On Tue, Oct 15, 2013 at 8:18 PM, Arnd Bergmann <arnd@xxxxxxxx> wrote:
> On Tuesday 15 October 2013, Haojian Zhuang wrote:
> Since all clocks are in the fc802000 page, I would prefer using a non-empty
> ranges to map the clocks into their own address space, like
>
> ranges = <0 0xfc802000 0x1000>;
>
>
>> + timer4_mux: timer4_mux@fc802018 {
>> + compatible = "hisilicon,clk-mux";
>> + #clock-cells = <0>;
>> + clocks = <&osc32k &timerclk45>;
>> + clock-output-names = "timer4_mux";
>> + reg = <0xfc802018 0x4>;
>> + clkmux-mask = <0x3>;
>> + clkmux-table = <0 0x1>;
>> + };
>
> which would turn this into
>
> timer4_mux: timer4_mux@18 {
> ...
> reg = <0x18 0x4>;
> ...
> };
>
> The code would not be impacted by this.
I'd rather question the addition of a reg =<>; specifier
for each and every clock.
Knowing the offsets of each and every register in a
memory-mapped device is usually overdesign, and
goes from specifying register base to specifying
register offsets and to me that is overspecifying.
I'd rather give each clock a unique ID or something,
then let the driver know the offsets, and specify the
base and size in the top node.
This becomes more obvious with a second silicon
revision moving the registers around - with this
design all of the clocks need updating, but if you're
lucky with less knowledge in the DTS you only need
a new compatible string and the knowledge of the new
whereabout for the regs is kept in the driver.
Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
