v01d commented on pull request #1661: URL: https://github.com/apache/incubator-nuttx/pull/1661#issuecomment-693743460
> > Debugging leads me here: > > https://github.com/apache/incubator-nuttx/blob/18cae2e78a7f4ab61a34e4e66b6b272d96b8e1a3/net/bluetooth/bluetooth_sockif.c#L909 > > > > And at that point since I'm trying to use BTPROTO_HCI this is not yet supported. As I mentioned nimBLE implements L2CAP on ther own so they produce HCI commands to be sent via the socket directly. > > That is correct BTPROTO_HCI in the **NuttX** stack is not implemented. That is what this PR will done when it is complete, but right now it is just stubbed out. > > Are you trying to run the nimble stack on top of NuttX? To do that we would need to implement the USER or RAW channels for BTPROTO_HCI. These were low priority for me because they bypass the kernel, instead userspace would use the L2CAP for GATT and (BTPROTO_HCI -- CONTROL_CHANNEL) for doing things like setting advertisements, scanning, etc... Correct. Sorry, I think I didn't gave enough context. I started last week looking into how to adapt NuttX host layer to work with L2CAP sockets, as I mentioned ealier here, but felt a bit overwhelmed and didn't know how to start. In the meantime, I stumbled upon nimBLE and realized that as it supports linux via RAW HCI sockets, considering you just added support for HCI using the sim, I thought maybe I could try to port nimBLE (at least to have another option for a host layer and also because it seemed a less daunting task). So I understand that supporting RAW HCI socket would not be part of this PR since you would implement the control channel and the rest of the communication would happen over L2CAP socket, right? Maybe I can see if it isn't too difficult to add the RAW option. Anyway, maybe it is best that I wait until this is merged, otherwise we would step on each other. > > What I was asking about was what interfaces were you running into that l2cap did not provide for creating the real NuttX GATT server/client? Because we could make sure to implement those first (should be one of the mgmt APIs that I pointed out) Well, the first point is that NuttX host layer should become a userspace library that abstracts L2CAP socket access. So to do that, the point of contact with L2CAP, at the ATT layer, needs to be modified to interfaces via sockets not direct calls. The point is that this change bubbles up to the GATT functions, since there's some structures used to identify an L2CAP connection all the way to the top of the host layer. Since now this would be replaced with a socket fd possibly, I didn't know how to proceed since the change will probably be quite massive. Besides this transformation, I also noticed that there does not seem to be a way to define a GATT server. The macros and such to declare a service are there, but I think there's something missing that takes a GATT service definition and creates the server itself. Do note that I'm not really familiar with all of this, I'm learning as I go, so maybe I'm confused in some areas. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
