Usage fault on MPSL Work thread (EPSR invalid PC ~0) on nRF52832 / NCS v3.2.1

[GokulPrasanth]

Environment

I can see a Usage fault on MPSL Work with invalid PC; addr2line shows Zephyr frames for some LRs and no symbol for an address in the SoftDevice/MPSL range. NCS v3.2.1, nRF52832, . Please help triage vs. known issues or corruption, this fault is making my device crashes and restarting after a certain period of run time>
please help me to fix this?

• SDK: nRF Connect SDK v3.2.1 (Zephyr 4.2.x as bundled with NCS)
• SoC / board: nRF52832 (Thingy52)
• Bluetooth: CONFIG_BT_LL_SOFTDEVICE=y, peripheral application
• Toolchain: Zephyr SDK / arm-zephyr-eabi-*

Symptom

• Fault type: Usage fault (e.g. Illegal use of EPSR / Error Code 35 in Zephyr fault report)
• Thread: MPSL Work (or equivalent name in thread list)
• Faulting PC: Very low / invalid (e.g. ~0x0000000f), strongly suggesting invalid execution address
• Frequency: Changes ever time i think 6 min for last time

What we already did

• Ran addr2line on zephyr.elf:
  • Some LR values resolve to Zephyr sources (examples: zephyr/lib/utils/notify.c, zephyr/subsys/logging/log_output.c).
  • PC or addresses in SoftDevice Controller / MPSL range often resolve to ??:? (no debug symbols — expected for closed-source libs).

 

[Einar Thorsrud]

Hi,

I agree that it looks like an invalid address is executed for some reason. Can you try increasing the MPSL work stack size and also add a stack guard by adding this to your prj.conf?

CONFIG_MPSL_WORK_STACK_SIZE=2048
CONFIG_MPU_STACK_GUARD=y

[GokulPrasanth]

i have already done that and it's not helping

[Einar Thorsrud]

Hi,

I see. This does not resemble a known issue as far as I can find.

Can you say more about what you are doing when this happens? I which features you are using, etc? Have you been able to narrow it down somehow? Could there be some other parts of the application that writes to memory it should not and thereby corrupts it?