Bug: 226572369
Test: In follow-up CL
Change-Id: I3cc67c9f1466fc8518c81160fbf3034b7ae6807b

Test: NA
Change-Id: Ifffacac7a34022b6c04ceaaf332a0f73fac3ea12

Change-Id: I9e10e7ef4e56a103bfd2c916b27920487e2ab173
Test: atest BluetoothDeviceConfigChangeTrackerTest
Bug: 242348618

Bug: 237447510
Test: atest pts-bot:HFP
Change-Id: Ie16aa3ed8fa6c49afec1e8814dd228320fb9be8e

`advertise set-scan-rsp <enable|disable>` enables or disables
scan response data of active advertising sets.

Bug: 233128828
Tag: #floss
Test: build.py --target test
Test: manual test with btclient
Change-Id: I73efe1a9f506474120c97b9b2753d1d19b90fd4f

If remote device uses gatt caching, LeAudio mighd start to do searvice
search and read on non encrypted link during pairing.

Bug: 257465691
Test: manual pair multiple times with devices supporting gatt caching
Test: atest BluetoothInstrumentationTests
Tag: #feature
Change-Id: Iec70fa30453871cd95fc494d9e130ae0dbc3c5b1

This function suppose to clear scheduled devices discovery or inquire
and not scheduled service discovery

Problem can be visible when pairing coorinated set and could end up with
one device being bonded but without any services connected.

Bug: 247511422
Bug: 242645864
Test: atest BluetoothInstrumentationTests
Tag: #feature

Change-Id: I83f55a173a17544c1a736a8218b769b038b9eecc

This reverts commit 29b0d287.

Reason for revert: Change breaks multiple tests, see b/257340301

Change-Id: I4a5d7ac69b32a6dc1b782e5d4643ca4529108c22

The command, 'advertise on', start an advertising sets with default
parameters and data. This patch allows this command to be executed
more than one time to start multiple advertising sets.
A potential deadlock between btclient command handler and D-Bus
callback while stopping multiple advertising sets is also fixed
in this patch.

Bug: 233128828
Tag: #floss
Test: build.py --target test
Test: manual test with btclient
Change-Id: I68c50670e9101ce0618516a14f75e7d10e8a301a

Test: atest pts-bot
Change-Id: I086b170637ac4456e9816949d9932fc1898761ea

* changes:
  RootCanal: Remove unecessary std::move in DualModeController
  RootCanal: Misc improvements

Bug: 255848788
Test: analyze logs, try to understand what the heck bta dm is doing
Change-Id: Ibaefca94554419ffe11a2bf370654f8a2ada8f8b

 - Add StorageModule to LeScanningManager
 - Add logic for removal

Bug: 236301249
Test: system/gd/cert/run --clean
Test: system/gd/cert/run --clean --sl4a LeAdvancedScanningTest
Test: system/gd/cert/run --clean --sl4a_sl4a LeIrkScanningTest
Tag: #refactor
Change-Id: I9c8bbb7aa6ef0b19eecf33f832301bd5ca494603

- Add definition of SetEventMaskPage2 in hci_packets.pdl
- Use generated class for SetResolvablePrivateAddressTimeout
- Remove unecessary validation of LeIsoCommand views

Test: m root-canal
Change-Id: I223c9f56839a734f6b11e37a7ae3417ef9c25153

- Remove some declared but un-implemented functions
  in DualModeController
- Move suggested_max_tx_octets and suggested_max_tx_time
  to the link layer controller
- Change the type of extended inquiry response
  to std::array<uint8_t, 240>, which accurately
  reflects the length of the response as passed
  from the HCI command

Test: m root-canal
Change-Id: I8837333f9c080454d987a9b7877d8c55e6d0c324

While the previous change was focused on the data type
system used for passing audio contexts, this one takes the
code logic into the account and improves the naming scheme
for better context awareness and code readability.

It renames functions around the context type management.
The 'active context' term was ambigious and was replaced
with 'available context', which the BT spec. is using.
Functions now use more coherent naming scheme and since
some of them were returning a bit masks which were
not bit-tested but logically ANDed with other booleans,
they now return boolean type to avoid the confusion and
implicit conversion bugs to happen.

Test: atest --host bluetooth_le_audio_test bluetooth_le_audio_client_test bluetooth_test_broadcaster bluetooth_test_broadcaster_state_machine  --no-bazel-mode
Bug: 249737696
Tag: #refactor
Change-Id: I4305dc64c04bedb904d5e23ef29f5282a27a2860

This eliminates the quirky static casts and type conversions making the
code around context types a bit more readable. By introducing stronger
types we can be sure if a particular part of the API expects a single
type or multiple context types. Previously there was to many ways of
passing multiple contexts: ORed enum class values (where the ored value
was not part of the enum value), integers, list of enums, raw bitsets,
and a type alias to bitset. It turned out the integer or bitset was
sometimes actually passing a single context. There could also be a case
where one parameter is passed as something completely different due to
compatible types. The explicit constructors should prevent any unwanted
conversions, while the additionall operators allows for clearly defined
type interactions.

Bug: 249737696
Tag: #refactor
Test: atest --host bluetooth_le_audio_test bluetooth_le_audio_client_test bluetooth_test_broadcaster bluetooth_test_broadcaster_state_machine  --no-bazel-mode
Change-Id: Ib4454b8c21e7bf4a2debcc5a1c37dc1526efa66f

This splits the client_audio.cc for sink and source files
and moves them into a new subdirectory for better separation
and mantainability. It was also put into the le_audio namespace.

It simplifies class naming - now named from the perspective of
the bluetooth profile role rather than mixing the bluetooth
domain with the framework point of view. This eliminates the
non-intuitive usages like registering LeAudioClientAudioSinkReceiver
in LeAudioClientAudioSource. It could be confusing who is Sink,
who is Source and what the Receiver is. After this change, the
AudioHalClient::LeAudioSourceCallbacks instance is registered
in AudioHalClient::LeAudioSource.

The interface was cleaned up, simplified and the confusing
const void* token has been removed. The object ownership was
redefined. Sink and Source endpoint classes have a static
Acquire() methods which gives the user an ownership of a HAL
session. Since HAL allows a single client for a single direction,
the acquired HAL client instance needs to be released to allow
subsequent Acquire() calls to proceed. The Hal client life cycle
is managed by a smart pointers. Releasing it, releases the HAL
audio session and deallocates the HAL client instance. Previously
the ownership was not clearly defined, which caused situations like
le audio service allocating and releasing it in the production run,
but only releasing it in the unit test runs. But in some test cases
this object was deallocated by the test case itself. This required
special handling to avoid double free or the usage of stack allocated
objects.

Lastly, it improves logs and switches to the new logging mechanism.

Test: atest --host bluetooth_le_audio_test bluetooth_le_audio_client_test bluetooth_test_broadcaster --no-bazel-mode
Bug: 249737696
Tag: #refactor
Change-Id: I41da290d7941b4a1a3001015cb9c15a89f9c3375

Bug: 191555414
Tag: #refactor
Test: gd/cert/run

Change-Id: Ie011725544accafc9c4cf2ad2b9f4ce6b1e61bac