Update process in FLASH memory (BANK1 to BANK0) from application

[Xav]

Hi,

We use the nRF51 (nRF51822) on our products with the S130 SDK 2.0.1. In order to update the BLE application that we developed, we use the bootloader (secure bootloader DFU : SDK 12.2.0). Currently the update is done via the BLE in DFU mode.

We now want to be able to update the Nordic application via the UART TTL link with our communication protocol that we have integrated into our application. The update would be done via the application by loading the new firmware in the memory area BANK1. Once loading is complete and checked, we would like to reboot on the bootloader (secure bootloader DFU) by indicating that an update is to make from the BANK1 to the BANK0.

Would you have already implemented update process? Do you have any examples that come close to this principle? Which libraries are needed to do that? How to tell the bootloader that an update is to be done (registry, ...)?

Regards,

[Tony B]

You were in contact with Xavier about updating nRF51 application via the UART TTL link.
Unfortunatly, Xavier leaft our company last week, so I have to continue it's work about this subjet.

I saw several examples but my code seems to have a problem (besauce it is not working); here are the different steps I do :

1) calling to nrf_dfu_settings_init (in main.c) to initialize s_dfu_settings structure

2) receiving new firmware and copying at adress calculated by :
    align_to_page(BANK0_ADDRESS + (uint32_t)s_dfu_settings.bank_0.image_size)
    with BANK0_ADDRESS =0x1B000

3) registering bank 0 as invalide :
    // Set the bank-code to invalid, and reset size/CRC
    memset(&s_dfu_settings.bank_0, 0, sizeof(nrf_dfu_bank_t));

    // Reset write pointer after completed operation
    s_dfu_settings.write_offset = 0;

    // Reset SD size
    //s_dfu_settings.sd_size = 0; Not needed ????

    // Promote dual bank layout
    s_dfu_settings.bank_layout = NRF_DFU_BANK_LAYOUT_DUAL;

    // Signify that bank 0 is empty
    s_dfu_settings.bank_current = NRF_DFU_CURRENT_BANK_1;

    // s_dfu_settings.bank_1.image_size = (uint32_t)ul_LngApp; // the application length
    // s_dfu_settings.bank_1.image_crc = (uint32_t)crc32_compute((uint8_t *)fs_config.p_start_addr, ul_LngApp, NULL); // the application CRC
    s_dfu_settings.bank_1.bank_code = NRF_DFU_BANK_VALID_APP;

    ret_code_t ret_code = nrf_dfu_settings_write(NULL);



4) rebooting the nRF51 :
    sd_power_gpregret_set(GPREGRET_BOOTLOADER_DFU_START);
    // Timer for delayed reboot (at least to let the answer be sent)
    app_timer_create(&m_reboot_timer_id, APP_TIMER_MODE_SINGLE_SHOT, (app_timer_timeout_handler_t)NVIC_SystemReset);
    app_timer_start(m_reboot_timer_id, APP_TIMER_TICKS(DELAYED_REBOOT_TIME, APP_TIMER_PRESCALER), NULL);

This methods does not work properly (the firmware is not updating after reboot).
So can you help me and tell me if I forgot something ?
Do I need to calculate CRC and size of bank1, and store them into s_dfu_settings.bank_1.image_crc and s_dfu_settings.bank_1.image_size ?

The problem also is that I have to deactivate NRF_LOG because of the the application code size (so I'm "blind" and it is very difficult to see what does not work).

[Hung Bui]

Hi Tony, 
It will be very difficult to debug the issue without any logging. 
My suggestion is to try implementing on a minimal application where you can enable logging and have a working prototype before you integrate the feature to your actual application. 

In your case here, I would assume you have managed to do all the DFU task of the application (receive image, and write to bootloader setting) what we need to check is the bootloader if it would do what we expect (swap the image). Please try testing with the debug version of the bootloader where you can step through the code (you can add a break point in the bootloader and wait for the app to jump to the bootloader) 

One thing you would need to check is to make sure that the flash write to bootloader setting has finished before you do a reset. Usually we wait for the flash complete event before we trigger a reset. i assume here you are doing a timer delay instead ? 

[Tony B]

Hi Hung,

It seems that the problem comes from writing/reading the bootloader settings but I don't understand why.

The application writes settings to invalidate bank_0 and validate bank_1
Erasing and storing at 0x3FC00 before rebooting seems to be OK
here are the logs :
     0> [00000000]
     0> [00000000]:INFO:TrtMessFF92: crc                = 0xA8168847
     0> [00000000]:INFO:TrtMessFF92: settings_version   = 0x00000001
     0> [00000000]:INFO:TrtMessFF92: app_version        = 0x0000270F
     0> [00000000]:INFO:TrtMessFF92: bootloader_version = 0x000003E8
     0> [00000000]:INFO:TrtMessFF92: bank_layout        = 0x00000000
     0> [00000000]:INFO:TrtMessFF92: bank_current       = 0x00000001
     0> [00000000]:INFO:TrtMessFF92: bank_0.image_size  = 0x0000F8C4
     0> [00000000]:INFO:TrtMessFF92: bank_0.image_crc   = 0x525D1F12
     0> [00000000]:INFO:TrtMessFF92: bank_0.bank_code   = 0x00000000
     0> [00000000]:INFO:TrtMessFF92: bank_1.image_size  = 0x00008C08
     0> [00000000]:INFO:TrtMessFF92: bank_1.image_crc   = 0x2E27F565
     0> [00000000]:INFO:TrtMessFF92: bank_1.bank_code   = 0x00000001
     0> [00000000]:INFO:TrtMessFF92: write_offset       = 0x00000000
     0> [00000000]:INFO:TrtMessFF92: sd_size            = 0x00000000
     0> [00000000]:INFO:    fstorage command successfully completed (FS_EVT_ERASE)
     0> [00000000]:INFO:TrtMessFF92: ERASING BOOTLOADER SETTINGS OK :)
     0> [00000000]:INFO:    fstorage command successfully completed (FS_EVT_STORE) @ 0x3FC00
     0> [00000000]:INFO:TrtMessFF92: WRITING BOOTLOADER SETTINGS OK :)
     0> [00000000]:INFO:TrtMessFF92: reboot to DFU mode requested
     0> [00000000]:INFO:TrtMessFF92: reboot in 2000 ms
     0> [00000000]:INFO:NVIC_SystemReset_: ******** REBOOT ********


But when the bootloader restarts and reads these settings (with nrf_dfu_settings_init function), values are not the same :(
I saw that step by step with breakpoint in debug mode
Here are the calls : nrf_bootloader_init() >> nrf_dfu_init() >> nrf_dfu_settings_init();

Do you have any suggestion ?

Regards,
Tony

[Hung Bui]

Hi Tony, 

Have you checked if the data is actually stored in the flash ? 
As I mentioned in last reply, we don't do a reset right after the write command, but wait for the call back in fds before doing a reset. I can see that you are doing a 2seconds delay, it should be enough for a write, but maybe you can put a check at the end of that delay and read the flash at 0x3FC00  to see if you have the correct data written ? 

[Tony B]

Hi Hung,

Yes !! I solved my problem; here is the explanation :

First I was sure that data were correctly stored in the flash (because "fstorage command successfully completed (FS_EVT_STORE) @ 0x3FC00" message came from event callback function and I waited 2 seconds to reboot after this message).

So, my problem came from GPREGRET_BOOTLOADER_DFU_START bit that was set before reboot (then bootloader entered in bad mode).

Now, the process is running very well.

Thank you for your help,

Tony