600,
> > for example: `aspeed,ast2600-i2c.yaml`
> > But also I think `aspeed,i2cv2.yaml` more better name, that compatible will 
> > support next generation such like AST2700 .....
> 
> The ship may have already sailed on that one, as you already have the
> existing compatible string describing existing hardware.
> 
> I would assume that the compatible string should be fixed for an
> instance of the specific hardware, but the DT maintainers may be able to
> provide some input/precedence on changing an existing binding, if
> necessary.
> 
> If this does get changed, I would expect that you would need a
> corresponding update in the old driver too.
> 
> Or, another option may be to keep the current generation ("v2 core with
> compat registers") as-is (ie., at ast2600-i2c-bus), and introduce a new
> string for the next - where the primary hardware change might be the
> removal of compat registers, but it's still new hardware

About the only place we use version numbers (without regret) is when the 
version can be traced back to the actual verilog. This is typically only 
IP targeted to FPGAs. The reality in SoCs is h/w designers can't help 
themselves to not change things. In reality, there are almost always 
some changes.

Rob


More information about the openbmc mailing list