AM625: Regarding Flash Write Issue During Nominal Download of Huge Data

Hello,

Allow me sometime to comment on this.

Thanks,

Vaibhav

Hi Pradeep,

Currently the observed behaviour is Read Back Data does not match with the known data and a bunch of FFh is seen randomly.

Can you please perform the following steps to verify if this is purely a write or read issue, as it is not an erase issue for sure:

1. Write x number of bytes, in your case, write the usual 1 MB of byte.
2. Now before reading back the data into a RX Buffer, pause and enable the dac bit, that is 7th bit of the register FC40000.
3. Here, I want you to go the offset where the flashing was done, so basically go to 0x60000000 + WRITEOFFSET and then check if the flash contents shows the random FFh in the middle or not?

If when DAC is enabled, and you see those random FFh, then write has not been performed correctly, and reading works fine. If you do not see the FFh randomly then this can be categorized as a read failure.

Hope this gives clarification on what needs to be performed.

Looking forward to the results.

Kind Regards,

Vaibhav

Hi Vaibhav,

We enabled dac bit in the Fls configuration for debugging, please find attached files bin.zip application.bin(input application) and dump_from_flash.bin(output application dump from flash).

1882.bin.zip

Please let us know if you need further information.

BR,

Shubham

Hi,

Looking at the flash dump, there are random places where the flash has not written correctly and on addition to this, I see blocks of 32 KB not written.

Please perform the following:

loop 1024 times:
        write 1024 bytes at a time
This ensure 1 MB written
Now check the flash area and see if the written value matches the application binary. It should ideally match as appotronics tested for 1 KB writes for sure.

Regards,

Vaibhav

Hi Vaibhav,

As shared bin.zip for application.bin(input application) and dump_from_flash.bin(output application dump from flash), this we have already configured with 1024 Buffer size with1024 times.

Please let us know if we can check anything.

BR,

Shubham

Hi Shubham,

Shubham Goel said:

As shared bin.zip for application.bin(input application) and dump_from_flash.bin(output application dump from flash), this we have already configured with 1024 Buffer size with1024 times.

Please let us know if we can check anything.

Thanks, this should be fine, lets wait for Appotronic's comments on the way they tested MCAL INDAC example. As it seems like they did not test independently without SBL NULL/SBL OSPI.

Please check the email for latest update from their side.

Thanks,

Vaibhav

Sharing updated Fls_delay and OSPI_flashExecCmd

 static Std_ReturnType OSPI_flashExecCmd(const CSL_ospi_flash_cfgRegs *pRegs)
 {
    uint32 retry = OSPI_DAC_WRITE_TIMEOUT;
    Std_ReturnType  retVal = E_OK;

    volatile uint32 tempCount = FLS_TIMEOUT_DURATION;

    if (FLS_TIMEOUT_DURATION > 9U)
    {
        /* Each unit of SW_delay equals to 9 clockcycles, so dividing by 9U */
        tempCount = (FLS_TIMEOUT_DURATION / 9U);
    }
    /* Wait for FSM to to come out of reset */
    do
    {
        if (tempCount == 0U)        
        {
            /* Timeout */
            break;
        }
        MCAL_SW_DELAY(tempCount);
    }while (CSL_ospiIsIdle(pRegs) == FALSE);

    /* Start to execute flash read/write command */
    CSL_ospiFlashExecCmd(pRegs);
    
    while (retry != 0U)
    {
        /* Check the command execution status */
        if (CSL_ospiFlashExecCmdComplete(pRegs) == TRUE)
        {
            break;
        }
        OSPI_delay(OSPI_CHECK_IDLE_DELAY);
        retry--;
    }

    if (retry == 0U)
    {
        retVal = E_NOT_OK;
    }

    if (FLS_TIMEOUT_DURATION > 9U)
    {
        /* Each unit of SW_delay equals to 9 clockcycles, so dividing by 9U */
        tempCount = (FLS_TIMEOUT_DURATION / 9U);
    }
    do
    {
        if (tempCount == 0U)        
        {
            /* Timeout */
            break;
        }
        MCAL_SW_DELAY(tempCount);
    }while (CSL_ospiIsIdle(pRegs) == FALSE);

    return (retVal);
 }
 
 static void Fls_delay(uint32 msecs)
 {
     volatile uint32 tempCount = 1000000U * msecs;

     /* Each unit of MCAL_SW_DELAY() equals ~9 clock cycles,
        so we scale tempCount accordingly */
     if (msecs > 9U) 
     {
         tempCount = msecs / 9U;
     }
 
     /* Wait for the delay to complete */
     do 
     {
         if (tempCount == 0U) 
         {
             /* Timeout reached */
             break;
         }
         MCAL_SW_DELAY(tempCount);
     } while (tempCount > 0U);
 }

Kind Regards,

Vaibhav

Hello Pradeep,

Thanks for the update.

Kind Regards,

Vaibhav