Bluetooth shell on xG24-DK2601B // Stack overflow when calling bt init

[Dermiste]

Hi all,

I am trying to build the shell sample for my xG24-DK2601B board.

The build works fine, I can flash fine, also run a couple of commands in the shell such as help or other bt commands, but when I run bt init I get the following:

So far the only clue I found is that there is no mention of any bluetooth device in the device tree and various DTS ... could it be possible that the board definition is not complete? Other BT example seem to run fine, at least without stack overflow.

If anyone had a similar issue on a bluetooth board any help would be much appreciated.

@carlescufi or @andreeaDumitrache , any thoughts ?

[Dermiste]

Update, after looking into it I've put together a sort of call stack:
subsys/bluetooth/shell/bt.c line 1100: err = bt_enable(bt_ready);
subsys/bluetooth/host/hci_core.c line 4065: err = bt_dev.drv->open();
drivers/bluetooth/hci/slz_hci.c line 152: ret = sl_btctrl_init_ll();

I can't go any deeper, I assume it directly calls an already compiled driver afterwards.

Running an example such as beacon seems to work fine and execution then goes beyond sl_btctrl_init_ll(), so I would assume it's more a context/thread issue than a driver issue.

[Dermiste]

Update, increasing CONFIG_SHELL_STACK_SIZE to 4096 instead of 2048 allowed execution to continue properly, new error is:

A little digging shows that we are now stuck here:
subsys/bluetooth/host/hci_core.c line 4075: k_work_submit(&bt_dev.init);

Deeper, some init function are executed:
subsys/bluetooth/host/hci_core.c line 3963: init_work()

Up to this one which never returns:
subsys/bluetooth/host/hci_core.c line 3009: common_init()

Hopefully blindly increasing stack sizes will help and we are not facing a much deeper issue such as a buffer not being emptied or sth like this.

[ndrs-pst]

@Dermiste, may need to increase CONFIG_SYSTEM_WORKQUEUE_STACK_SIZE let's try to set to 2048.

By the way, the assert shown on indicate that assertion fail due to net_buf which seem link to this piece of code.

zephyr/subsys/bluetooth/host/hci_core.c

Lines 131 to 133 in 3b99fb1

	#define CMD_BUF_SIZE MAX(BT_BUF_EVT_RX_SIZE, BT_BUF_CMD_TX_SIZE)
	NET_BUF_POOL_FIXED_DEFINE(hci_cmd_pool, CONFIG_BT_BUF_CMD_TX_COUNT,
	CMD_BUF_SIZE, sizeof(struct bt_buf_data), NULL);

[Dermiste]

Both BT_BUF_EVT_RX_SIZE and BT_BUF_CMD_TX_SIZE are equal to 255, it is also maximum value I believe

zephyr/subsys/bluetooth/common/Kconfig

Lines 111 to 125 in 3b99fb1

	config BT_BUF_EVT_RX_SIZE
	int "Maximum supported HCI Event buffer length"
	default 255 if (BT_EXT_ADV && BT_OBSERVER) || BT_PER_ADV_SYNC || BT_DF_CONNECTION_CTE_RX
	# LE Read Supported Commands command complete event.
	default 68
	range 68 255
	help
	Maximum supported HCI event buffer size. This value does not include
	the HCI Event header.
	This value is used by both the Host and the Controller for buffer
	sizes that include HCI events. It should be set according to the
	expected HCI events that can be generated from the configuration.
	If the subset of possible HCI events is unknown, this should be set to
	the maximum of 255.

[ndrs-pst]

@Dermiste Another suggestion, please take a look in the following tests directory and seems that zephyr test it with nrf52840dk_nrf52840 board.

https://github.com/zephyrproject-rtos/zephyr/tree/main/tests/bluetooth/shell

[ndrs-pst]

@Dermiste, just to let you know, I got this .config file from doing the build in the tests/bluetooth/shell folder for efr32xg24_dk2601b. You might want to check it against yours to see if there's anything different.

tests_bluetooth_shell_board_efr32xg24_dk2601b.zip

[Dermiste]

Yes I figured all the test scenarios are designed for another board BUT I was definitely hoping they would work for mine, which is the main purpose of Zephyr I assume :)

Do you have the efr32xg24_dk2601b board available too? I assume the build output of yours vs mine would definitely be the same, unless you have tuned a working example.

After checking, yes .config files are the same.

[Dermiste]

Hi all, it's fixed, what did the trick for the first error:
CONFIG_SHELL_STACK_SIZE=4096

Proposed fix: in a efr32xg24_dk2601b.conf in the sample folder OR here:

zephyr/boards/arm/efr32xg24_dk2601b/Kconfig.defconfig

Lines 28 to 45 in 3b99fb1

	if BT
	config FPU
	default y
	config MINIMAL_LIBC_MALLOC_ARENA_SIZE
	default 8192
	config MAIN_STACK_SIZE
	default 2304
	choice BT_HCI_BUS_TYPE
	default BT_SILABS_HCI
	endchoice
	endif # BT
	endif # BOARD_EFR32XG24_DK2601B

And for the second error:
BT_HCI_RESERVE=1

Proposed fix:

Add a new case linked to BT_SILABS_HCI with default 1 in:

zephyr/subsys/bluetooth/host/Kconfig

Lines 75 to 93 in 3b99fb1

	config BT_HCI_RESERVE
	int
	default 0 if BT_H4
	default 1 if BT_H5
	default 1 if BT_HCI_IPC
	default 1 if BT_SPI
	default 1 if BT_STM32_IPM
	default 1 if BT_STM32WBA
	default 1 if BT_USERCHAN
	default 1 if BT_ESP32
	default 0 if BT_B91
	default 1 if BT_AMBIQ_HCI
	# Even if no driver is selected the following default is still
	# needed e.g. for unit tests.
	default 0
	help
	Headroom that the driver needs for sending and receiving buffers. Add a
	new 'default' entry for each new driver.

[ndrs-pst]

@Dermiste, Good to hear that.
So according to MAINTANERS.yml I will tag @fkokosinski and @tgorochowik for the suggestion.

[tgorochowik]

@Dermiste could you please create a PR with your suggested fix? Please link it here afterwards, we will continue the discussion there.

[Dermiste]

#67991