Thread (138 messages) 138 messages, 6 authors, 2013-10-25

Re: [PATCH 1/2] pinmux: Add TB10x pinmux driver

From: Haojian Zhuang <haojian.zhuang@linaro.org>
Date: 2013-06-08 08:31:08
Also in: lkml

On 8 June 2013 03:18, Stephen Warren [off-list ref] wrote:
On 06/06/2013 09:30 AM, Christian Ruppert wrote:
quoted
On Thu, Jun 06, 2013 at 10:32:21PM +0800, Haojian Zhuang wrote:
quoted
On 6 June 2013 22:11, Christian Ruppert [off-list ref] wrote:
quoted
On Wed, Jun 05, 2013 at 09:44:27AM +0800, Haojian Zhuang wrote:
quoted
On 3 June 2013 20:30, Christian Ruppert [off-list ref] wrote:
quoted
OK, here's a simplified example of what we would like to do (this seems
pretty common so I suppose there is a way I haven't understood). Our
situation is slightly more complex but for the purpose of discussion
let's assume a chip with 8 pins which can be configured for the
following functions:

Pin   GPIO-A    I2C    SPI0     SPI1
------------------------------------
 1    GPIOA0    SDA             MISO1
 2    GPIOA1    SCL             MOSI1
 3    GPIOA2                    SS1_B
 4    GPIOA3                    SCLK1
 5    GPIOA4           MISO0
 6    GPIOA5           MOSI0
 7    GPIOA6           SS0_B
 8    GPIOA7           SCLK0

We can now define the following pinctrl-single:

pinmux: pinmux@0xFFEE0000 {
        compatible = "pinctrl-single";
        reg = <0xFFEE0000 0x8>;
        #address-cells = <1>;
        #size-cells = <0>;
        #gpio-range-cells = <3>;
        pinctrl-single,register-width = <32>;
        pinctrl-single,function-mask = <0xffffffff>;
        pinctrl-single,gpio-range = <&range 1 8 0>;
        gpioa_pins: pinmux_gpioa_pins {
                pinctrl-single,pins = <0x0 0 0x4 0>
        };
        i2c_pins: pinmux_i2c_pins {
                pinctrl-single,pins = <0x0 1>
        };
        spi0_pins: pinmux_spi0_pins {
                pinctrl-single,pins = <0x1 1>
<0x1 1>?

If each pinmux register is only for one pin in your SoC.
I think that your definitions are wrong above. We use
register offset as the first argument, not pin number.
And the second argument should be pin function number.
In our case each pinmux register (bit field) actually controls an entire
group of pins.
quoted
If multiple pins are sharing one register with different bits,
you need to enable "pinctrl-single,bit-per-mux".
Multiple pins are sharing the same bits in the same register. Do you
think this prevents us from using pinctrl-single?
Could you give me your register definition? Then I can understand you
better.
In our example, the register map would look a bit like the following.
Note that every register configures four pins at a time.

Register 0x0:
 Mode  GPIO-A    I2C    SPI1
 Value 0x0       0x1    0x2
 ---------------------------
 Pin1  GPIOA0    SDA    MISO1
 Pin2  GPIOA1    SCL    MOSI1
 Pin3  GPIOA2           SS1_B
 Pin4  GPIOA3           SCLK1

Register 0x4:
 Mode  GPIO-A    SPI0
 Value 0x0       0x1
 ---------------------
 Pin5  GPIOA4    MISO0
 Pin6  GPIOA5    MOSI0
 Pin7  GPIOA6    SS0_B
 Pin8  GPIOA7    SCLK0
My suggestion here is that pinctrl-single isn't appropriate. The only
way it could work is if you pretend that each group-of-pins is actually
a single pin.

However, then the correlation between these pretend pins (i.e. really
the groups) and GPIOs won't work, because each "pin" is really 4 pins,
and hence 4 GPIOs, and hence you won't be able to gpio_get() more than 1
GPIO per pin group, I think.
Actually we can get each GPIO in the SoC. But we need to do some workaround.

1. As we discussed, we need to pretend a pin group as a single pin.

2. In DTS, we need to define "gpio-ranges" in gpio node and
"pinctrl-single,gpio-range"
in pinmux node as below.

    gpio {
             /* gpio offset, pin offset, nr pins */
             /* skip GPIOA1 & GPIOA3, PIN0 means pin1/pin2, PIN1 means
pin3/pin4 */
            gpio-ranges = <&pmx 0 0 1 &pmx 2 1 1>;
    };

    pmx {
              /* pin offset, nr pins, gpio function */
             pinctrl-single,gpio-range = <&range 0 1 0 &range 1 1 0>
    };
    range {
             #pinctrl-single,gpio-range-cells = <3>;
    };

Because we pretend pin1/pin2 as one single pin (PIN1), we skip to define it
in gpio-ranges. This range is only help you to find right pinmux controller.

Yes, I agree that pinctrl-single driver isn't 100% appropriate. But it
could work.
I verified it.

Regards
Haojian
It's not hard (although possibly data intensive depending on your SoC)
to represent your HW just fine with a native pinctrl driver; pinctrl
itself has the ability to separate the concepts of pins, groups-of-pins,
and the mux-functions-that-are-assigned-to-groups. If any of your HW
registers actually do control only a single pin, you can simply create
both a pin and a group that contains only that one pin. This is all very
similar to how Tegra works, although it sounds like your registers may
be a bit more regular than Tegras - Tegra has a very variable number of
pins in each grop, and even some overlap between groups (mux function
groups and pin configuration groups aren't aligned).
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help