Uart RTS Not Working

[bwilcutt]

For RS485, I've enabled the rts gpio on my uart using OS v4.1:

            uart1: serial@48022000 {
                    compatible = "ti,omap-serial-rs485";
                    ti,hwmods = "uart2";
                    clock-frequency = <0x2dc6c00>;
                    reg = <0x48022000 0x2000>;
                    interrupts = <0x49>;
                    dmas = <&edma 0x1c &edma 0x1d>;
                    dma-names = "tx", "rx";
                    rs485-rts-active-high;
                    rs485-rts-delay = <0 0>;
                    rts-gpio = <&gpio3 19 GPIO_ACTIVE_HIGH>;
                    linux,rs485-enabled-at-boot-time;
                    status = "okay";
            };

gpio3 19 points to P9_27 which is my rts line as verified by a scope (and a working program on the 3.14 of the OS).

My serial driver does successfully find the gpio ("rts-gpio") however I get a kernel crash during boot up:

Running uname_boot ...
loading /boot/vmlinuz-4.1.13-ti-rt-r77 ...
[    5.576334] kobject: 'omap_serial_rs485' (bf003550): kobject_add_internal: parent: 'module', set: 'module'
[    5.576402] kobject: 'holders' (df555500): kobject_add_internal: parent: 'omap_serial_rs485', set: '<NULL>'
[    5.576589] kobject: 'notes' (df58cc80): kobject_add_internal: parent: 'omap_serial_rs485', set: '<NULL>'
[    5.576618] kobject: 'omap_serial_rs485' (bf003550): kobject_uevent_env
[    5.576638] kobject: 'omap_serial_rs485' (bf003550): fill_kobj_path: path = '/module/omap_serial_rs485'
[    5.585686] kobject: 'omap-serial-rs485' (df4efc80): kobject_add_internal: parent: 'drivers', set: 'drivers'
[    5.586061] of_get_named_gpiod_flags: parsed 'rts-gpio' property of node '/ocp/serial@48022000[0]' - status (0)
[    5.586085] Unhandled fault: external abort on non-linefetch (0x1028) at 0xfa1ae134
[    5.586095] pgd = df5d4000
[    5.586103] [fa1ae134] *pgd=48011452(bad)
[    5.586128] Internal error: : 1028 [#1] PREEMPT SMP ARM
[    5.586140] Modules linked in: omap_serial_rs485(+)
[    5.586165] CPU: 0 PID: 174 Comm: modprobe Not tainted 4.1.13-ti-rt-r77 #262
[    5.586174] Hardware name: Generic AM33XX (Flattened Device Tree)
[    5.586184] task: df5daac0 ti: df50a000 task.ti: df50a000

This code worked in OS 3.14 but not in 4.1. The biggest change, for this system, is the dts/dtsi files which I've tried to make sure were set up properly.

The crash occurs during the gpio_direction_output function call, see below..
.
.
.
if (omap_up_info->tx_enable_gpio = of_get_named_gpio(dev->of_node, "txe-gpio", 0))
{
gpio_direction_output(omap_up_info->tx_enable_gpio, PIN_HIGH);
.
.
.

By manipulating P9_27 (gpio 115) directly, the problem goes away and the software works as designed.

root@localhost:~# cd /sys/class/gpio/
root@localhost:/sys/class/gpio# echo 115 > export
root@localhost:/sys/class/gpio# cd gpio115
root@localhost:/sys/class/gpio/gpio115# ls
active_low  device  direction  edge  power  subsystem  uevent  value
root@localhost:/sys/class/gpio/gpio115# cat value 
0
root@localhost:/sys/class/gpio/gpio115# cat active_low  
0
root@localhost:/sys/class/gpio/gpio115# ls
active_low  device  direction  edge  power  subsystem  uevent  value
root@localhost:/sys/class/gpio/gpio115# cat direction 
in
root@localhost:/sys/class/gpio/gpio115# echo "out" > direction 
root@localhost:/sys/class/gpio/gpio115#                

Note that the pin is already configured to be "out" yet the devicetree is reporting it set to "in".

[bwilcutt]

PROBLEM SOLVED:

Perhaps it was "magic" that my original code worked but from 3.14 to 4.1 I had to add the following lines in my probe function to get the code to work properly:

omap_up_info->tx_enable_gpio = of_get_named_gpio(dev->of_node, "txe-gpio", 0);
if (gpio_is_valid(omap_up_info->tx_enable_gpio)) {
        ret = gpio_request(omap_up_info->tx_enable_gpio, "omap-serial-rs485");
} else {
        return -EPROBE_DEFER;
}
gpio_direction_output(omap_up_info->tx_enable_gpio, 0);

I was unsuccessful at getting the gpio (115) to be set as "out" through the device tree and it's obfuscations. The "gpio_request" function turned the 'direction' from 'in' to 'out'. Go figure.

Dr. B
"I drank what?!" --Socrates