Halt on SOFTDEVICE_HANDLER_INIT after previous access of flash via nrf_nvmc driver

[vik]

Hi all,

I am facing issues when initializing and enabling the Softdevice after using the nrf_nvmc driver to erase and write flash pages. The project shall have the ability to access write the flash before the Softdevice is enabled to update parts of our own bootloader.

In the steps to reproduce the crash on my system are:

1. System startup
2. Erasing a single flash page using nrf_nvmc_page_erase()
3. Writing to the previouse erase flash page nrf_nvmc_write_words()
4. Verify the written data -> successfull
5. Erasing the flash page again using nrf_nvmc_page_erase()
6. Initialize and enable the Softdevice via SOFTDEVICE_HANDLER_INIT()

The macro calls sd_softdevice_enable() within softdevice_handler_init(). Calling sd_softdevice_enable() halts the processor. Debugging with eclipse shows:

When we remove the erasing and writting prior SOFTDEVICE_HANDLER_INIT() the Softdevice is enabled successfully. The code is running on nRF51 and nRF52 and we are facing the same problem.

Our setup is:

• PCA10028 (V.1.0.0) and PCA10040 (V.1.1.0)
• SDK 11
• Softdevice S130 and S130 in Version 2.0.1

Any hints on that issue?

[Susheel Nuguru]

which page are you erasing and writing to? And what do you mean by halts the processor? do you mean that execution is stuck or the CPU is really halted?

[vik]

On nRF51 it is page 230. When we remove the code to erase and write the page prior enabling the Softdevice, the page is erasable and writeable with sd_flash_page_erase() and sd_flash_write() after enabling the Softdevice.

I do not think that the CPU execution is stuck. It just stops/halts and I am not able to resume it when in debug mode. Looking at the call stack it seems that the signal_handler is called somewhere in the address space. Does not make sense to me.

[Susheel Nuguru]

seems like a hardfault or APP_ERROR_CHECK failure to me ... but you have check if it is in application space or inside softdevice.

Put the breakpoint in the next line of softdevice_enable and see if it returns any error code.

[vik]

We implemented a Hardfault_handler for debugging purposes, but it is not called. We also checked other interrupt vectors but it seems that no one of them is called.

I already put a breakpoint after the sd_softdevice_enable() call but the debugger already stops directly in sd_softdevice_enable() before reaching the breakpoint as seen in the call stack above. So it seems not to be a APP_ERROR_CHECK failure.

[vik]

It would be helpfull the see what sd_softdevice_enable() is doing or somebody can verify the behaviour of the Softdevice. Currently I do net see any reason why it should not be possible to write the flash using nrf_nvmc driver and enable the Softdevice later.