NRF5340: net core crashes in sdc_hci_cmd_le_set_adv_enable() if CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP=0

[ronbo999]

The NRF5340 net core crashes in sdc_hci_cmd_le_set_adv_enable() if I have CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP set to 0.

To reproduce the issue, do the following:

1) Use the attached peripheral_gatt_dm project. In that project, I enabled MCUBOOT which caused a linking issue that I temporarily resolved by setting CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP to 0. See devzone.nordicsemi.com/.../bug-in-ncs-v1-7-1-with-32khz-rc-oscillator-and-mcuboot-enabled as to the reason why I had to set CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP=0.

2) Add the following lines into the C:\ncs\v1.8.0\zephyr\boards\arm\nrf5340dk_nrf5340\nrf5340dk_nrf5340_cpunet_defconfig file:

CONFIG_CLOCK_CONTROL_NRF_K32SRC_XTAL=n
CONFIG_CLOCK_CONTROL_NRF_K32SRC_RC=y
CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP=0

3) Use the following command to build and flash the image:

west build -p -b nrf5340dk_nrf5340_cpuapp
west flash --erase

4) In the app core side, you'll see the following console logs:

*** Booting Zephyr OS build v2.7.0-ncs1 (Jan 21 2022 13:27:32) ***
Starting GATT Discovery Manager example
[00:00:00.082,641] <inf> bt_hci_core: HW Platform: Nordic Semiconductor (0x0002)
[00:00:00.082,672] <inf> bt_hci_core: HW Variant: nRF53x (0x0003)
[00:00:00.082,672] <inf> bt_hci_core: Firmware: Standard Bluetooth controller (0x00) Version 223.20160 Build 1719410646
[00:00:00.089,691] <inf> bt_hci_core: Identity: ED:B3:AF:58:A1:9C (random)
[00:00:00.089,691] <inf> bt_hci_core: HCI: version 5.2 (0x0b) revision 0x22b0, manufacturer 0x0059
[00:00:00.089,691] <inf> bt_hci_core: LMP: version 5.2 (0x0b) subver 0x22b0
ASSERTION FAIL [err == 0] @ WEST_TOPDIR/zephyr/subsys/bluetooth/host/hci_core.c:306
k_sem_take failed with err -11
[00:00:10.113,281] <err> os: r0/a1: 0x00000003 r1/a2: 0x00022004 r2/a3: 0x00000001
[00:00:10.113,311] <err> os: r3/a4: 0x00018061 r12/ip: 0x00001000 r14/lr: 0x00010cf9
[00:00:10.113,311] <err> os: xpsr: 0x49000000
[00:00:10.113,311] <err> os: Faulting instruction address (r15/pc): 0x00010d04
[00:00:10.113,311] <err> os: >>> ZEPHYR FATAL ERROR 3: Kernel oops on CPU 0
[00:00:10.113,311] <err> os: Current thread: 0x20000f88 (unknown)

The k_sem_take failure is caused by the net core crashing.

I would like to know if there's a way to not have to set CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP=0 to fix the linking issue or is there a work-around to prevent this crash?

The crash only occurs with the Nordic Bluetoooth Controller. It does not crash with the Zephyr Bluetooth Controller.peripheral_gatt_dm.zip

Setting the CONFIG_CLOCK_CONTROL_NRF_K32SRC_RC_CALIBRATION=n or CONFIG_CLOCK_CONTROL_NRF_K32SRC_SYNTH=y instead of CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP=0 is not an option that I could use.

[ovrebekk]

Hi 

Sorry for the slow response on this. 

The developers have looked into it, and found that for legacy reasons the MPSL library considers a CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP value of 0 special, and won't function properly in this case. 

A possible workaround is to change the value in the code here to anything but 1 (the developer tested with 0 and 2, and neither produced a crash). 

This issue will be fixed by this pull request, which will be included in the first NCS release after v1.9.0. 

Best regards
Torbjørn

[ronbo999]

Hi Torbjørn,

Will the pull request also fix the CLOCK_CONTROL_NRF_USES_TEMP_SENSOR dependency with the NET bootloader. This is the error that I get if I don't set CONFIG_CLOCK_CONTROL_NRF_CALIBRATION_MAX_SKIP=0.

=== child image b0n - CPUNET (inherited) begin ===
CMake Warning (dev) at C:/ncs/projects/SP-1253/build_nrf5340_cliplav_cpuapp/hci_rpmsg/b0n/child_image_preload.cmake:12 (set):
  Syntax error in cmake code at

    C:/ncs/projects/SP-1253/build_nrf5340_cliplav_cpuapp/hci_rpmsg/b0n/child_image_preload.cmake:12

  when parsing string

    C:\ncs\v1.7.1\toolchain\segger_embedded_studio/html/configure_nordic_project_menuconfig.py

  Invalid escape sequence \v

  Policy CMP0010 is not set: Bad variable reference syntax is an error.  Run
  "cmake --help-policy CMP0010" for policy details.  Use the cmake_policy
  command to set the policy and suppress this warning.
This warning is for project developers.  Use -Wno-dev to suppress it.

Including boilerplate (Zephyr base): C:/ncs/v1.9.0/zephyr/cmake/app/boilerplate.cmake
-- Using NCS Toolchain 1.9.0 for building. (C:/ncs/v1.9.0/toolchain/cmake)

warning: TEMP_NRF5 (defined at drivers/sensor/nrf5/Kconfig:13) has direct dependencies HAS_HW_NRF_TEMP && (MULTITHREADING || TEMP_NRF5_FORCE_ALT) && SENSOR with value n, but is currently being y-selected by the following symbols:
 - CLOCK_CONTROL_NRF_USES_TEMP_SENSOR (defined at drivers/clock_control/Kconfig.nrf:119), with value y, direct dependencies HAS_HW_NRF_TEMP && CLOCK_CONTROL_NRF_DRIVER_CALIBRATION && CLOCK_CONTROL_NRF_K32SRC_RC_CALIBRATION && CLOCK_CONTROL_NRF && CLOCK_CONTROL (value: y), and select condition HAS_HW_NRF_TEMP && CLOCK_CONTROL_NRF_DRIVER_CALIBRATION && CLOCK_CONTROL_NRF_K32SRC_RC_CALIBRATION && CLOCK_CONTROL_NRF && CLOCK_CONTROL (value: y)

error: Aborting due to Kconfig warnings

[ovrebekk]

Hi Ronald 

Sorry for the slow response, I was out in travel last week. 

I didn't realize the PR was for the MPSL library source code, which is not publicly available, my bad. 

The NCS PR can be found here: https://github.com/nrfconnect/sdk-nrfxlib/pull/675

It doesn't show the code changes, but you should be able to apply the PR to your local repo if you want to test it. 

ronbo999 said:

Will the pull request also fix the CLOCK_CONTROL_NRF_USES_TEMP_SENSOR dependency with the NET bootloader.

I have asked the team for input on this and will let you know once I hear from them. 

Best regards
Torbjørn

[ovrebekk]

Hi again

Apparently the TEMP_SENSOR issue has been reported already, here:
https://github.com/zephyrproject-rtos/zephyr/issues/40538

The issue contains a couple of workarounds, most of which should be familiar to you already (setting MAX_SKIP to 0 for instance). 

Best regards
Torbjørn