[U-Boot] [Uboot-stm32] [PATCH v1 4/5] pinctrl: stm32: Add pinmux_show() ops

[Patrice CHOTARD]

Hi Simon

On 09/26/2018 07:41 AM, Simon Glass wrote:
> Hi Patrice,
> 
> On 20 September 2018 at 07:37, Patrice Chotard <patrice.chotard at st.com> wrote:
>> pinmux_show allows to display the muxing of all pins
>> belonging to pin-controller.
>>
>> Signed-off-by: Patrice Chotard <patrice.chotard at st.com>
>> ---
>>
>>  drivers/pinctrl/pinctrl_stm32.c | 79 +++++++++++++++++++++++++++++++++++++++++
>>  1 file changed, 79 insertions(+)
>>
> 
> The only thing I don't like about this is that it is quite
> SoC-specific. I am hoping we can make it more generic.

I will rework it to make it more independant.

> 
>> diff --git a/drivers/pinctrl/pinctrl_stm32.c b/drivers/pinctrl/pinctrl_stm32.c
>> index 31285cdd5784..a477035bf420 100644
>> --- a/drivers/pinctrl/pinctrl_stm32.c
>> +++ b/drivers/pinctrl/pinctrl_stm32.c
>> @@ -14,6 +14,16 @@ DECLARE_GLOBAL_DATA_PTR;
>>  #define OTYPE_MSK                      1
>>  #define AFR_MASK                       0xF
>>
>> +#define PINMUX_MODE_COUNT              5
>> +
>> +static const char * const pinmux_mode[PINMUX_MODE_COUNT] = {
>> +       "gpio input",
>> +       "gpio output",
>> +       "analog",
>> +       "unknown",
>> +       "alt function",
>> +};
>> +
>>  static int stm32_gpio_config(struct gpio_desc *desc,
>>                              const struct stm32_gpio_ctl *ctl)
>>  {
>> @@ -176,12 +186,81 @@ static int stm32_pinctrl_set_state_simple(struct udevice *dev,
>>  }
>>  #endif /* PINCTRL_FULL */
>>
>> +static int stm32_pinctrl_get_af(struct udevice *dev,  unsigned int offset)
>> +{
>> +       struct stm32_gpio_priv *priv = dev_get_priv(dev);
>> +       struct stm32_gpio_regs *regs = priv->regs;
>> +       u32 af;
>> +       u32 alt_shift = (offset % 8) * 4;
>> +       u32 alt_index =  offset / 8;
>> +
>> +       af = (readl(&regs->afr[alt_index]) &
>> +             GENMASK(alt_shift + 3, alt_shift)) >> alt_shift;
>> +
>> +       return af;
>> +}
>> +
>> +static int stm32_pinmux_show(struct udevice *dev)
>> +{
>> +       struct udevice *child;
>> +       struct udevice *dev_gpio;
>> +       const char *bank_name;
>> +       const char *label;
>> +       int offset;
>> +       int ret;
>> +       int num_bits;
>> +       int mode;
>> +       int af_num;
>> +
>> +       /* parse pin-controller sub-nodes, ie gpio bank nodes */
>> +       list_for_each_entry(child, &dev->child_head, sibling_node) {
>> +               ret = uclass_get_device_by_name(UCLASS_GPIO, child->name,
>> +                                               &dev_gpio);
> 
> I wonder to what extend this is actually different from 'gpio status'?

I agree, currently, "pinmux show" and "gpio status" comments are
overlapping..
At the end, as on kernel side, only pinmux should be used to know the
pin state (configured as a GPIO or as alternate function).

pinmux command indicates if pins is configured as an alternate function
and indicates which one, what "gpio status" can't do.

> 
>> +
>> +               if (ret < 0 && ret != -ENODEV) {
>> +                       dev_err(dev, "Failed to find %s device ret = %d\n",
>> +                               child->name, ret);
>> +                       return ret;
>> +               }
>> +
>> +               if (!ret) {
>> +                       bank_name = gpio_get_bank_info(dev_gpio, &num_bits);
>> +
>> +                       printf("\nBank %s:\n", bank_name);
>> +                       for (offset = 0; offset < num_bits; offset++) {
>> +                               mode = gpio_get_raw_function(dev_gpio,
>> +                                                            offset, &label);
>> +                               printf("%s%d: %s", bank_name, offset,
>> +                                      pinmux_mode[mode]);
>> +                               switch (mode) {
>> +                               case GPIOF_FUNC:
>> +                                       af_num = stm32_pinctrl_get_af(dev_gpio,
>> +                                                                     offset);
>> +                                       printf(" %d", af_num);
>> +                                       break;
>> +                               case STM32_GPIO_MODE_OUT:
> 
> How come you cannot use GPIOF_OUTPUT here?

Good catch, GPIOF_OUPUT/INPUT must be used instead of
STM32_GPIO_MODE_OUT/IN.

> 
>> +                               case STM32_GPIO_MODE_IN:
>> +                                       printf(" %s", label ? label : "");
>> +                                       break;
>> +                               }
>> +                               printf("\n");
>> +                       }
>> +               }
>> +
>> +               if (!child)
>> +                       break;
>> +       }
>> +
>> +       return 0;
>> +}
>> +
>>  static struct pinctrl_ops stm32_pinctrl_ops = {
>>  #if CONFIG_IS_ENABLED(PINCTRL_FULL)
>>         .set_state              = stm32_pinctrl_set_state,
>>  #else /* PINCTRL_FULL */
>>         .set_state_simple       = stm32_pinctrl_set_state_simple,
>>  #endif /* PINCTRL_FULL */
>> +       .pinmux_show            = stm32_pinmux_show,
>>  };
>>
>>  static const struct udevice_id stm32_pinctrl_ids[] = {
>> --
>> 1.9.1
>>
> 
> Regards,
> Simon
> _______________________________________________
> Uboot-stm32 mailing list
> Uboot-stm32 at st-md-mailman.stormreply.com
> https://st-md-mailman.stormreply.com/mailman/listinfo/uboot-stm32
>