On Tue, 26 Jun 2007 22:02:53 +0200, "Franck Bui-Huu" <vagabon.xyz@xxxxxxxxx> wrote: > It seems that clock configuration highly depends on the board, not the > arch. For example, a board can have only static clocks which won't be > unregistered later. In that case most of code provided by the patch is > useless. > > So maybe the best thing is to let each board implements their own > generic clock API (exactly like the generic GPIO API you did > recently). Or make a default implementation (your patch) and allow > others boards to make their own implementation. Hmm... I had thought the default implementation would be useful, but now I change my mind ;) What people expected on its implementation would vary too much. Now I think leaving all implementation to platform code is better. I'll send rbtx4938 implementation (minimum one just for SPI driver) soon. Ralf, please do not apply or queue this patch :) --- Atsushi Nemoto
