Re: Fwd: HDP profile support

[Marcel Holtmann <marcel@holtmann.org>]

Hi Nathan,

> > Is there currently a timetable in place for adding the HDP profile to
> > Bluez?  I'm doing some concept work for an Android device and one of
> > the Bluetooth requirements is to support the HDP profile.
> 
> Full HDP support (especially on the Android platform) requires a few
> supporting features that haven't quite made the mainline of development
> yet.  It is possible--I did so earlier this summer--and the situation
> has improved drastically due the efforts of various people.
> 
> The first step is to get L2CAP enhanced retransmission and streaming
> mode support in your kernel.  Gustavo Padovan worked on this over the
> summer for a Google Summer of Code project.  He has done an excellent
> job taking my initial patches, extending the functionality and cleanly
> integrating them into the bluetooth-testing kernel.

they will be in 2.6.32 btw.

> >From there, HDP depends upon the Bluetooth Multi-Channel Adaptation
> Protocol.  This provides an additional control channel layered on top of
> L2CAP sockets.  Additionally, some simple SDP work is required (which
> can use BlueZ's existing SDP services).

The part of MCAP and HDP needs to be done similar to how we did AVDTP
and AVRCP and thus as a plugin for bluetoothd. Mainly because I assume
it mostly interact with other system components. So a concept similar to
obexd is not feasible.

> As for implementation upon Android, there's a bit more work.  The
> Android kernel (on the Cupcake/Donut branches at least) already includes
> basic L2CAP statically linked into the kernel.  This causes conflicts
> when attempting to load a module with Enhanced L2CAP support built in.
> In order to do so, I've had to build my own kernel and system images,
> and then flash them to a development phone.  If you go this way, you'll
> have to backport the applicable patches onto the 2.6.29 kernel used in
> Donut.  I've never even tried to backport to the 2.6.27 kernel in
> Cupcake.

Backporting them to 2.6.29 is kinda hard and also a pretty stupid
endeavorer. I would have to check, but most likely we are talking about
100 or more patches that would need backporting. If Google would finally
get their act together and polish and merge patches upstream there would
have been no problem in just taking a newer upstream kernel.

Regards

Marcel