nrf_dfu_mbr_init_sd blocking bootloader after reset?

[jmr]

Hej allesammen!

I'm having trouble with the bootloader example "bootloader_secure_ble", which seems to stuck when calling nrf_dfu_mbr_init_sd. But one thing after another:

preconds: using nRF5_SDK_13.0.0, Segger embedded studio 3.26

bootloader_secure_ble has been ported by me to SES, using Bjørn's memory map and flash allocation examples from here

now, if I flash SD+MBR first like that:

nrfjprog.exe --program s132_nrf52_4.0.2_softdevice.hex -f nrf52 --chiperase

And then program the bootloader:

nrfjprog.exe --reset --program myBootloader.hex --sectoranduicrerase -f nrf52

nrfjprog does its job, resets and runs the bootloader at the end. Bootloader works as expected, I can update an application using nrf Connect and it will even be executed after a reset. I can even do a soft-reset using nrfjprog -r and the bootloader is still working.

BUT If I do a hard-reset using reset button (or cutting power), the bootloader will run but it seems getting stuck somewhere around that nrf_dfu_mbr_init_sd With enabled logging, output looks like this when working well (just after flashing as explained above):

 0> :INFO:Inside main
 0> :DEBUG:In nrf_bootloader_init
 0> :DEBUG:In real nrf_dfu_init
 0> :DEBUG:running nrf_dfu_settings_init
 0> :DEBUG:Enter nrf_dfu_continue
 0> :ERROR:Single: Invalid bank
 0> :DEBUG:Application sent bootloader request
 0> :DEBUG:In nrf_dfu_transports_init
 0> :DEBUG:num transports: 1
 0> :DEBUG:vector table: 0x0006e000
 0> :DEBUG:vector table: 0x0006e000
 0> :DEBUG:Error code - sd_softdevice_vector_table_base_set: 0x00000000
 0> :DEBUG:Before SOFTDEVICE_HANDLER_APPSH_INIT
 0> :DEBUG:After SOFTDEVICE_HANDLER_APPSH_INIT
 0> :INFO:Error code - sd_ble_cfg_set: 0x00000000
 0> :DEBUG:Enabling softdevice.
 0> SDH:DEBUG:RAM start at 0x20002100.
 0> SDH:WARNING:RAM start should be adjusted to 0x20002060.
 0> SDH:WARNING:RAM size should be adjusted to 0xdfa0.
 0> :DEBUG:Softdevice enabled
 0> :DEBUG:After nrf_dfu_transports_init
 0> :DEBUG:nrf_dfu: init dfu req handler:DEBUG:------- nrf_dfu_flash_init-------
 0> :DEBUG:Waiting for events

and after a reset of any kind, looks like that:

 0> :INFO:Inside main
 0> :DEBUG:In nrf_bootloader_init
 0> :DEBUG:In real nrf_dfu_init
 0> :DEBUG:running nrf_dfu_settings_init
 0> :DEBUG:Enter nrf_dfu_continue
 0> :ERROR:Single: Invalid bank
 0> :DEBUG:Application sent bootloader request
 0> :DEBUG:In nrf_dfu_transports_init
 0> :DEBUG:num transports: 1

and obviously, no normal bootloader operation anymore. It seems as if it got stuck somewhere somehow. Trying to locate where it got stuck pointed me to that nrf_dfu_mbr_init_sd called at line 827 in ble_stack_init in file nrf_ble_dfu.c :

uint32_t           err_code;
nrf_clock_lf_cfg_t clock_lf_cfg = NRF_CLOCK_LFCLKSRC;
if (init_softdevice)
{
    err_code = nrf_dfu_mbr_init_sd();
    VERIFY_SUCCESS(err_code);
}

NRF_LOG_DEBUG("vector table: 0x%08x\r\n", BOOTLOADER_START_ADDR);

The documentation about nrf_dfu_mbr_init_sd leaves me still clueless and the reason for above boolean switch "init_softdevice" as well. Ah.. and I have not fiddled with mergehex or any other trickery so there is no application in the flash nor bootloader settings (but the default zeroes).

Has anyone suggestions what I might have done wrong? Any hint is very much appreciated..

[Hung Bui]

Hi jmr,

Have you modified the bootloader ?

I assume when you test you don't have any buttonless DFU application, only the bootloader is flashed on the chip ?

init_softdevice: it's the flag, telling if the softdevice is already running and nrf_dfu_mbr_init_sd() can be skipped or not. You can try to skip nrf_dfu_mbr_init_sd() function and see if the bootloader still works. If it is, meaning you already initialize the softdevice prior to the bootloader.

Are you testing with nRF5 DK ? Could you updace the question to add the hex file for the bootloader so we can test here ? Also state the logging method you use (RTT or UART, which pin?)