Fix processing of received ESP_HCI_IF (EHM-77)

[OYTIS]

Greetings!

We have found that in order for esp-hosted-mcu firmware to work as an HCI controller for our host, we need to get rid of decrementing the payload pointer. We are not using the host code from this repo, but are reasonably sure that our outgoing ESP_HCI_IF packets are well-formed, consisting of esp_payload_header followed by H4 HCI packet. The last esp-idf version we've been testing with is espressif/esp-idf@67e5e59

[mantriyogesh]

@SohKamYung-Espressif PTAL..

[SohKamYung-Espressif]

Hi @OYTIS

In our low level transport code for the host, for example here for SDIO

esp-hosted-mcu/host/drivers/transport/sdio/sdio_drv.c

Line 431 in 70e9ffd

// copy first byte of payload into header

we move the HCI header into the Hosted payload header and copy the payload, less the header. On the slave, this HCI header is restored by moving the pointer back one byte so it points to the HCI header data in the Hosted payload header.

This was done to maintain compatibility with the slave BT code for the ESP-Hosted-FG implementation.

Can you check your host transport implementation whether it also moves the HCI header to the Hosted payload header?

[OYTIS]

@SohKamYung-Espressif Thank you! We had to change it for an earlier version too, but never made the change public.

So as I understand this can't be fixed, because it would break backwards compatibility, so instead host code should deviate from the protocol?

[SohKamYung-Espressif]

@OYTIS

So as I understand this can't be fixed, because it would break backwards compatibility, so instead host code should deviate from the protocol?

Yes, the change would break backward compatibility in ESP-Hosted-MCU.

[OYTIS]

@SohKamYung-Espressif Got it! How valuable is the backwards compatibility here? Say, protobuf protocol seems to be rewritten, so the host code would not be compatible with esp-hosted-fg.

If it's a good point to break it, we could fix it on the host side too.

[SohKamYung-Espressif]

How valuable is the backwards compatibility here?

We may also want to unify the underlying slave transport code between ESP-Hosted-FG and ESP-Hosted-MCU in the future, so we want to keep the underlying transport details the same.

The RPC Protobuf commands between FG and MCU differ, but this in the upper layer. The underlying transport still remains the same for both.