Also added bonded device management because this is needed for A2DP to
work with most devices.

Bug: 32657860
Test: Run on device, modified tests pass.
Change-Id: I862e41b6594b7cec9f61094d300577574cc8bbc6

We cannot simply assume the write is terminated in reliable write. When
the reliable write value is longer than MTU allows, the current
implementation can only send whatever MTU allows and naively set the
status to GATT_SUCCESS, in the name of "application should verify handle
offset and value are matched or not". That's why MTU negotiation is a
workaround as people mention in b/37031096, which just fits all the write
value into a single request.

This also blocks our test on CtsVerifier.

Bug: 37031096
Test: Manual test and confirm that we don't simply send partial value
Change-Id: I907877608f4672f24c002e630e58bf9133937a5e

Usecase:
1) Pair and connect with remote hid device.
2) Disconnect with remote hid device.
3) Unpair DUT from remote hid device.
4) Send connect request from hid device by pressing any button.
5) Try to pair and connect DUT with remote hid device.

Expected Result
1) Connection should fail after step 4 if remote device does not initaite pairing.
2) Connection should succeed with remote device and hid device should be added to
paired list and hid operations should work properly.

Observed Result
HID operations from remote device starts working without device added to paired list
and UI stucked at pairing.

Root Cause:
After step 4, device should be removed from database when incoming connection is
rejected.

Fix:
Perform virtual unplug for device when connection request for hid is received from
unpaired device.

Test: Issue is not reproducible after stress testing with above change

Bug: 68176469
Change-Id: I9bc72ccc7a849c0d6acf3230a75e1fe95949b3fe

Bug: 112285337
Test: Manual
Change-Id: Ibb019d11cf93051a198df4efb7bd79f31c15a9c0

Some inactive devices can't send hung up key to host when
HFP connected and active call exist.
And other inactive devices can send hung up call by
sending hung up command.

Solution: To keep consistency of this behavior, 
only allow active BT device to hung up call.

Test: Active device can hung up call and inactive can't hung up

Change-Id: Iec1ae689f1d40c15199707e9e925814f71ded619

event_finish_startup and startup_timer_expired can race, leading one of
them to try to use startup_future when it is NULL.

Check startup_future still exists before using it.

Bug: 112114060
Test: Use Bluetooth
Change-Id: I018d45f070b3f931c1114abf0471b6ea1d524f2a

Bug: 112451773
Test: compiled and local device test

Change-Id: Ia39f46b3b9667cba34465e3ff310df85a765c3e1

bug: 112350218
Test: manual
Change-Id: I9e4297b936e7a7db164e7ff3bf680c43600c5bda

Right now we often get this error in logs:

E bt_btif_gattc: btif_gattc_upstreams_evt: Unhandled event (7)!

This is often misinterpreted during log analysis as error, where this
really means just uninteresting event we don't want to handle.

Change-Id: Ia7804b2e1802a117198a31e025d05a80f973e896

* changes:
  Reduce Media State change callbacks
  AVRCP Controller metadata feature

Eliminate the 2 second timer and only update media state when it
changes.

Bug: 28917369
Test: Play music, pause and change its position observe status is
correct.

Change-Id: Iaf17e8cde570651706c55603b01c5100d4c60aa4

Bug: http://b/112277682
Bug: http://b/79419131

Lld, which is the default linker for the platform, is able to resolve
symbols across libraries irrespective of the order they appear in the
command line.  This was not the case for ld.gold and other linkers.
Unfortunately, there are some targets that still rely on ld.gold
(pgo-taimen-config1 in internal release branch being one of
them).

With http://aosp/721652, libbt-stack should be added before libbt-common
so MessageLoopThread::~MessageLoopThread is resolved correctly.  If not,
the following error is reported by gold:

system/bt/stack/btu/btu_init.cc:41: error: undefined reference to
'bluetooth::common::MessageLoopThread::MessageLoopThread(std::__1::basic_string<char,
std::__1::char_traits<char>, std::__1::allocator<char> > const&)'

Test: m USE_CLANG_LLD=false
Change-Id: I9c8208218ffb4bb027ad75b51bcd484e9bb37368

Process the METADATA feature flag separately from the VENDOR feature
flag to ensure it doesn't get overwritten.

Bug: 110039732
Test: connect avrcp from phone verify getCapabilities events supported
is sent.

Change-Id: Ib5d54c13d6c912904f20e8760f1632fa5737b98f

Bug: 112104996
Change-Id: I242b3c68bcb25f29f8e979487f507a0585f1a8ae

am: b7e311c7

Change-Id: I859ce7751ee787fcbf9b5e46f07690a28fa91a1f

Bug: 112135987
Test: compilation
Change-Id: Ifaf0189f512ef2abdc7179bda5d6ff1e291fde54

am: 628a65f0

Change-Id: I1614d74f1f6980623ef38f00be4a1024a7f8e306

* host_supported is broken in cc_benchmark on Mac

Bug: 77585931
Test: make
Change-Id: I2d2b4d52a108f90e3ce2d9c8c9f5510918f1fcff

am: 63d2fa3e

Change-Id: I1093576b83d19cac3a11ff4d81173d97604d2a4d

am: 0ac5b1b6

Change-Id: I040d328e642b351420f9a1f4eb35e6c0dc09a74e

* Replace OSI thread with MessageLoopThread in
  - btif_a2dp_source (A2DP encoding thread)
  - btif_core (JNI thread)
  - btu_init (BTA thread)
  - hci_layer (HCI thread)
  - module (module bring up thread)
  - various unit test threads
* Code that uses reactor from OSI thread is not replaced in this change
  because they need to move from using reactor to message loop first
* Main difference:
  - HCI layer no longer has access to mutex protecting message loop
    set up and tear down. Messages posted to thread after ShutDown()
    is called become no-op
  - In all cases QuitClosure() is used instead of QuitWhenIdleClosure().
    This means that we will never force kill a thread. An infinite loop
    will delay the shutdown of Bluetooth stack.
* do_in_hci_thread is actually do_in_bta_thread
* Move the definition of do_in_bta_thread from bta to btu since btu is
  where the thread actually lives
* Remove bta_closure_api.h and replace it with stack/include/btu.h
* Remove header inclusion in bta_sys.h and include stack/include/btu.h
  in individual compilation units
* Fixed a bug in btif_cleanup_bluetooth where btif_jni_disassociate()
  was not called on the JNI thread
* Make setting real time priority a requirement
* Crash during the following scenario:
  - When btif_profile_queue cleanup failed to be scheduled on JNI thread
  - When A2DP encoder thread failed to gain real time priority
  - When BTA thread failed to start up
  - When BTA thread failed to gain real time priority
  - When BTA thread failed to schedule its first callback
* Turn off Bluetooth in the following scenario:
  - When HCI thread failed to gain real time priority
  - When bt_workqueue thread failed to gain real time priority

Bug: 110303473
Test: make, native and Java unit tests,
      Connect to headset and listen to music through A2DP,
      testplans/details/184455/3975
Change-Id: Ib448992fc0ba5af82c2d117dd65f1abd45d2acb6

* Add MessageLoopThread to abstract thread implementation to our code
* Add ExecutionBarrier utility class to help with thread synchronization
* Add more performance tests and performance benchmarks to evaluate
  execution efficiency in both batch and sequential usages, for various
  thread implementations including:
  - libosi reactor on pthread
  - MessageLoop on libosi reactor on pthread
  - MessageLoop on STL std::thread
  - MessageLoop on Posix pthread
  - MessageLoop on libchrome base::Thread
  - MessageLoop on MessageLoopThread

Bug: 110303473
Test: make, native and Java unit tests,
      test/run_benchmarks.sh net_benchmark_thread_performance,
      test/run_unit_tests.sh net_test_performance,
      test/run_unit_tests.sh bluetooth_test_common,
      testplans/details/184455/3975
Change-Id: I5b4ce2ee910a0f1d2edf95e0296916dea04d3f89

am: 4d024528

Change-Id: Ibba4bdb0f337d743a7148d49a1ad23eb2f8a8b5e

am: b8b3655f

Change-Id: I313af6b96913b72fc140c7873c1952f24b75c12a

Use case:
Run PTS test cases TP/LE/CFC/BV-10-C to BV-16-C

Steps:
1. These test cases executed with two DUT as PTS support not available
2. Set desired security level on DUT and Remote with tool
3. Initiate LE COC connection request from DUT

Failure:
LE COC connection accepted though even remote security are not met.

Fix:
Added security handling for COC connection

Test: Ran sl4a test for BleCocTest
Change-Id: I5f6385d462698582e701b5a07953c0ccb932bbbf

Reason: When the BT device name is updated is getting saved to the
config data pointer but not to the persistent data (i.e. bt_config.conf).
So, when the reboot is happening it is not able to get the updated device
name from the persistent data (i.e. bt_config.conf) as during reboot
bt_config_flush is not called.

Fix: Saving the BT Device name to persistent data using btif_config_flush
once it is set.

Test: Manual Reboot test
Bug: 110301897
Change-Id: Ie329b475eaaf208ed667f27e271a00e230e4b95c
Signed-off-by: Aiswarya Cyriac <aiswarya.cyriac@intel.com>
Signed-off-by: Gaganpreet kaur <gaganpreetx.kaur@intel.com>

am: cc5b28bf

Change-Id: I7e432dd8dcc8e389807998e2f07d368414412dd1

am: 6252ace3

Change-Id: If9ab310626b5acb2364a2403ba831de3021b5ce2

When setting a browsed player memory for folder depth was incorrectly
allocated based on folder item count not depth.

Bug: 111370532
Test: Set a browsed player that has a depth greater than item count.
Change-Id: I6984c5852f623e7e9478429bdab3dcbcf57196e0

Reject AVDTP Codec Capabilities that contain invalid fields

Bug: 110918549
Test: Manual - connect to a device that sends invalid codec capability
Change-Id: Ib697cc9ad13ed666fbafa9219447592fe3f56684

am: 5c9ab99a

Change-Id: Ibe4f75b8eb8b50dd9aaa105bdbafe766bedb8ed4