AM620-Q1: Failed to use mcu only mode

Hello Walter,

Lets do this step by step:

1. First, lets check using the default Linux SDK code with NO changes on a TI EVM (like SK-AM62-LP), does MCU only work?

I've tested SDK 10.1 MCU only on SK-AM62-LP which does work correctly on my end using this exact image: https://www.ti.com/tool/download/PROCESSOR-SDK-LINUX-AM62X/10.01.10.04

2. Then lets run MCU Only with your IPC echo example changes on the TI EVM. Does MCU only work?

If this does not, then it might be an caused by the IPC echo firmware and we can debug from there.

3. If that does works, then we can check on your custom hardware.

Walter Wang said:

modify the routine, using ipc_rpmsg_echo_linux_am62x-sk-lp_m4fss0-0_freertos_ti-arm-clang+EVB board test still can not work,

Please share all changes you've made to the IPC example code.

Walter Wang said:

it can not wake up in R5 serial port input,

Please explain how you are testing this wake up source. The simplest way is:

HOST PC~$ echo a > /dev/ttyUSB2

Best Regards,

Anshu

Hello Walter,

Can you also explain how the IPC echo firmware that you used was developed? 

The standard IPC RPMSG echo Linux file has all the components to support low power modes: https://github.com/TexasInstruments/mcupsdk-core-k3/blob/k3_main/examples/drivers/ipc/ipc_rpmsg_echo_linux/ipc_rpmsg_echo.c

Its recommended to use the this file as the base and build upon it.

Best Regards,

Anshu

Hi Anshu,

In the above picture, we have already shared all the modifications made in the example.

We need to first prove that the MCU continues to operate after entering the mcu only mode.

Thanks

Hi Anshu,

I update the latest phenomenon.

After pulling out the U disk, AM62 can be woked up through the RTC and the wake-up source configured by us.

The MCU still does not continue to run after sleeping, but does not restart when waking up,

but continues to run from the place stopped when sleeping

Thanks

Hi Anshu,

According to the description in the ti sdk documentation, in the "MCU Only" mode,

the MCU should be operating normally.

However, our tests have shown that the MCU did not continue to run after entering the "MCU Only" mode.

Do you have any logs to support this?

Perhaps you could refer to my modifications and then test it again, and provide the MCU log as evidence.

Thanks

Hi Walter,

After some debugging, we determined that when a USB device is connected to the processor then entered Deep Sleep or MCU only, the SoC sometimes crashes. We believe it might be caused by the TIFS which is the reason it will reset after 3 minutes because its the timer expires.

We're still trying to root cause the issue, but we were able to replicate the issue.

Thanks,

Anshu

Hi Walter,

Can you confirm if USB will be one of the possible wake up sources?

Thanks,
Anshu

Hi Walter,

Can you share the schematic for the USBx_DRVVBUS signal from the SoC?

Are there any pull resistors on that signal?

Best Regards,

Anshu

Hi Anshu,

Only test point,no pull resistors on USBx_DRVVBUS.

BR

Chason

Hi Anshu,

The test phenomena and test procedures I mentioned were all tested on the EVB board with the modified program.

It's not our own project board.

Thanks

Hello Walter

Thanks for clarifying that you are doing the experiments on the TI EVM. 

Can you clarify which version of the EVM you have ( silkscreen should have that detail) 

I believe on the EVM there is a 10K pull down on the USB1 DRVBUS that should be removed to test this feature reliably 

Additionally The VBUS_5V0_TYPEA should be 5V in Deep sleep/ MCU only mode in order for wakeup to work , you can check this at the TP203 to confirm if that is the case or not for your EVM test. 

Hi, Mukul & Anshu

Update debug status, beside USB wakeup

I think we create this thread is try to discussion why MCU do not keep working in "MCU Only mode"

Because as far as I know, MCU program should be keep running when AM62x in "MCU only mode".

Pls check attachment, and I found some problems

(1) Pls ref main,c, even we already set "MCU_GPIO0_14" as an wakeup source in linux device tree, but customer seems re-init this gpio in MCU code again, is it reasonable? because as far as I know, even "deep sleep mode" or "MCU only mode" gpio wakeup, we only support Linux device tree setting officially, isn't? Does it a concern?

static Pinmux_PerCfg_t fii_PinMuxMcuDomainCfg[] =
{
 /* MCU_GPIO0 pin config */
     /* MCU_GPIO0_14 -> MCU_MCAN0_RX (C4) MCU_CAN1_INH*/
     {
         PIN_MCU_MCAN0_RX,
         0x28054087
     },
    {PINMUX_END, 0U}
};

(2) Pls reference "ipc_rpmsg_echo.c" for function "lpm_mcu_wait_for_uart()"

Once AM62 works in mcu only mode, this function (lpm_mcu_wait_for_uart) should be keep running, and print the string "MCU is running" in while loop.

but we found this function in suspend until AM62 wakeup.

Does uart output will disable in MCU only mode?

Source code as attachment.

update status first, and keep debugging.

#if defined(ENABLE_MCU_ONLY_LPM)
static void lpm_mcu_wait_for_uart()
{
    UART_Transaction trans;
    uint8_t uartData;
    int32_t status;
    uint32_t cnt = 0U;

    UART_Transaction_init(&trans);

    /* Read 1 byte */
    trans.buf   = &uartData;
    trans.count = 1U;

    DebugP_memLogWriterPause();

    gNumBytesRead = 0u;

    while(1)
    {
        if(cnt < 0xFFFFF)
        {
            cnt++;
        }
        else
        {
            DebugP_log("MCU is running\r\n");
            cnt = 0U;
        }
# if 0
        if (cnt == 0xFFFFFF)
        {
            //SOC_triggerMcuLpmWakeup();
            /* Wait for resuming the main domain */
            //SemaphoreP_pend(&gLpmResumeSem, SystemP_WAIT_FOREVER);
            DebugP_log("cnt = %u\r\n",(cnt / 0xFFFF));
        }
#endif
    }


# if 1
    /* Wait for any key to be pressed */
    status = UART_read(gUartHandle[CONFIG_UART0], &trans);
    DebugP_assert(status == SystemP_SUCCESS);

    while (gNumBytesRead == 0u && gbSuspended == 1u)
    {
    }

    if (gNumBytesRead != 0)
    {
        DebugP_log("[IPC RPMSG ECHO] Key pressed. Notifying DM to wakeup main domain\r\n");
        SOC_triggerMcuLpmWakeup();
        /* Wait for resuming the main domain */
        SemaphoreP_pend(&gLpmResumeSem, SystemP_WAIT_FOREVER);

        DebugP_log("[IPC RPMSG ECHO] Main domain resumed due to MCU UART \r\n");
    }
    else if (gbSuspended == 0u)
    {
        UART_readCancel(gUartHandle[CONFIG_UART0], &trans);
        DebugP_log("[IPC RPMSG ECHO] Main domain resumed from a different wakeup source \r\n");
    }
#endif
    DebugP_memLogWriterResume();
}

ipc_rpmsg_echo_linux_am62x-sk-lp_m4fss0-0_freertos_ti-arm-clang.zip

Thank You.

Gibbs

Hi, Dear Expert

Update status,

Board : SK-AM62-LP

SW : 10.00.00.04

We need an example code,

we need MCU uart (M4 uart) always "keep string output"  in normal operation mode and MCU olny mode.

I think we can base on this project for modification.

--> ipc_rpmsg_echo_linux_am62x-sk-lp_m4fss0-0_freertos_ti-arm-clang

After few day testing, base on this sample project, we found MCU core (M4) is keep running, but root cause should be any I/O (like uart) which can not normal operation.

Thank You.

Gibbs

Hi, Ansu

Thanks your replies.

Here are some quick feedback and few question

Q1/A1 : When did the SDK versions change? In prior threads, SDK 10.1 was being used. 

I use SDK  SDK 10.1 for testing, customer use SDK  10.00 for testing. Both of these two version has the same problem. But basically, I think it should be coding issues.

Q2/A2 : Do you think place customized code in thread function "lpm_mcu_wait_for_uart()" is reasonable?

code modify as below, concept discussion.

we disable uart wake up, but we remain mcu_gpio make up from MCU only mode.

Because customer say they place some gpio program in this function, it also does not trigger.

I suspect it should be have RTOS Task priority problems.

static void lpm_mcu_wait_for_uart()
{
    UART_Transaction trans;
    uint8_t uartData;
    int32_t status;

    //UART_Transaction_init(&trans);

    /* Read 1 byte */
    //trans.buf   = &uartData;
    //trans.count = 1U;

    //DebugP_memLogWriterPause();

    //gNumBytesRead = 0u;

    /* Wait for any key to be pressed */
    //status = UART_read(gUartHandle[CONFIG_UART0], &trans);
    //DebugP_assert(status == SystemP_SUCCESS);

    //while (gNumBytesRead == 0u && gbSuspended == 1u)
    //{
    //}

    while(1){

	// Place customized code here
        // and do something.
	// example another uart / gpio action

    }


    // Because we do not need mcu uart wakeup
    // but keep another mcu gpio wake up
    // this gpio test in Linux device tree, and it works.
    if (gNumBytesRead != 0)
    {
        //DebugP_log("[IPC RPMSG ECHO] Key pressed. Notifying DM to wakeup main domain\r\n");
        //SOC_triggerMcuLpmWakeup();
        /* Wait for resuming the main domain */
        //SemaphoreP_pend(&gLpmResumeSem, SystemP_WAIT_FOREVER);
        //DebugP_log("[IPC RPMSG ECHO] Main domain resumed due to MCU UART \r\n");
    }
    else if (gbSuspended == 0u)
    {
        //UART_readCancel(gUartHandle[CONFIG_UART0], &trans);
        DebugP_log("[IPC RPMSG ECHO] Main domain resumed from a different wakeup source \r\n");
    }

    //DebugP_memLogWriterResume();
}

Thanks

Gibbs

The customer reported that the problem that mcu cannot control IO in mcu only mode has been resolved.

The future communication about IPC crash in resume has been moved to

https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1532839/am620-q1-ipc-crash-in-resume-from-mcu-only-mode

Hello Gibbs,

How is the customer using the MCU domain GPIO module? 

Anshu showed me a diagram where an external CAN device is connected to the MCU CAN interface, and an MCU GPIO pin.

1) Which core is controlling the CAN device on the MCU CAN interface?

2) Which core is controlling the MCU GPIO module?

lpm_mcu_wait_for_uart() - should you modify it? 

The customer can modify any part of our examples that they like. However, if they try to do something that nobody at TI has tested, then we will be limited in the support that we can offer.

The "low power GPIO wakeup" usecase that TI has tested is Linux controlling the MCU GPIO module, and Linux setting up MCU GPIO as a wakeup source as per https://software-dl.ti.com/processor-sdk-linux/esd/AM62X/10_01_10_04/exports/docs/linux/Foundational_Components/Power_Management/pm_wakeup_sources.html .

If Linux is also the owner of MCU GPIO in the customer design, I would suggest using the TI-tested wakeup implementation.

Regards,

Nick

Hi, Nick

Simple answer your question, discussion is under MCU Only Mode

Q1/A1 : Which core is controlling the CAN device on the MCU CAN interface?

MCU M4 core control MCU MCAN

Q2/A2 : Which core is controlling the MCU GPIO module?

It you talk about "wake up GPIO", I think A53 core control this wakeup gpio, because we coding it in linux device tree (DTS) .

But follow our SDK guide, we only support wakeup up gpio setting in linux DTS, isn't?

https://software-dl.ti.com/processor-sdk-linux/esd/AM62X/10_01_10_04/exports/docs/linux/Foundational_Components/Power_Management/pm_wakeup_sources.html

--> 3.3.6.2. MCU GPIO define

so I recommand customer set any wakeup source in Linux DTS, and tell them pls DO NOT USE THIS wakeup source (gpio) in MCU code.

Thanks

Gibbs

Thanks, Anshu

Pls close this thread.

Hi Walter, 

The TIFS timeout will be discussed on this thread: e2e.ti.com/.../am620-q1-tifs-timeout
Thanks,

Anshu