Re: Supporting the new Xbox One controller with wireless/BT connectivity

On 07/12/2016 01:34 AM, Benjamin Tissoires wrote:

On Mon, Jul 11, 2016 at 7:47 PM, Cameron Gutman [off-list ref] wrote:

quoted

I'm trying to come up with the best way to support wirelessly connected Xbox
One controllers. Microsoft has previously released a wireless adapter that
uses a proprietary protocol to talk to the Xbox One controllers. Recently,
they have also announced an Xbox One controller with Bluetooth connectivity.
Though we can't be sure yet since it hasn't come out yet, I would guess that
they're using the same protocol as they do when attached via USB. If it's a
standard HID gamepad, we don't have any work to do but I'm not holding my
breath ;)

That's assuming they won't be using BLE (Bluetooth 4), in which case
they will likely not use the HID BLE protocol. Most input
manufacturers don't use it when they have weird devices.

We'll see once they release.

quoted

Unfortunately, our xpad driver is quite tied to USB so it will be difficult to
support HID communication over Bluetooth. My initial thinking is that we can
add a new hid-xpad driver that will initially take over this single Bluetooth
Xbox One controller. After it is tested and feature parity with xpad, we can
possibly expand it to handle all Xbox One controllers to avoid duplicated
logic.

I don't think you'll be able to have a hid-xpad driver for the USB
controllers. From what I can read in the driver, my memories and the
various lsusb I can find on the internet, the Xbox (360 or one) are
not using HID at all. So they are not bound to HID, and unless you add
some weird logic, you can't have HID core handle non HID devices.

We might have to bend the rules of HID to make it work, but I think
it can be done. hid-sony for PlayStation controllers doesn't exactly
scream standard HID either. Microsoft does use HID to interact with
the controller on Windows, so I think we should at least explore that
possibility.

quoted

1) Write a driver for the Xbox One wireless adapter that will expose a virtual
   USB controller that enumerates Xbox One controllers and allows our existing
   xpad driver to run them. The OZWPAN driver in staging does something like
   this and would be a good reference. The advantage is that it doesn't require
   a new hid-xpad driver. The disadvantages are that there's a bunch of
   complexity in writing a virtual USB host controller driver and even after
   doing it, we still can't support Bluetooth Xbox One Controllers.

I personally don't like it as you are getting up and down in the
abstraction layers, which is doomed to be bugs prone.

Yes, I agree.

quoted

2) Write a HID transport for the Xbox One wireless adapter that enumerates HID
   devices for the hid-xpad driver. By moving the abstracting up a layer, it
   really cuts down on the amount of code required. This would also let us

If the Xbox One wireless adapter is HID, then yes, please add this hid-xpad.

The adapter itself isn't a HID device nor even really related to the Xbox
One controller at all. It's just a WLAN NIC that can connect to them and
allow a protocol driver to communicate with the controllers over the link.

The thought would be to expose a HID bus via a driver that sits atop the
WLAN interface itself. Microsoft's xboxgip.sys driver does just this.

solution 3 would be: ignore the Xbox One wireless adapter in HID and
let xpad.c handle it if both protocols are similar.

xpad.c shouldn't handle it though. We would have to write a MediaTek
NIC driver into xpad ourselves to handle it. That's why the
Xbox GIP <-> HID transport driver makes sense.

quoted

   support USB, proprietary wireless, and Bluetooth attached Xbox One
   controllers with a single xpad driver.

That would be a sweet dream. Not sure we can make it happen.

I really think we can. FWIW, Microsoft has done it. We obviously may decide
that they did it in a very messy way and not go that way, but I think it's
worth considering.

In scenario #2, the (simplified) device stacks would look like this:

For USB controller: USB hub -> xpad (possibly hid-xpad later)
For proprietary wireless adapter:
	USB hub -> MediaTek NIC driver -> xgip-transport -> hid-xpad
For BT controller (speculating): BT stack -> hid-xpad

xgip-transport would be a HID transport driver that uses the Xbox Game
Input Protocol (GIP) to enumerate and pass HID reports to each connected
controller. hid-xpad would attach to the device(s) exposed by
xgip-transport.

quoted

As far as I'm concerned, #2 seems like the obvious choice.

Jiri and Dmitry, do either of you see anything wrong with #2 or have any other
comments or concerns?

I'd say it's difficult to build a plan on future devices when we have
no information besides marketing.

In the end, we always prefer the solution that has less code
duplication, but sometimes we can't make apples looks like oranges, so
we have barely no choices but to have several similar drivers.

2 of 3 devices are out today. We're only waiting on the BT-compatible
Xbox controller. I'm definitely not advocating that people start work
on a driver for unreleased hardware based on speculation.

I just think it's worth having a little design discussion before anyone
begins working on something that will lead to nowhere or a big design
mess.

Cheers,
Benjamin

Thanks,
Cameron