nimble-linux: ble_hci_sock_init: Assertion `rc == 0' failed. about mynewt-nimble

[andrzej-kaczmarek]

ble_hci_sock_init calls should be removed from apps.

ble_transport_ll_init initializes LL side of transport which in case of external controller is a HCI socket so the socket should be initialized there.

[CW-B-W]

Hello @Fanhang32,

I'm only a user of this project, and I only played it with FreeRTOS, therefore the below might not be completely correct.

---

In mynewt-nimble v1.5.0, the architecture of HCI transport had been modified.
(You may refer to nimble/doc/transport.md for the current architecture.)

When I was migrating my personal project from v1.4.0 to v1.5.0, I found the examples at that time were not compatible with v1.5.0.
For adapting to the v1.5.0 architecture, I opened PR #1597 to resolve the issue.

Is the example you used porting/examples/linux?

As I saw porting/examples/linux has not been maintained for a long time, I think this example might be kinda outdated and need to be adapted to the new HCI transport architecture.

Would it be better if we keep ble_hci_sock_init() in ble_transport_ll_init()

mynewt-nimble/nimble/transport/socket/src/ble_hci_socket.c

Lines 836 to 841 in 64067f5

	void
	ble_transport_ll_init(void)
	{
	ble_hci_sock_init();
	}

and remove ble_hci_sock_init() in the example (porting/examples/linux)?

mynewt-nimble/porting/examples/linux/main.c

Lines 69 to 70 in 64067f5

	nimble_port_init();
	ble_hci_sock_init();

I think ble_transport_ll_init() is a more upper-level API for users to initialize the controller, while ble_hci_socket_init() may be the platform-specific implementation for ble_transport_ll_init()?

[Fanhang32]

@andrzej-kaczmarek @CW-B-W Got it , thanks a lot！ I‘ve removed ble_hci_sock_init in example apps in #1670

[Fanhang32]

If it is indeed necessary to do so，i can fix it by remove ble_hci_sock_init from ble_transport_ll_init in ble_hci_socket.c