Issues NCS(1.3.2) and Nordic BLE LL implementation

[Filip T.]

Hi Nordic team,

I'm having a custom board and custom application made for that board. Application is based on the zephyr v2.3.0.

I decided to move to the NCS because i want to try out your Nordic BLE LL controller instead of zephyr BLE controller.
Since the NCS version v1.3.2 uses the same zephyr v2.3.0 underhood i was able to run my application without any problems, but only if i use zephyr BLE LL (CONFIG_BT_LL_SW_SPLIT=y) config.

When configured my application to use Nordic BLE Link Layer implementation (CONFIG_BT_LL_NRFXLIB=y) my device didn't want to bootup at all.

After debugging the issue i have found that the device asserts on some condition inside the hci_core.c file:

[00:00:01.675,140] <inf> bt_ctlr_hci_driver: BLE controller build revision:
6b 05 51 03 45 62 05 0e  65 d1 a0 45 8d a7 b6 24 |k.Q.Eb.. e..E...$
3a 17 ea 5c                                      |:..\             
ASSERTION FAIL [!bt_hci_evt_is_prio(hdr->evt)] @ WEST_TOPDIR/zephyr/subsys/bluetooth/host/hci_core.c:5238
[00:00:01.675,872] <err> os: r0/a1:  0x00000003  r1/a2:  0x2002bf50  r2/a3:  0x00000000
[00:00:01.675,903] <err> os: r3/a4:  0x20030cf4 r12/ip:  0xaaaaaaaa r14/lr:  0x0005f763
[00:00:01.675,903] <err> os:  xpsr:  0x61000000
[00:00:01.675,933] <err> os: Faulting instruction address (r15/pc): 0x0005f76e
[00:00:01.675,933] <err> os: >>> ZEPHYR FATAL ERROR 3: Kernel oops on CPU 0
[00:00:01.675,964] <err> os: Current thread: 0x20019140 (BT RX)

The event code that it asserts on is the 0x0e. Looking in the hci.h file the 0x0e is the BT_HCI_EVT_CMD_COMPLETE code.

My questions would be:

1. Am i missing some of the configuration options for the BLE in order to utilize your Nordic BLE Link Layer implementation?
2. Or using Nordic BLE Link Layer is still not supported with the NCS(1.3.2) version i'm using?

My environment:

• NCS version: v1.3.2
• Hardware used:
  •  Custom board with nrf52840 chipset
• BLE stack configuration
  • CONFIG_BT_LL_NRFXLIB=y
  • CONFIG_BT_LL_NRFXLIB_DEFAULT=y
  • CONFIG_BT_LL_NRFXLIB_VS_INCLUDE=y

Thank you in advance!

[Hakon]

Hello,

it could be related to the NCSDK-8223 issue here. Can you try implementing the workaround

"Set CONFIG_BT_L2CAP_TX_BUF_COUNT >= CONFIG_BT_ATT_TX_MAX + 2."?

[Filip T.]

Hi Hakon,

I have implemented the workaround with the next configuration:

• CONFIG_BT_L2CAP_TX_BUF_COUNT=8
• CONFIG_BT_ATT_TX_MAX=5

It seems that this is not related to a known issue, as the behaviour is still the same. Later i even increased the CONFIG_BT_L2CAP_TX_BUF_COUNT=10, no results.

Do you have any further points on what could possibly cause this?

If you need more info please let me know.

[Hakon]

Looks like the assert is caused by a line in the code that has been removed in later releases. Can you try newer NCS releases and see if the problem persists?

[Filip T.]

Hi Hakon, sorry for my late response.

Prior to trying out latest NCS release version we had to port our application so that it can be compatibile with the zephyr that the latest NCS provides.

Using SoftDevice works fine now, we have been using the NCS v1.5.0/v1.5.1 versions.

Thank you for your support.