GitLab

* commit '654c376f':
  DO NOT MERGE Reduce TIMER_INTERVAL_FOR_WAKELOCK_IN_MS to 100ms

The bluedroid wakelock locked time has increased 2.5x since K.
Reduce the 3s interval to reduce the time that the wakelock is
held, allowing the kernel to suspend more frequently. By
reducing this, wakelock usage returns to roughly K levels.

This in turn reduces BT process power usage.

Bug: 18357273
Change-Id: I7c3bddffafb0fa59468b8dd5718220be8ac85506

Bug:18045480
Change-Id: Ic433f4aa26bd08e84c53c447be9d5278e16ebd55

In an attempt to make the I/O capability configuration more reasable and
overwritable per platform, the BR/EDR specific I/O capabilities were
used for LE. In SMP, the I/O capabilities flag is mapped differently,
thus a separate flag is required to configure the LE I/O capabilities.

Bug: 18245986
Change-Id: I6b4ba8f8a1d60ea25729e49d305e3653e94928c6

* commit 'a59f47f2':
  DO NOT MERGE Remove overflow rate limiting when reading A2DP frames

DUMO device can do both BLE and BREDR operations. It was wrong
to assign BLE or BREDR type to it.
Bug: 18275230

Change-Id: I8d6c17e5157ba29a1af459629de9e2d6f6dc5e99

A2DP now supports adjusting the number of A2DP frames sent per timer
tick to adjust for timer drift and missed media task ticks. When the
signal to the headset/speakers becomes interrupted, the number of
packets to be read can queue up and rate limiting will not allow for the
queue to be cleared.

The overflow mechanism introduced in commit 4aebca46 will cause the media
task to stop sending packets completely, which can underflow the jutter
buffer on the remote device and lead to audio drop-outs.

This patch removes the overflow mechanism and also adds code do discard
audio frames that could not be sent to the remote device (weak signal
etc) to allow the device to stay in sync and not build up audio delays.

Also added additional debug logging and changed the UIPC flush mechanism
to address an issue where reading byte by byte causes an endless flush
loop if remote UIPC producer writes data faster than the flush loop
consumes it.

Bug: 18244205
Change-Id: I9a424984806bb2a464877399804b3355b2c439c3
(cherry picked from commit c5916e97)

A2DP now supports adjusting the number of A2DP frames sent per timer
tick to adjust for timer drift and missed media task ticks. When the
signal to the headset/speakers becomes interrupted, the number of
packets to be read can queue up and rate limiting will not allow for the
queue to be cleared.

The overflow mechanism introduced in commit 4aebca46 will cause the media
task to stop sending packets completely, which can underflow the jutter
buffer on the remote device and lead to audio drop-outs.

This patch removes the overflow mechanism and also adds code do discard
audio frames that could not be sent to the remote device (weak signal
etc) to allow the device to stay in sync and not build up audio delays.

Also added additional debug logging and changed the UIPC flush mechanism
to address an issue where reading byte by byte causes an endless flush
loop if remote UIPC producer writes data faster than the flush loop
consumes it.

Bug: 18326405
Change-Id: I9a424984806bb2a464877399804b3355b2c439c3

When HOGP keyboard is connecting and BT off is requested, if there
is any HOGP data that was received during this state, it causes the
following side-effects:
- LE data was not thrown away in disconnecting state, instead it is
now being used as indication of link establishement
-The above triggers the link open callback for fix channel and
causes the BT off failure.

Check L2CAP link state and always discard the fixed channel data
received in disconnecting state solves the problem.

Bug: 17793437
Change-Id: I633d5e6346e952daa2abd642effad2f260954fdf

* commit 'b407515f':
  Fix "no instances found for adv" when stop BLE scanning.

Bug:17703806
Change-Id: I103d3637ad38d33b4370850a300b0d72bdaeb562

* commit '2933de99':
  HOGP: Do not remove device from whitelist on encryption failure

* commit 'f37767a1':
  HOGP: Do not remove device from whitelist on encryption failure

If the encryption times out during the re-connect process for a HOGP
device, the device is removed from the whitelist and thus will not
reconnect until it is added back to the list. On platforms that do not
have a way to reconnect to a device manually, this means the device will
never reconnect again until the stack is restarted.

This patch checks the reason for the encryption failure and does not
remove the device from the whitelist if the encryption failed to a non
authentication/key related error.

Bug: 18233442
Change-Id: Ifaec4a9224ff9281956715b38d34c2d7c3fecb01

am e50c7933: DO NOT MERGE Do not set sig_id in gerneral reject message per AVDTP SPEC 1.2 automerge: 595bcb4c  -s ours automerge: 23699356  -s ours
automerge: 0570fd9c  -s ours

* commit '0570fd9c':

am 87303c76: Increment seq before adding to packet in avdt_scb_hdl_write_req_no_frag automerge: 98ad505f automerge: f81ffdb1
automerge: 6d11a01a

* commit '6d11a01a':

am e50c7933: DO NOT MERGE Do not set sig_id in gerneral reject message per AVDTP SPEC 1.2 automerge: 595bcb4c  -s ours automerge: 23699356  -s ours

* commit 'e50c7933':

am 87303c76: Increment seq before adding to packet in avdt_scb_hdl_write_req_no_frag automerge: 98ad505f automerge: f81ffdb1

* commit '87303c76':

DO NOT MERGE Do not set sig_id in gerneral reject message per AVDTP SPEC 1.2 automerge: 595bcb4c  -s ours
automerge: 23699356  -s ours

* commit '23699356':

automerge: f81ffdb1

* commit 'f81ffdb1':

am 020cccca: am 5fe98634: am ba2f6d99: Revert "anthias:17614185:Bluetooth: Getbuf returning invalid buffer pointer" automerge: 8403c6d2

* commit '020cccca':

am 5fe98634: am ba2f6d99: Revert "anthias:17614185:Bluetooth: Getbuf returning invalid buffer pointer" automerge: 8403c6d2

* commit '5fe98634':

automerge: 8403c6d2

* commit '8403c6d2':

am 36eac517: am 27c93b25: am 46bfe63d: DO NOT MERGE AVRCP: Proper handling of Fragmented AVCTP packet automerge: f9336d34  -s ours

* commit '36eac517':

am 9624c4d6: am 36ba410e: am c1ccb0d9: anthias:17614185:Bluetooth: Getbuf returning invalid buffer pointer automerge: 0e74efe5

* commit '9624c4d6':

am 27c93b25: am 46bfe63d: DO NOT MERGE AVRCP: Proper handling of Fragmented AVCTP packet automerge: f9336d34  -s ours

* commit '27c93b25':

am 36ba410e: am c1ccb0d9: anthias:17614185:Bluetooth: Getbuf returning invalid buffer pointer automerge: 0e74efe5

* commit '36ba410e':

automerge: f9336d34  -s ours

* commit 'f9336d34':

automerge: 0e74efe5

* commit '0e74efe5':

Bug: 17916028.
Change-Id: I4badb9208eeb973a08b4f8463ee287f6bfb3dd01

* commit '903285cb':
  Don't starve a BT ACL link with 0 xmit quota

* commit '472c2a7b':
  Don't starve a BT ACL link with 0 xmit quota

automerge: 595bcb4c  -s ours

* commit '595bcb4c':
  DO NOT MERGE Do not set sig_id in gerneral reject message per AVDTP SPEC 1.2

automerge: 98ad505f

* commit '98ad505f':
  Increment seq before adding to packet in avdt_scb_hdl_write_req_no_frag

On some platforms clock_gettime may sometimes return a current
time that is earlier than a previous time. While it rarely
happens, we need cover the case anyway.

Bug: 18154020
Change-Id: I3846487aa45d0b11aeb63fa68af8826b864248fa

Bug: 17753401
Change-Id: I288ea8b6194d91d18485d0f3f1bbf997f97a26dd

With new link management of controller buffers between
different ACL connections, it is possible that a ACL
conn would get 0 buffers - thus no profile can run.
This affects classic and LE ACL connections.

This work around allows a floor of 1 buffer as the
floor for all connection, irrespective of controller
capability.

Bug: 17976311
Change-Id: I7cd0d34af47c441c5ea7f2b836b591a4fec62ad1

* commit '5e808468':
  LE: Do not remove bond on encryption failure