This patch implements the initialization of Bluetooth 5 advertising.

Bug: 30622771
Test: manual
Change-Id: If06a6c827f1d8097aa3f019b99ebcb3933742d6c

Add ability to check maximum advertising data length.

Bug: 30622771
Test: manual
Change-Id: I0f3c806046157633f8e2106c1b2700c4277a2b4d

Bug: 36372989
Test: Running ASAN build
Change-Id: I6016e7609ba24db632222ff4613017e3be30b09c

This fixes stack-buffer-overflow issue found using ASAN.
Previously, the original data had type "struct tBTA_PAN_SET_ROLE"
and similar, and eventually "memcpy(..., sizeof(tBTA_PAN))" would
copy data beyond the end of the data buffer.

Bug: 36367964
Test: Running ASAN build
Change-Id: I47210a501378023168a0dd71381e93a5051a4c71

This patch wires up fist methods of AdvertisingSet, making it possible
to start advertising and stop advertising. It also replaces legacy
implemementation with calls to new implementation.

Bug: 30622771
Test: sl4a ConcurrentBleAdvertisingTest
Change-Id: I843d8fc12ba1928c87e256a0fb6d730c9fa200e6

Change-Id: Ic28aa4a120d973bbf3f05d8022bc2c2e619562e2

There is a race condition when calling
event_command_ready() -> transmit_fragment() -> hci_transmit()

If right after hci_transmit() there is thread context switch
and another thread executes filter_incoming_event() for the same
command, the corresponding packet/command will be taken off the
commands_pending_response list and free()-ed.

However, after the execution on the first thread continues
within transmit_fragment(), the execution logic will continue using
the "packet" that was already free()-ed by the other thread.

To prevent this from happening, the "commands_pending_response_mutex"
within event_command_ready() has to protect the transmit_fragment()
execution and the update_command_response_timer() function right after it.

Also:
 * Changed the "commands_pending_response_mutex" to recursive_mutex
 * Added "commands_pending_response_mutex" protection in few other
   places where "commands_pending_response" itself is used.

Bug: 36205494
Test: Running ASAN build
Change-Id: I63677ad1f2b28683c321631e9e29e4f01628d269

Test: net_test_bluetooth BluetoothTest.AdapterRepeatedEnableDisable
Change-Id: I72f5c9105b535ad5a5c41bc15dc9a3ad5f4e023f

When BTM_LOCAL_IO_CAPS == BTM_IO_CAP_NONE the code fails to compile due
to a missing comma. This was introducted a couple of months ago by
70556f32 but it went undetected.

This patch simply adds back the missing comma.

Bug: 35804180
Test: `make` on a device with BTM_LOCAL_IO_CAPS set to BTM_IO_CAP_NONE.
Change-Id: I0a924abec4fe6612c5dbb72416123ca391551523

Test: manual
Bug: 30622771
Change-Id: I286b8d636115385ee5951284629a1894026d9b5b

Bug: 36043456
Test: Code compilation
Change-Id: Ib588c95a64100f5b2b3567f1bac9fe5e9160e4c0

Bug: 30622771
Test: manual
Change-Id: Ie0df73475359c69af8df672454fe0c0374523c88

Bug:36026072
Test: Switch users (toggle Bluetooth under load)
Change-Id: I3fb901b5c47021c708c1fb548858406225065d00

Bug: 30622771
Test: manual
Change-Id: I358b872c9ffb52f007db4f111520c05b9521aa7f

This prevents errors in case of event callbacks being received or timers
being fired after the pending command queue has been deleted.

Bug: 36026072
Test: Turn Bluetooth off/on while scanning and file transfer
Change-Id: I7bb699e929a8b40e42a3543b5445885ae6b1b975

Wire the new feature check API to actual values received from the
controller.

Test: manual
Bug: 30622771
Change-Id: I2eefb7d5bb00d89ac7ea803191ba05f830080ade

Test: manual
Change-Id: Ia38604d496b8ae50934c42b68f2424afd1c6320c

Test: Bluetooth starts/stops, Bluetooth scan works
Change-Id: I4b04d71d57f45956898ff1a99e495195b54f4e8e

* BTM_ReadRemoteDeviceName checks if a device is connectable over LE before
  making a connection. However, when it is not connectable over LE, it could
  be connectable over BR/EDR.
* For Dual Mode device, the device type should be 0x3 and is not equal to
  BT_DEVICE_TYPE_BREDR, hence we cannot check equality to determine whether
  a device is BR/EDR, but need to AND them and check the result.
* btm_ble_read_remote_name() already checks if device is connectable in LE,
  and we do not need to do it in BTM_ReadRemoteDeviceName().
* This fix removes the check from BTM_ReadRemoteDeviceName
* This fix also adds error checking to return values of BTM_ReadRemoteDeviceName
* This fix moves inquiry database query to respective remote name
  request procedures, as the result of the query is used in those
  procedures instead of in BTM_ReadRemoteDeviceName()

Bug: 35348392
Test: make, pair with device that advertise on both LE and BR/EDR
Change-Id: I10d1aa9dd6e2447f9b9d40b1fea9370b5088a57d

Change-Id: I9c7dbd8dfdb440a31e4c98e93ee533fe96ee7b0e

Using BTM_IO_CAP_NONE currently causes compilation to fail since
bta_dm_sp_cback ends up unused out a previous #ifdef that only kicks
in if BTM_IO_CAP_NONE is not set. Plugging the entire function into
a #ifdef so consistency is maintained and the framework can compile.

Bug: 35257446
Test: Built locally and run on device. Pairing without display using
JustWorks works.

Change-Id: Ibd7af6454253ccd731ff79602f327a52b8fe585b

Some devices claim they can handle large MTU and large SBC bitpool.
However, they are struggling and eventually over time they
start delaying the "Rcvd Number of Completed Packets" reports.
Now we always limit the payload MTU to MAX_2MBPS_AVDTP_MTU(663) -
with the overhead of (4 bytes L2CAP Header + 12 bytes AVDTP Header)
such packet can fit within 2-DH5 payload size of 679 bytes.

Bug: 34127145
Test: A2DP SBC streaming to headsets and carkit
Change-Id: I2f32becba703d77d36be46158f68b8a9ef3f299a

Some Sink devices might advertise AAC zero bitrate or a very small
value (e.g., 24576). Ignore values that are below a threshold
(currently set to 64000), because those seem invalid.
For such devices, the actual streaming bitrate will be computed
based on the MTU.

Test: Code compilation
Change-Id: Ibe3de6e53681a618f388b41a01f366ad462b7a1d

Bug: 35690239
Test: Switch users with Bluetooth enabled
     (Disables/Enables Bluetooth under high load)
Change-Id: I6e09547cc8ffa3f01bc523ce4c18fdabfe08df99

Some assigns snuck into our conditionals, making the code harder to
read / follow.

Also remove some unnecessary null checks after osi_malloc that are
nearby.

Test: compiles, sanity check with device
Change-Id: Ia544da44060d9fc4363af9fbcb4112f3190f75ab

Test: Switch users with Bluetooth enabled
     (Disables/Enables Bluetooth under high load)
Change-Id: I2414fd26a0380644700d66a8fdec1fc54a1cd2f6

Test: Switch users with Bluetooth enabled
     (Disables/Enables Bluetooth under high load)
Change-Id: I9039f97c090f693e798b8082fbb027e2028c8b40

Test: Switch users with Bluetooth enabled
     (Disables/Enables Bluetooth under high load)
Change-Id: I34c5579b75620970f756fff28ccc07cf76599a94

Use Case: DUT unable to pair with Xbox HID device

Steps:
1. Turn on BT
2. Start Inquiry
3. Pair with device with three extended feature pages

Failure: Unable to pair with HID device

Root Cause: DUT not responding for btm_read_remote_ext_features_complete
as we are not supporting three extended pages. So Security procedure also
not getting started.

Fix: Increase max remote ext feature pages to three from
2 as some latest remotes supporting three extended pages.
As per spec we can support 255 extended pages.

Test: Pair with device with three extended feature pages

Bug: 32999575

Change-Id: I061c78a7aaa1ab58a2fc554d026305a0124eee77

In build/BUILD.gn we set EXPORT_SYMBOL value, and it's used across the
whole stack. However, third party libraries require this value to be
cleared - otherwise the symbols are not exported properly.

Test: manual

Set the address during SetParameters, for VSC implementation, and right
after setting parameters for BT5 implementation.

Test: sl4a ConcurrentBleAdvertisingTest
Bug: 33922058
Bug: 34962013
Change-Id: Idcc60a36884607de191066d8be1ee3bc459e86fc

Make btsnd_hcic_ble_rand accept base::Callback instead of function
pointer. This will result in great code simplifications in further
patches.

Test: sl4a ConcurrentBleAdvertisingTest
Change-Id: I9933caa78b20d0011ce3e17fd72d5740bcc4ae6d