Hi,

El 03/08/14 a las 12:55, Chen-Yu Tsai escibi?:

On Sun, Aug 3, 2014 at 8:50 PM, Maxime Ripard
[off-list ref] wrote:

quoted

On Thu, Jul 31, 2014 at 05:46:09PM -0400, jonsmirl at gmail.com wrote:

quoted

Would it be better to name this "allwinner,sun4i-a10-pll2-clk" instead
of "allwinner,sun4i-a10-b-pll2-clk"? By encoding the b in it everyone
is going to wonder what to do on the 'c' revision which is the most
common revision.

Not really, the way we works usually is that the compatible is the one
from the first SoC that implemented that IP. If the rev C has the same
IP than rev B, then we're using the rev B compatible.

quoted

The revision based rename would then be from
"allwinner,sun4i-a10-pll2-clk" to "allwinner,sun4i-a10-a-pll2-clk".

Though, I'd agree with you. We should have a single compatible in the
DT, a generic one, that would trigger the auto-detection, and might
change it to the rev A one, but the rev B doesn't make much sense.

I agree. The rev B would make people reading the DT wonder what happened
to rev A.

Would you like to see "allwinner,sun4i-a10-pll2-clk" on the DT then? 
Should we document it as a magic property triggering an autodetect on 
the clock binding document?

Cheers!

Emilio