Le 02/05/2013 16:06, jonsm...@gmail.com a écrit :
> On Thu, May 2, 2013 at 9:49 AM, Michel Stempin
> <michel.stem...@wanadoo.fr> wrote:
>> Hi,
>>
>> It looks like the GPIOs are broken since the introduction of DTS for the 
>> ramips architecture.
>>
>> In the DTS include file ("target/linux/ramips/dts/rt3050.dtsi" in my case, 
>> by this is also true for the other ones), only the number of GPIOs for each 
>> gpio-controller is defined, but not the GPIO base number for each chip. For 
>> example:
>>
>>                 gpio0: gpio@600 {
>>                         compatible = "ralink,rt3052-gpio", 
>> "ralink,rt2880-gpio";
>>                         reg = <0x600 0x34>;
>>
>>                         gpio-controller;
>>                         #gpio-cells = <2>;
>>
>>                         ralink,num-gpios = <24>;
>>                         ralink,register-map = [ 00 04 08 0c
>>                                                 20 24 28 2c
>>                                                 30 34 ];
>>
>> This part of DTS is handled by the ralink_gpio_probe() function added by the 
>> "0130-GPIO-MIPS-ralink-adds-ralink-gpio-support.patch" file for the ramips 
>> architecture, which sets the GPIO base number to -1 to get a 
>> dynamically-allocated base GPIO number.
>>
>> This feature was introduced in Linux kernel 2.6.26-rc1 
>> (<http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/gpio/gpiolib.c?id=8d0aab2f16c4fa170f32e7a74a52cd0122bbafef>),
>>  but unfortunately, it is not applicable here, since to avoid using any 
>> numbers that may have been explicitly assigned but not yet registered, this 
>> dynamic allocation assigns GPIO numbers from the biggest number on down, 
>> instead of from the smallest on up.
>>
>> This assignment is performed by function gpiochip_find_base() in the 
>> kernel's "drivers/gpio/gpiolib.c" file, and as the same patch above defines 
>> ARCH_NR_GPIOS to 128, this result in a base GPIO number of 128 - 24 = 104 
>> instead of the expected 0.
>>
>> Since changing the dynamic GPIO number feature in the kernel is out of 
>> question as it would break the userspace ABIs 
>> (<https://patchwork.kernel.org/patch/2084411/>), what is the best option to 
>> get the GPIOs back working?
>>
>> My first guess would be to add a "ralink,base-gpio" property for each 
>> gpio-controller chip to the DTS files for ralink...
>>
>> Any suggestions?
> 
> Your code should not care about the assigned GPIO number.
> 
> gpio_ebi_i2stx_0: gpio@13003040 {
>   #gpio-cells = <2>;
>   compatible = "nxp,lpc31xx-gpio";
>   reg = <0x13003040 0x40>;
>   gpio-controller;
> };
> gpio_spi: gpio@13003240 {
>   #gpio-cells = <2>;
>   compatible = "nxp,lpc31xx-gpio";
>   reg = <0x13003240 0x40>;
>   gpio-controller;
> };
> 
> 
> Spi device that uses the GPIO...
> 
> spi@15002000 {
>   gpios = <&gpio_spi 4 0  /* chip selects */
>     &gpio_ebi_i2stx_0 3 0>;
>   s25sl032a@0 {
>     compatible = "code,s25sl032a";
>     reg = <0>;
>     spi-max-frequency = <1000000>;
>   };
> };
> 
> then use something like:
> 
> gpio = of_get_gpio_flags(pdev->dev.of_node, i, &flags);
> or
> cs_gpio = of_get_named_gpio(pdev->dev.of_node, "cs-gpios", i);
> 
> To attach your gpio.

What about the /sys/class/gpio user interface, then?

I get a gpiochip124 with a gpiochip124/base = 124 in my case...

-Michel
> 
> 
> 
>>
>> -Michel
>> _______________________________________________
>> openwrt-devel mailing list
>> openwrt-devel@lists.openwrt.org
>> https://lists.openwrt.org/mailman/listinfo/openwrt-devel
> 
> 
> 
> --
> Jon Smirl
> jonsm...@gmail.com
> _______________________________________________
> openwrt-devel mailing list
> openwrt-devel@lists.openwrt.org
> https://lists.openwrt.org/mailman/listinfo/openwrt-devel
> 
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel

Reply via email to