AM263P4: R5FSS0-1 TCMB problem of SBL OSPI multicore elf and HS-SE state

Jeeuk Chang

Intellectual 290 points

Part Number: AM263P4
Other Parts Discussed in Thread: UNIFLASH, SYSCONFIG

Hi, TI experts.

The first phase of my AM263P4 development work is almost complete in HS-FS state.

Before product test in customer site, I am trying to changing my board to secure mode.

(1) I made one of my custom boards to HS-SE device.

(2) I made HSMRt firmware for my HS-SE device.

(3) I modified devconfig.mak from the original file below

DEVICE_TYPE?=HS
KD_SALT={somewhere}/kd_salt.txt
CUST_MPK={somewhere}/mcu_custMpk.pem
CUST_MEK={somewhere}/mcu_custMek.key
ENC_ENABLED?=yes

(3) I built SBL_UART_UNIFLASH and SBL_OSPI again for my HS-SE device with modified devconfig.mak and HSMRt firmware.

(4) I built my application again for my HS-SE device with modified devconfig.mak.

Then I boot the result in OSPI boot mode but the progress is stopped at SBL_OSPI program.

After some inspection of SBL_OSPI program progress, I found the stop point below.

status = Bootloader_parseAndLoadMultiCoreELF(bootHandle, &bootImageInfo);

Instead of my application, I tested empty system project in the same process and succeeded in operation of the program.

After changing and modification of project files, I conclude that if TCMB of R5FSS0-1 is used in HS-SE state, SBL OSPI stops in Bootloader_parseAndLoadMultiCoreELF function.

SBL OSPI program had no problem in HS-FS state.
Only R5FSS0-1 TCMB made a problem in SBL OSPI boot.
When I removed initialized variable and executable code in R5FSS0-1 TCMB, SBL OSPI had no problem.

I attached my simple test codes for R5FSS0-1 of empty system project.

empty.c (a modified file)

#include <stdio.h>
#include "ti_drivers_config.h"
#include "ti_drivers_open_close.h"
#include "ti_board_open_close.h"


volatile float temp_data;

// case 1 : initialized data in TCMB
float data_preset __attribute__((section(".data.TCMB"))) = {0};

// case 2 : place code in TCMB
void __attribute__((section(".text.TCMB"))) background_variable(void)
{
    // make this function size not too small in order not to make this function optimized
    temp_data = data_preset;    temp_data = data_preset+1;  temp_data = data_preset+2;
    temp_data = data_preset+3;  temp_data = data_preset+4;  temp_data = data_preset+5;
    temp_data = data_preset+6;  temp_data = data_preset+7;  temp_data = data_preset+8;
    temp_data = data_preset+9;  temp_data = data_preset+10;
}

void empty_main(void *args)
{
    Drivers_open();
    Board_driversOpen();

    while(1)
    {
        background_variable();
    }

    Board_driversClose();
    Drivers_close();
}

user.cmd (an additional file)

SECTIONS
{
    .data.TCMA : {} > R5F_TCMA, palign(8)
    .data.TCMB : {} > R5F_TCMB, palign(8)
    .data.OCRAM : {} > OCRAM, palign(8)

    .text.TCMA : {} > R5F_TCMA, palign(8)
    .text.TCMB : {} > R5F_TCMB, palign(8)
    .text.OCRAM : {} > OCRAM, palign(8)
}

MEMORY
{
}

Case 1 or Case 2 make the memory usage in TCMB.

(Not-initialized data defined in TCMB did not make a problem.)

My application must use TCMB memory due to its size and execution speed limitations.

May I get the solution for this problem as soon as possible?

Regard,

Jeeuk Chang

5 months ago

0 Nilabh Anand 5 months ago

TI__Guru 60570 points

Hi Jeeuk,

It could be related to firewall settings in HSSE state, I can check it early next week and get back to you.

0 Jeeuk Chang 5 months ago in reply to Nilabh Anand

Intellectual 290 points

Hi Nilabh Anand.

I compared actual MPU firewall settings (sysconfig) and firewall configuration table in the TIFS document.

I found some settings in sysconfig are slightly different in their size (End Address).

4-4, 5-4, 6-4, 7-4, 10-2, 11-1

And some settings in sysconfig are different in permission CPUID.

14-0, 15-0, 16-0, 17-0, 19-0, 19-1, 19-2, 20-0

And new items are found in sysconfig.

10-5, 10-6

But I thought those changes are not important.

Rather, I thought TCMB sizes in sysconfig and TIFS table are mismatched to actual TCMB sizes (96kB in AM263P4).

4-1, 5-1, 6-1, 7-1

I tried to make the area size enlarge but sysconfig blocked the trials.

This is my another question but I leaved it for later.

If you check MPU settings again, please answer my question about TCMB size settings also.

Regard,

Jeeuk Chang

0 Nikhil Dasan 4 months ago in reply to Jeeuk Chang

TI__Guru* 89881 points

Hi Jeeuk Chang,

Sorry for delay in response as Nilabh is OOO.

I'm trying this out at my end.

Meanwhile, may I know at what stage in the Bootloader_parseAndLoadMultiCoreELF , is it getting stuck?

It is working in case of non-initialized variables, because these regions are not parsed by the API.

Thanks and Regards,

Nikhil Dasan

0 Jeeuk Chang 4 months ago in reply to Nikhil Dasan

Intellectual 290 points

Hi, Nikhil Dasan.

Unfortunately I do not know the secure debug method yet.

I used log outputs via UART at every steps of SBL OSPI and found the stop point.

Regards,

Jeeuk Chang

0 Jeeuk Chang 4 months ago in reply to Jeeuk Chang

Intellectual 290 points

Hi everyone.

It has been almost 20 days since I posted the question.

Because the bootloader function is located within the library, making it difficult to modify or trace them in the HS-SE environment.

Before identifying the cause, I am curious whether the same issue occurs on the HS-SE boards of TI.

When can I expect to hear an update on the progress?

Regards,

Jeeuk Chang

+1 Jeeuk Chang 3 months ago in reply to Jeeuk Chang

Intellectual 290 points

Hi everyone.

I made another thread related to this question.

AM263P4: MPU firewall and SBL booting problem of HS-SE state - Arm-based microcontrollers forum - Arm-based microcontrollers - TI E2E support forums

Regards,

Jeeuk Chang