On Thu, 29 Nov 2012, Mark Brown wrote:

On Thu, Nov 29, 2012 at 10:08:08AM +0000, Lee Jones wrote:

quoted

On Wed, 28 Nov 2012, Mark Brown wrote:

quoted

quoted

The I2C subsystem guesses at a compatible string by default but it's
much better to explicitly set one as conflicts do arise from time to
time (eg, Wolfson parts are called WMxxxx but the WM prefix is also used
by at least WonderMedia).

quoted

It uses the exact device name, rather than guessing. I don't think
you're allowed to have duplicate device names in the system, or there
would be clashes at registration time.

quoted

The system is the same for platform data and DT alike.

Right, which is why this mostly works, but it's still better to provide
an actual compatible string which we can be 100% certain will avoid
conflicts.  This is very low cost when one is already defining DT
bindings.

I know that it's an easy thing to do, but I'm more worried about
what would happen if the an I2C device is registered using the
current way (stripping the compatible string and using it as a
client look-up) and we also provide a compatible entry in the
driver itself. Do you know if there is any danger of duplicate
registration or suchlike?

quoted

Hence, there should be no need to have a compatible string in any i2c
driver registration information.

We're getting away with it at present (and frankly nobody's going to
build in two different drivers for a chip of the same name for practical
systems anyway).

Right. In the same way as we're getting away with it when we
register a platform_device using the register/add routines.

-- 
Lee Jones
Linaro ST-Ericsson Landing Team Lead
Linaro.org ? Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog