AM2634: AM2634 how to started Core2 by Core0 without bootloader and image?

Hi,

Where are you keeping the image for core 2? Is SBL running on core0 copying that to TCM?

Hi，Nilabh Anand！

Maybe I didn't express myself clearly. Let me be as detailed as I can.

I want to give Core0 functions to Core2 to run, I don't know whether this can be achieved?

The overall process is as follows：

1. All of the code provided above is Core0 code，I want to give the following function (ipc_remote_core_main) to Core2 to run.

void ipc_remote_core_main()
{
    remote_start_flag = 0x01;

    while(1)
    {

    }
}

2. I manually built the vector table for Core2 (modeled after the first Section of the RPRC file), as shown below：

uint32_t tcm_memory_data[] =
{
     0xE5FF9018,0xE5FF9018,
     0xE5FF9018,0xE5FF9018,
     0xE5FF9018,0xE5FF9018,
     0xE5FF9018,0xE5FF9018,

     /* 向量表 */
     (uint32_t)&ipc_remote_core_main, // entry point
     (uint32_t)&exception_handler,    // undefined
     (uint32_t)&exception_handler,    // svc
     (uint32_t)&exception_handler,    // prefetch
     (uint32_t)&exception_handler,    // data abort
     (uint32_t)&exception_handler,    // reserved
     (uint32_t)&exception_handler,    // irq
     (uint32_t)&exception_handler,    // fiq
};

3. Then I copied the way the bootloader started the kernel, and i Instead of loading RPRC with mine code.

as shown below，you can see that I executed Bootloader_socCpuPowerOnReset to initialize TCM memory, wrote the Core2 vector table to 0x74800000, and then Bootloader_runCpu to start Core2.

void ipc_remote_core_start()
{
    uint8_t i = 0;
    int32_t status;

    Bootloader_Handle gBootloaderHandle;
    Bootloader_CpuInfo gBootImageInfo;

    gBootImageInfo.cpuId = IPC_TEST_REMOTE_CORE;
    gBootImageInfo.clkHz = Bootloader_socCpuGetClkDefault(IPC_TEST_REMOTE_CORE);

    status = Bootloader_socCpuSetClock(gBootImageInfo.cpuId, gBootImageInfo.clkHz);

    if(status != SystemP_SUCCESS)
    {
        log_i("Bootloader_socCpuSetClock failed\r\n");
        return;
    }

    status = Bootloader_socCpuPowerOnReset(IPC_TEST_REMOTE_CORE, &socCoreOpMode);

    if(status != SystemP_SUCCESS)
    {
        log_i("Bootloader_socCpuPowerOnReset failed\r\n");
        return;
    }

    uint32_t *tcm_memory_addr = (uint32_t *)0x78400000;

    for(i = 0; i < (sizeof(tcm_memory_data) / sizeof(tcm_memory_data[0])); i++)
    {
        *tcm_memory_addr++ = tcm_memory_data[i];
    }

    status = Bootloader_runCpu(NULL, &gBootImageInfo);

    if(status == SystemP_SUCCESS)
    {
        log_i("core(%d) start success\r\n", IPC_TEST_REMOTE_CORE);
    }
    else
    {
        log_i("core(%d) start failed\r\n", IPC_TEST_REMOTE_CORE);
    }
}

Thanks.

Hi ,

The core booting is done by SBL and if your intention is to just boot core 2 only then it would make sense to create a application image which contains binary of core 2 this way SBL itself would boot the second core.

Hi，Nilablh Anand.

Maybe I haven't been clear enough.

I don't want to launch Core2 through SBL, I want to launch Core2 through Core0 application, so can I customize Core2 vector table, then launch Core2 and run the vector table 0x00 function?

I mean is the first function that the kernel runs after startup on the vector table 0x00 in TCM memory? Can I set the function to Core2's Vector Table 0x00, then restart Core2 and Core2 will execute the function?

Thanks.

Hi Nilabh Anand.

As shown below, I write tcm_memory_data to 0x78400000 (Core2 TCM) and then start Core2. My expectation is that Core2 will execute the table mount ipc_remote_core_main function when it starts

/*  This table -> Write to Core2 TCM memory */
uint32_t tcm_memory_data[] =
{
     0xE5FF9018,0xE5FF9018,
     0xE5FF9018,0xE5FF9018,
     0xE5FF9018,0xE5FF9018,
     0xE5FF9018,0xE5FF9018,

     /* vector table */
     (uint32_t)&ipc_remote_core_main, // entry point?
     (uint32_t)&exception_handler,    // undefined
     (uint32_t)&exception_handler,    // svc
     (uint32_t)&exception_handler,    // prefetch
     (uint32_t)&exception_handler,    // data abort
     (uint32_t)&exception_handler,    // reserved
     (uint32_t)&exception_handler,    // irq
     (uint32_t)&exception_handler,    // fiq
};
    
    
uint32_t *tcm_memory_addr = (uint32_t *)0x78400000;

for(i = 0; i < (sizeof(tcm_memory_data) / sizeof(tcm_memory_data[0])); i++)
{
    *tcm_memory_addr++ = tcm_memory_data[i];
}

Hi 

The approach should be feasible. As long as the you are ensuing the program loaded on TCM is correct.

Hi Nilabh Anand.

 Can you provide me with relevant routines, I do not know the internal mechanism of the chip, I do not know where my code is wrong.

Thanks.