2020-10-11 18:00:13

by Peter Mullen

[permalink] [raw]
Subject: Bose QC 35 Battery/ANC Support

Hi all,

I've been working on a plugin that adds support for battery level
reporting and Active Noise Cancelling (ANC) control for the Bose QC 35
headphones. The patch is nearly ready to go, but because of the
non-standard way it's implemented I figured it might be necessary to get
some preliminary feedback on it prior to submission (or to see if it's
worth submitting).

As a brief overview of device control, the QC35 has a set of additional
controls operating over rfcomm channel 8 where messages are passed back
and forth between devices. The messages consist of a 3-byte opcode, a
1-byte payload length, and N-bytes of payload. There's some more
information on the specifics here:
https://blog.davidventura.com.ar/reverse-engineering-the-bose-qc35-bluetooth-protocol.html

As this is a non-standard "profile", there's no profile UUID assigned to
it which makes adding a profile somewhat more complicated. My solution
here was to add a profile under the iPod Accessory Protocol (iAP)
profile UUID which the QC35 lists, and to then filter out devices in the
profile probe callback based on manufacturer ID, device class, and
product ID. This isn't ideal as the probe will be called for any device
listing the iAP UUID, so suggestions for alternative approaches (if
necessary) are appreciated.

For session state control, I've hooked into the AVDTP state-change
callback. Again, possibly non-standard but this seemed to be the best
way to trigger the initiatiation/tearing down of the rfcomm connection.

For battery level integration, I've basically just wrapped the dbus
interface from the standard GATT battery profile. This has been working
fine with my DE; the battery level shows immediately after connection
and updates as the level drops.

For ANC integration, I've added a new dbus interface under
"org.bluez.Anc1". The interface contains a read-only "Range" property,
which indicates the number of discrete values the ANC can be set to, and
a read-write property "Level" which gets/sets the level. This interface
layout was chosen to be device agnostic, so that ANC support could
potentially be added for more devices in future.

Let me know if there are any thoughts on this; I can get the patch ready
and submitted if desired.

Kind regards,
Peter


2020-10-11 18:02:40

by Bastien Nocera

[permalink] [raw]
Subject: Re: Bose QC 35 Battery/ANC Support

On Sat, 2020-10-10 at 22:00 -0700, Peter Mullen wrote:
> For battery level integration, I've basically just wrapped the dbus
> interface from the standard GATT battery profile. This has been
> working
> fine with my DE; the battery level shows immediately after connection
> and updates as the level drops.

Glad this works. I'd like to eventually look at the code you have so
far for this functionality, so don't hesitate to share it.

Cheers

2020-10-12 16:47:39

by Luiz Augusto von Dentz

[permalink] [raw]
Subject: Re: Bose QC 35 Battery/ANC Support

Hi Peter,

On Sun, Oct 11, 2020 at 1:04 PM Peter Mullen <[email protected]> wrote:
>
> Hi all,
>
> I've been working on a plugin that adds support for battery level
> reporting and Active Noise Cancelling (ANC) control for the Bose QC 35
> headphones. The patch is nearly ready to go, but because of the
> non-standard way it's implemented I figured it might be necessary to get
> some preliminary feedback on it prior to submission (or to see if it's
> worth submitting).
>
> As a brief overview of device control, the QC35 has a set of additional
> controls operating over rfcomm channel 8 where messages are passed back
> and forth between devices. The messages consist of a 3-byte opcode, a
> 1-byte payload length, and N-bytes of payload. There's some more
> information on the specifics here:
> https://blog.davidventura.com.ar/reverse-engineering-the-bose-qc35-bluetooth-protocol.html
>
> As this is a non-standard "profile", there's no profile UUID assigned to
> it which makes adding a profile somewhat more complicated. My solution
> here was to add a profile under the iPod Accessory Protocol (iAP)
> profile UUID which the QC35 lists, and to then filter out devices in the
> profile probe callback based on manufacturer ID, device class, and
> product ID. This isn't ideal as the probe will be called for any device
> listing the iAP UUID, so suggestions for alternative approaches (if
> necessary) are appreciated.

We could in possibly add pid/vid to btd_profile that way it can be
probed by pid/vid in addition to UUID, that way one can make a driver
that is specific to a vendor or a product.

> For session state control, I've hooked into the AVDTP state-change
> callback. Again, possibly non-standard but this seemed to be the best
> way to trigger the initiatiation/tearing down of the rfcomm connection.

I would recommend hooking to the btd_service state, like for example
the policy plugin is doing.

> For battery level integration, I've basically just wrapped the dbus
> interface from the standard GATT battery profile. This has been working
> fine with my DE; the battery level shows immediately after connection
> and updates as the level drops.

We had some discussion on weather we should continue using a custom
D-Bus interface or just emulate the battery reports over UHID given
that is probably better integrated with the system.

> For ANC integration, I've added a new dbus interface under
> "org.bluez.Anc1". The interface contains a read-only "Range" property,
> which indicates the number of discrete values the ANC can be set to, and
> a read-write property "Level" which gets/sets the level. This interface
> layout was chosen to be device agnostic, so that ANC support could
> potentially be added for more devices in future.

I'd probably add Bose to the name of the interface if it is bose specific.

> Let me know if there are any thoughts on this; I can get the patch ready
> and submitted if desired.
>
> Kind regards,
> Peter
>


--
Luiz Augusto von Dentz

2020-10-14 09:20:17

by Peter Mullen

[permalink] [raw]
Subject: Re: Bose QC 35 Battery/ANC Support

Hi Luiz,

On 2020-10-12 9:43 a.m., Luiz Augusto von Dentz wrote:
> Hi Peter,
>
> On Sun, Oct 11, 2020 at 1:04 PM Peter Mullen <[email protected]> wrote:
>>
>> Hi all,
>>
>> I've been working on a plugin that adds support for battery level
>> reporting and Active Noise Cancelling (ANC) control for the Bose QC 35
>> headphones. The patch is nearly ready to go, but because of the
>> non-standard way it's implemented I figured it might be necessary to get
>> some preliminary feedback on it prior to submission (or to see if it's
>> worth submitting).
>>
>> As a brief overview of device control, the QC35 has a set of additional
>> controls operating over rfcomm channel 8 where messages are passed back
>> and forth between devices. The messages consist of a 3-byte opcode, a
>> 1-byte payload length, and N-bytes of payload. There's some more
>> information on the specifics here:
>> https://blog.davidventura.com.ar/reverse-engineering-the-bose-qc35-bluetooth-protocol.html
>>
>> As this is a non-standard "profile", there's no profile UUID assigned to
>> it which makes adding a profile somewhat more complicated. My solution
>> here was to add a profile under the iPod Accessory Protocol (iAP)
>> profile UUID which the QC35 lists, and to then filter out devices in the
>> profile probe callback based on manufacturer ID, device class, and
>> product ID. This isn't ideal as the probe will be called for any device
>> listing the iAP UUID, so suggestions for alternative approaches (if
>> necessary) are appreciated.
>
> We could in possibly add pid/vid to btd_profile that way it can be
> probed by pid/vid in addition to UUID, that way one can make a driver
> that is specific to a vendor or a product.
Had a go at that this evening as a proof-of-concept; so far it's working
well, so this could be viable. I'll finish it off and test it out properly.

>> For session state control, I've hooked into the AVDTP state-change
>> callback. Again, possibly non-standard but this seemed to be the best
>> way to trigger the initiatiation/tearing down of the rfcomm connection.
>
> I would recommend hooking to the btd_service state, like for example
> the policy plugin is doing.
I was able to get the plugin working based on the btd_service state cb,
but had a couple of concerns with it, mainly because we'll be receiving
events for all services on all devices, so the filtering of unused
events adds a bit of overhead.
I was also able to use the AVCTP state callback without issue, which may
be better suited than the AVDTP one.

>> For ANC integration, I've added a new dbus interface under
>> "org.bluez.Anc1". The interface contains a read-only "Range" property,
>> which indicates the number of discrete values the ANC can be set to, and
>> a read-write property "Level" which gets/sets the level. This interface
>> layout was chosen to be device agnostic, so that ANC support could
>> potentially be added for more devices in future.
>
> I'd probably add Bose to the name of the interface if it is bose specific.
While the plugin is Bose specific, I had tried to keep the interface
itself generic so that integration with the DE or similar won't need any
changes if ANC support is added for other devices later.

--
Peter Mullen

2020-10-14 17:43:41

by Luiz Augusto von Dentz

[permalink] [raw]
Subject: Re: Bose QC 35 Battery/ANC Support

Hi Peter,

On Tue, Oct 13, 2020 at 10:06 PM Peter Mullen <[email protected]> wrote:
>
> Hi Luiz,
>
> On 2020-10-12 9:43 a.m., Luiz Augusto von Dentz wrote:
> > Hi Peter,
> >
> > On Sun, Oct 11, 2020 at 1:04 PM Peter Mullen <[email protected]> wrote:
> >>
> >> Hi all,
> >>
> >> I've been working on a plugin that adds support for battery level
> >> reporting and Active Noise Cancelling (ANC) control for the Bose QC 35
> >> headphones. The patch is nearly ready to go, but because of the
> >> non-standard way it's implemented I figured it might be necessary to get
> >> some preliminary feedback on it prior to submission (or to see if it's
> >> worth submitting).
> >>
> >> As a brief overview of device control, the QC35 has a set of additional
> >> controls operating over rfcomm channel 8 where messages are passed back
> >> and forth between devices. The messages consist of a 3-byte opcode, a
> >> 1-byte payload length, and N-bytes of payload. There's some more
> >> information on the specifics here:
> >> https://blog.davidventura.com.ar/reverse-engineering-the-bose-qc35-bluetooth-protocol.html
> >>
> >> As this is a non-standard "profile", there's no profile UUID assigned to
> >> it which makes adding a profile somewhat more complicated. My solution
> >> here was to add a profile under the iPod Accessory Protocol (iAP)
> >> profile UUID which the QC35 lists, and to then filter out devices in the
> >> profile probe callback based on manufacturer ID, device class, and
> >> product ID. This isn't ideal as the probe will be called for any device
> >> listing the iAP UUID, so suggestions for alternative approaches (if
> >> necessary) are appreciated.
> >
> > We could in possibly add pid/vid to btd_profile that way it can be
> > probed by pid/vid in addition to UUID, that way one can make a driver
> > that is specific to a vendor or a product.
> Had a go at that this evening as a proof-of-concept; so far it's working
> well, so this could be viable. I'll finish it off and test it out properly.
>
> >> For session state control, I've hooked into the AVDTP state-change
> >> callback. Again, possibly non-standard but this seemed to be the best
> >> way to trigger the initiatiation/tearing down of the rfcomm connection.
> >
> > I would recommend hooking to the btd_service state, like for example
> > the policy plugin is doing.
> I was able to get the plugin working based on the btd_service state cb,
> but had a couple of concerns with it, mainly because we'll be receiving
> events for all services on all devices, so the filtering of unused
> events adds a bit of overhead.
> I was also able to use the AVCTP state callback without issue, which may
> be better suited than the AVDTP one.

I guess we could add support for subscribing to state changes of a
specific service, that way you don't need to filter anything at the
plugin, how about that?

> >> For ANC integration, I've added a new dbus interface under
> >> "org.bluez.Anc1". The interface contains a read-only "Range" property,
> >> which indicates the number of discrete values the ANC can be set to, and
> >> a read-write property "Level" which gets/sets the level. This interface
> >> layout was chosen to be device agnostic, so that ANC support could
> >> potentially be added for more devices in future.
> >
> > I'd probably add Bose to the name of the interface if it is bose specific.
> While the plugin is Bose specific, I had tried to keep the interface
> itself generic so that integration with the DE or similar won't need any
> changes if ANC support is added for other devices later.
>
> --
> Peter Mullen



--
Luiz Augusto von Dentz

2020-10-15 04:27:45

by Peter Mullen

[permalink] [raw]
Subject: Re: Bose QC 35 Battery/ANC Support

Hi Luiz,

On 2020-10-14 10:15 a.m., Luiz Augusto von Dentz wrote:
> Hi Peter,
>
> On Tue, Oct 13, 2020 at 10:06 PM Peter Mullen <[email protected]> wrote:
>>
>> Hi Luiz,
>>
>> On 2020-10-12 9:43 a.m., Luiz Augusto von Dentz wrote:
>>> Hi Peter,
>>>
>>> On Sun, Oct 11, 2020 at 1:04 PM Peter Mullen <[email protected]> wrote:
>>>>
>>>> Hi all,
>>>>
>>>> I've been working on a plugin that adds support for battery level
>>>> reporting and Active Noise Cancelling (ANC) control for the Bose QC 35
>>>> headphones. The patch is nearly ready to go, but because of the
>>>> non-standard way it's implemented I figured it might be necessary to get
>>>> some preliminary feedback on it prior to submission (or to see if it's
>>>> worth submitting).
>>>>
>>>> As a brief overview of device control, the QC35 has a set of additional
>>>> controls operating over rfcomm channel 8 where messages are passed back
>>>> and forth between devices. The messages consist of a 3-byte opcode, a
>>>> 1-byte payload length, and N-bytes of payload. There's some more
>>>> information on the specifics here:
>>>> https://blog.davidventura.com.ar/reverse-engineering-the-bose-qc35-bluetooth-protocol.html
>>>>
>>>> As this is a non-standard "profile", there's no profile UUID assigned to
>>>> it which makes adding a profile somewhat more complicated. My solution
>>>> here was to add a profile under the iPod Accessory Protocol (iAP)
>>>> profile UUID which the QC35 lists, and to then filter out devices in the
>>>> profile probe callback based on manufacturer ID, device class, and
>>>> product ID. This isn't ideal as the probe will be called for any device
>>>> listing the iAP UUID, so suggestions for alternative approaches (if
>>>> necessary) are appreciated.
>>>
>>> We could in possibly add pid/vid to btd_profile that way it can be
>>> probed by pid/vid in addition to UUID, that way one can make a driver
>>> that is specific to a vendor or a product.
>> Had a go at that this evening as a proof-of-concept; so far it's working
>> well, so this could be viable. I'll finish it off and test it out properly.
>>
>>>> For session state control, I've hooked into the AVDTP state-change
>>>> callback. Again, possibly non-standard but this seemed to be the best
>>>> way to trigger the initiatiation/tearing down of the rfcomm connection.
>>>
>>> I would recommend hooking to the btd_service state, like for example
>>> the policy plugin is doing.
>> I was able to get the plugin working based on the btd_service state cb,
>> but had a couple of concerns with it, mainly because we'll be receiving
>> events for all services on all devices, so the filtering of unused
>> events adds a bit of overhead.
>> I was also able to use the AVCTP state callback without issue, which may
>> be better suited than the AVDTP one.
>
> I guess we could add support for subscribing to state changes of a
> specific service, that way you don't need to filter anything at the
> plugin, how about that?

Had a look into this, but since the state callback is being registered
during profile-probe, we're not guaranteed that the other services will
have been created yet, so we won't be able to specify which service we
want to subscribe to.
I've removed the AVDTP/AVCTP state callback hook anyway, and switched
over to the existing service-state cb; after moving things around the
overhead turned out to be pretty low, so I'll get it all tidied up and
submit a patch in a couple of days (pending any further feedback).

>>>> For ANC integration, I've added a new dbus interface under
>>>> "org.bluez.Anc1". The interface contains a read-only "Range" property,
>>>> which indicates the number of discrete values the ANC can be set to, and
>>>> a read-write property "Level" which gets/sets the level. This interface
>>>> layout was chosen to be device agnostic, so that ANC support could
>>>> potentially be added for more devices in future.
>>>
>>> I'd probably add Bose to the name of the interface if it is bose specific.
>> While the plugin is Bose specific, I had tried to keep the interface
>> itself generic so that integration with the DE or similar won't need any
>> changes if ANC support is added for other devices later.
>>
>> --
>> Peter Mullen
>
>
>