Test: build on linux
Change-Id: I72c7621fde1f229e3dd2a8cfa0cf0339e601914c

* changes:
  Use base::Callbacks in BleScannerInterface (2/3)
  Remove unnecessary extern declarations

The remaining four native callbacks need more internal refactoring
before they can be turned into a callbacks.

Bug: 30622771
Test: sl4a FilteringTest

Change-Id: I8e1532bdfa2b0f22545f095eb427fd5ac1a75b56

Test: manual
Change-Id: I21bdb5952fff359a9ee7091d740ecac13aad640a

This CL fixes the actual crash location on pairing within the code.
https://android-review.googlesource.com/#/c/338167/ fixed the macro.

Bug:35154933
Test: Verified by building and pairing with RPI3/NXP Arm boards.

Change-Id: Ic35f446e53aa724ecab72801b7f89c70fd11ba62

* add jni_thread_wrapper
* add FROM_HERE parameter to do_in_jni_thread, this will be used in
later patches to make thread switching logs nice.

Bug: 30622771
Test: manual
Change-Id: I4c8f9e0f32c501a6126fe04f2476ab6453fe637e

If the HAL can be opened, but initialization fails, Bluetooth
can not start.

Test: Force an initialization failure, BT dies
Change-Id: I7dd5ed49c9e7873a666dc001e68bdc13660648ba

true -> TRUE
false -> FALSE

Test: builds
Change-Id: I2573b6ab81aa384cc6d20bb8beb84303aba06a66

Recent changes in libbionic cause pthread_join to SIGABRT if called with
an invalid thread id. The select_thread_id variable in the HDP
sub-system was truncated from long->int causing pthread_join to not find
the correct thread and triggering the error.

Using the correct type for select_thread_id fixes this problem.

Change-Id: Id332fa86cf7f30abd211588d32899d41e157b462
Fixes: 35432169
Test: manual; boot device, turn BT off; ensure it stays off

* changes:
  Use LE Extended Create Connection when possible
  Add LE Extended Create Connection HCI command

After switching to HIDL, capture() can be called by multiple threads.
Capturing a packet and writing to the btsnoop/btsnooz logs
require multiple operations, hence we need a mutex to make those
operations atomic.
Previously, there were occasional btsnoop_hci.log files
with corrupted data as reported by Wireshark.

Test: Manual - captured multiple btsnoops_hci.log files.
Change-Id: Ib89d9156ad7e27a7dee652f71a0cf38c85d4447c

Bug: 30622771
Test: manual
Change-Id: Idcbbc7eb6f9b32b2f258d6f8913ead6986510da8

Bug: 30622771
Test: manual
Change-Id: I3dc48800bd2dfb2f00c824c89888c9f4e8271497

This patch fixes overflow of position variable, and possible read
outside of vector boundaries when parsing advertise data.

Parsing 1004 bytes of hex "112233112233.." was causing infinite loop.

Bug: 30622771
Test: manual
Change-Id: I0d669f7958de73f5d53350fb293ff27fea172f44

Usecase
1) Pair and connect with Apple Magic Mouse
2) Scroll over HID mouse and see if pointer moves up or down
   on DUT.

Observation:
Scroll functionality does not work on DUT and mouse pointer
stays still.

Root Cause:
From kernel 3.18 onwards, UHID flags are updated and following new
flags are added. Support for handling same is missing in user
space hid driver (bta_hh_co.cc)
UHID_GET_REPORT
UHID_GET_REPORT_REPLY
UHID_SET_REPORT
UHID_SET_REPORT_REPLY

Fix:
Add support for set report and get report in HID user
space because of changes in uhid kernel driver with kernel changes
for kernel 3.18. Apple Magic Mouse scroll functionality does not
work without this fix, as Apple Mouse requires set report from
HID host for it to start working properly wrt Scroll functionality.

Bug: 35306202
Change-Id: I90230b76d68e9275dcacd9f46f03e312ddf2fa30

Apparently, the p_data argument could be NULL if the reconfig
was triggered by the local device.

Test: Tested with Bluetooth headsets
Bug: 35351216
Change-Id: Ic51d849232944b5978772d222c41f66824e14dd4

Bluetooth stack crashes with a SIGBUS memory unaligned error since the
toolchain optimizes memcpy to generate register instructions which are
unaligned in memory. Similar issue was seen earlier in
https://android-review.googlesource.com/#/c/181713/
which was solved by casting pointers to (void *). That fix misses edge
cases where the destination pointer needs to be casted as well.

Change-Id: I23cc16142deaf3dc1fff91588d736abdfcb3bfde
Fixes: 35154933
Test: Verified by building and pairing with RPI3/NXP Arm boards.

Previously, the MTU was saved inside the AVDTP module after the first
codec setup. However, the MTU wasn't updated after codec reconfiguration.
As a result, if a Sink device uses different (smaller) MTU for
a codec like SBC, then switching fron another codec to SBC will
result in generating SBC frames that are too large to transmit,
and there will be no audio.

Bug: b/35351216
Test: A2DP streaming and codec switch to various Sink devices.
Change-Id: Ieb3caf4ef7a30a7af31ceb2f6132807739829441

Test: builds
Change-Id: I691a40ad37b0cbe3ce56a09dc986df8857883b23

The current default of 20 seconds is too high especially considering
that BLE tries to optimize (re)connection times. A connection
supervision timeout of 20 seconds essentially means that an initiator
of the connection could be in the connecting state for 20 seconds
before giving up on that connection attempt.

Checking other platforms such as iOS for default connection supervision
timeout reveals that they use sub-second (750ms) values.

Change-Id: I6ff7a20ceab0df8986191dd73676cadbe23a175e

Chained advertisement data can be up to 1650 bytes.

Test: manual
Bug: 30622771
Change-Id: I67af5a8130ac2d3f0dbd8f5e3f28b72b152256c2

Some CHECK macros had logs that printed the function name, which
duplicated the logs in the functions outside the checks, producing too
many logs.

Remove some unused macros.

Changes:
02-01 09:34:11.757  1259  1755 D bt_btif : get_folder_items_list_rsp
02-01 09:34:11.757  1259  1755 D bt_btif : get_folder_items_list_rsp: item_cnt: 0 len: 0
to
02-01 09:34:11.757  1259  1755 D bt_btif : get_folder_items_list_rsp: item_cnt: 0 len: 0

and similar.

Test: run logcat, check that we have less duplicated logs
Change-Id: Id6e57f38b34d176ed8be0c3baedfbcf55e61a9a4

Based on AVDTP Spec 1.3, Section 9.8 StreamChangeParameters, the RECONFIG
procedure (AVDTP_RECONFIGURE_CMD) happens in OPEN state, and there is no
state transition that would prohibit calling the RECONFIG procedure
back-to-back.

Test: A2DP codec reconfiguration and audio play with several Sink devices
Bug: 35284763
Change-Id: Ie4ddca8545e2146fbd863e55713fd2d49994b4b0

Bug: 30622771
Test: sl4a FilteringTest
Change-Id: I94e62ca0d785b20f97e3397131b70eac06e15003

scanning_phys is a bitmap, not a count of PHYs. This error was resulting
in invalid HCI command being sent. This had no impact so far, because
only 1M PHY was used.

Bug: 30622771
Test: manual
Change-Id: I2be4f5ddbe0d453dc49cdee4605821b10853d7d8

Test: Build
Change-Id: I378883fdeb09fadaac04ace2dc499abd27b7d3ab

- Include <mutex> for heart_rate_server
- Remove unnecessary assignment

Test: Build
Change-Id: Ic2368d08789b95e71c083460148c70b9d0ef0b20

* use non-hardcoded advertise data size
* use cache for keeping non-complete advertising data, waiting either
for scan scan response, or secondary channel data.

Bug: 30622771
Test: sl4a FilteringTest
Change-Id: I689edcb4aee30361df15340029559cd19dac552b

Test: manual
Change-Id: I1ff4ad6b2543246351d4d34919ba443e5617a9c0

* Logcat uses gettimeofday for its timestamp, the same as on-disk BT
  Snoop log
* Although in-memory BT Snooz Log uses the same method to get time, it
  is calling it separately, resulting in mismatch between timestamps of
  two snoop logs
* This CL let them uses the same timestamp_us value and put the function
  definition to libosi
* Note that preserved on-disk BT Snoop logs timestamp postfix at
      btsnoop_hci_<timestamp>.log
  will be changed to microsecond since epoch at current device timezone
  instead of the shifted BT Snoop timestamp value
* New unit tests for gettimeofday

Bug: 35113514
Test: Make, unit tests, run BT activities and check both snoop logs
Change-Id: I5b3f87bc523b272ced2c69a4595d0e0cbe29bcb3