[U-Boot] [PATCH] fdt: Do not report bad cell count when #size-cells = <0>

[Przemyslaw Marczak]

Hello Michal,

+CC Stephen Warren

On 04/14/2016 08:20 AM, Michal Simek wrote:
> Upcomming pca953x gpio driver (tca6416 is compatible with it)
> is using for example this DT description (ZynqMP zcu102):
> i2c0: i2c at ff020000 {
> 	compatible = "cdns,i2c-r1p10";
> 	interrupt-parent = <&gic>;
> 	interrupts = <0 17 4>;
> 	reg = <0x0 0xff020000 0x1000>;
> 	#address-cells = <1>;
> 	#size-cells = <0>;
>
> 	tca6416_u61: gpio at 21 {
> 		compatible = "ti,tca6416";
> 		reg = <0x21>;
> 		gpio-controller;
> 		#gpio-cells = <2>;
> 	};
> };
>
> I2C bus is using #size-cells = <0> which current code detects as
> "Bad cell count":
> OF: reached root node
> OF: ** translation for device gpio at 20 **
> __of_translate_address: Bad cell count for gpio at 20
> Command 'gpio' failed: Error -19
>
> This patch fixes OF_CHECK_COUNTS macro not to report problem when
> size-cells is 0 which is also valid bus configuration.
>
> Signed-off-by: Michal Simek <michal.simek at xilinx.com>
> ---
>
>   common/fdt_support.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/common/fdt_support.c b/common/fdt_support.c
> index ced119e70d9f..5f5b49c6210b 100644
> --- a/common/fdt_support.c
> +++ b/common/fdt_support.c
> @@ -941,7 +941,7 @@ void fdt_del_node_and_alias(void *blob, const char *alias)
>   #define OF_MAX_ADDR_CELLS	4
>   #define OF_BAD_ADDR	FDT_ADDR_T_NONE
>   #define OF_CHECK_COUNTS(na, ns)	((na) > 0 && (na) <= OF_MAX_ADDR_CELLS && \
> -			(ns) > 0)
> +			(ns) >= 0)
>
>   /* Debug utility */
>   #ifdef DEBUG
>

Some time ago I tried to solve GPIO issue on Exynos boards, by proposing 
change similar to your patch. After some discussion with Stephen Warren 
and Simon Glass, I was aware, that it was a wrong approach to solve this 
problem.

You can follow it here:

http://comments.gmane.org/gmane.comp.boot-loaders.u-boot/247651

https://patchwork.ozlabs.org/patch/564246/
https://patchwork.ozlabs.org/patch/557008/
https://patchwork.ozlabs.org/patch/557010/
https://patchwork.ozlabs.org/patch/557009/

As a conclusion, for MMIO addresses we are using 'reg' property with a 
range. If you need use this as I2C address or I2C device's register - 
e.g. 'reg = <0x4>' - it's good until you don't use dev_get_addr() 
function to get it - because it's a different address space - non-MMIO.

I didn't saw your GPIO driver, but I can assume, that you should use 
fdtdec_get_int() to get the GPIO register addres instead of dev_get_addr().

Please look at function i2c_chip_ofdata_to_platdata() in i2c-uclass.c, 
the reg is taken by using fdtdec_get_int()...

Best regards,
-- 
Przemyslaw Marczak
Samsung R&D Institute Poland
Samsung Electronics
p.marczak at samsung.com