• State TI Thinks Resolved
  • Replies 1 reply
  • Answers 1 answer
  • Subscribers 99 subscribers
  • Views 34 views
  • Users 0 members are here
Support feedback
Options
Options
Related

AM6442: Bidirectional GPIO Loopback Testing Without Device Tree Changes

Anjana Rao
Prodigy 40 points
Part Number: AM6442

Tool/software:

Hello,

We are trying to implement a GPIO loopback between the following pins on AM64x:

GPIO1_58  and GPIO1_37 

With this configuration, the loopback works in one direction — writing to GPIO1_58 and reading from GPIO1_37 works correctly.

However, when we try the reverse — writing to GPIO1_37 and reading from GPIO1_58 — it doesn't work.

How can we test bidirectional GPIO loopback without having to modify the Device Tree each time?

Here are the screenshots for the above mentioned cases:

Writing to the output pin GPIO1_58 and reading through the GPIO1_37 is successful.



But the other way loopback where writing to GPIO1_37 and reading from the GPIO1_58 is not working as expected.


Could you please suggest what might be causing this issue?

  • 0
    TI__Intellectual 1040 points

    Hello Anjana Rao,

    Do you change the configuration of the pins in the second case - setting GPIO1_37 as output (in your pin initialization routine) vs GPIO1_58 as input (in the devtree dtsi). I am asking because you do not have a screenshot of the changed dtsi configuration corresponding to your second testcase.

    Anjana Rao said:
    How can we test bidirectional GPIO loopback without having to modify the Device Tree each time?

    To avoid modifying the device tree each time, I would propose that:

    1. Within the same dtsi "static" configuration, you also enable the input LVCMOS buffer of the GPIO1_58 corresponding pin.

    You can try replacing your GPIO_A_1:GPIO1_58 line with the following line:

    gpio1_pins_default: gpio1-pins-default {
    pinctrl-single,pins = <

    AM64X_IOPAD(0x0248, PIN_INPUT, 7) /* (D16) UART1_CTSn.GPIO1_58 */

    .........................................

    2 .Then comes the question regarding GPIO1_37 and GPO1_58 output buffers control which is NOT part of the device tree, because of the need of run-time control over the pins output buffer. I am not a Linux expert, so I would describe it from a low level (hardware) perspective:

    2.1. Initially disable GPIO1_37 and GPIO1_58 pins output buffers in their corresponding padconf registers.

    Make sure that you are aware which GPIO user space command (or program routine) will disable / enable a pin output buffer, to appropriately follow the below sequencies.

    2.2. Setup the Muxmodes of the two pins accordingly.

    2.3A When GPIO1_37 shall transmit to GPIO1_58:

    First set the GPIO1_58 & GPIO1_37 output buffers disabled (to avoid output vs output contention), then set GPIO1_37 output value and then enable the GPIO1_37 output buffer in the corresponding padconf registers.

    2.3B When GPIO1_58 shall transmit to GPIO1_37 (your first usecase)

    First set the GPIO1_37 &  GPIO1_58 output buffers disabled (to avoid output vs output contention), then set GPIO1_58 output value and then enable the GPIO1_58 output buffer in the corresponding padconf registers.

    This will result in need to change control of the GPIO1_37  and GPIO1_58 output buffers in your program / terminal session when you flip the direction, but always use the same devtree dtsi configuration.   


    Please make sure that your code takes care to synchronize outputs enabling/disabling via some "direction change" protocol.

    Thanks

    Kind Regards,

    Anastas Yordanov