AM335x GPIO based RTS for RS-485

Thanks for the speedy response.

It appeared the 8250 driver wasn't quite ready for rs-485 support yet, so defaulted to omap-serial. I am porting code from existing product which is using this driver just fine, but with an older kernel and without devicetree. I can't be switching kernel versions this late in our schedule, so really need to make this work on 3.14.xx.

I see in omap-serial.c where it is reading devicetree data and setting internal variables for when/how to set RTS with the defined rts-gpio. And I can see in /sys/kernel/debug/gpio where omap-serial has the correct GPIO and with our 485 app running, I never see this change. I also have a scope tapped on this signal.

cat /sys/kernel/debug/gpio

GPIOs 0-31, gpio:
 gpio-7   (vtt                     ) out hi    
 gpio-22  (omap-serial  ) out lo    

GPIOs 32-63, gpio:
 gpio-60  (sysfs               ) out lo    
 gpio-63  (sysfs               ) in  lo    

GPIOs 64-95, gpio:
 gpio-76  (sysfs               ) out lo    
 gpio-84  (sysfs               ) out lo    
 gpio-86  (sysfs               ) out lo    
 gpio-87  (sysfs               ) out hi    

GPIOs 96-127, gpio:
 gpio-99  (wlan-en-regulator   ) out hi    
 gpio-117 (kim                            ) out lo    

Below are code snippets from my latest .dts:

    uart4_pins: pinmux_uart4_pins {
        pinctrl-single,pins = <
            0x020 (PIN_OUTPUT_PULLDOWN | MUX_MODE7)    /* gpmc_ad8.gpio0_22, AM335X_GPIO0_22 */
            0x070 (PIN_INPUT_PULLUP    | MUX_MODE6)    /* gpmc_wait0.uart4_rxd, AM335X_UART4_RXD */
            0x074 (PIN_OUTPUT_PULLDOWN | MUX_MODE6)    /* gpmc_wpn.uart4_txd, AM335X_UART4_TXD */
        >;
    };

&uart4 {
    pinctrl-names = "default";
    pinctrl-0 = <&uart4_pins>;
    rts-gpio =  <&gpio0 22 0>;
    rs485-rts-delay = <0 0>;
    enable-active-high;
    rs485-rts-active-high;
    linux,rs485-enabled-at-boot-time;
    status = "okay";
};

If I don't have 'enabled-at-boot-time' declared, I am able to toggle this signal manually, via sysfs /sys/class/gpio, and change the RTS to the transceiver, allowing the Tx/Rx to operate. But this just proves that I have the correct GPIO configured for this device.

It's not like the rts-gpio signal is inverse from what transceiver needs, I never see omap-serial change the state. Even though it has 'grabbed' it.

It's probably not the issue, but I don't think "enable-active-high" should be in there at all.  The omap-serial driver looks for "rs485-rts-active-high", which you have.

I suggest adding some instrumentation into the omap-serial driver.  I don't see anything wrong with your dts aside from the superfluous "enable-active-high" I mentioned.

Mark,

The following is the Device Tree configuration that I have used successfully:

    uart4_pins: uart4_pins {
        pinctrl-single,pins = <
            0x4c (PIN_OUTPUT | MUX_MODE7)        /* gpmc_a3.gpio1[19] */
            0x70 (PIN_INPUT | MUX_MODE6)        /* gpmc_wait0.uart4_rxd */
            0x74 (PIN_OUTPUT | MUX_MODE6)        /* gpmc_wpn.uart4_txd */
        >;
    };

&uart4 {
    pinctrl-names = "default";
    pinctrl-0 = <&uart4_pins>;
    status = "okay";
    rts-gpio = <&gpio1 19 GPIO_ACTIVE_HIGH>;
    rs485-rts-active-high;
    rs485-rts-delay = <0 0>;
    linux,rs485-enabled-at-boot-time;
};

The pull-downs on the output pins are not required, but should be fine.  The pull-up on the input should also be fine.

I agree with the assessment that enable-active-high should be removed completely, and I've highlighted one difference in my configuration that upon a code review appears to me like it could be significant.  I would want to correct your device tree configuration before moving forward with further debugging.

Brad is also correct that while the omap_serial.c driver supports RS-485 TX enable on any GPIO pin, the 8250 driver that we have moved to in our latest releases only supports this on the RTS line, so keep in mind for future designs that you will have less flexibility on the RS-485 TX enable GPIO.

Thanks,

Stuart

Thank you Stuart and Brad, I appreciate the feedback.

With the above changes I still cannot see omap-serial picking up rts-gpio. I have a scope on this pin and its always low. However, when not using "rs485-enabled-at-boot-time" in the .dts, I can manually toggle the gpio value, via /sys/class/gpio/ functions, whereby I see the scope change. And, if I am running communication test apps on either side of this port I can both Tx and Rx. But, obviously I'm bypassing omap-serial and talking directly to the transceiver. This tells me hardware is OK and it is again something in the driver.

This code base is so mature, I am surprised I am having issues. Hard to explain.