GitLab

Required now that we always return success.

Bug: 30025777
Change-Id: I7db51321672c128039545cc0fb604c615f169bc3
(cherry picked from commit f4a37b25)
(cherry picked from commit 5274ab7a)

Fixes: 30004157
Change-Id: Id170856fbee5ea345f02f55d3a6e3376d846c2de
(cherry picked from commit e8b81851c33095fe3b3ac78329073309be5421e1)
(cherry picked from commit d2a37eff)

Bug: 29157087
Change-Id: If227963b939a9b0ed55e6d900364c71932d3cfe8

Currently, our TX queue does not allow for quick recovery of minor
congestion. As we only clear as much room in the queue as is needed to
fix the news frames needed for a given timer tick, any minor congestion
on the TX path can result in single frames to be dropped over an
excessively long amount of time.

From the "just rip the band-aid off" school of thought, drop the whole
queue instead to cause a single drop-out on congestion and catch up to
the live stream immediately, hoping that the stream recovers...

Bug: 29601962
Change-Id: I482c3cfef6d8b51143e54733eb2da5f70663636f

Bug: 29601962
Change-Id: I30ce09f1c7550a1be9fd1c8ed70745f9aa0cdb44

Bug: 29601962
Change-Id: I4a86b786609b0203a72b750a1c4a38c82fcc5f97

am: edb7ce12

Change-Id: Ie7ea41474e11479b54ba526294977dfaa60c8cd7

am: 8e80a260

Change-Id: I54a690024ccd316058d9d59ff7b16206f7fcd543

am: f15ca6d2  -s ours

Change-Id: Ic1a23e2c6d0284b9fdf346a681f59cdcbaa599e0

Change-Id: I7278875d21c59316217cb1b99d902f4fc248ca06

Bug: 28885210
Change-Id: I63866d894bfca47464d6e42e3fb0357c4f94d360

- Handled Buffer over flow for uint16_t
- Discard invalid HCI packets from Codenomican test
  tool as data length and actual data not matching
  during reassembly

Use case:
Execute L2CAP test suit from Defensics Codenomican

Steps:
1. Pair and connect DUT to Codenomican tool
2. Execute L2CAP test suit from Defensics Codenomican

Failure:
Crash observed on DUT and Codenomican tool stuck in execution.

Root cause:
Codenomican tool sending invalid HCI packets to DUT and
there are no checks to handle buffer over flow and other invalid data
from Codenomican tool.

Bug: 29498064
Change-Id: I6f93c80244fc39d607ad285185136bbbca83d7ae

Each pointer that was freed by one of the following functions should
be reset to NULL:
 - config_free()
 - data_dispatcher_free()
 - fixed_queue_free()
 - list_free()

Bug: 29421693
Change-Id: Ie55a04ed986393994564edcf872d7413b0767a85
(cherry picked from commit 8cbc2910)

By polling, we ensure the socket buffer is nearly always full,
as blocking socket writes may not wake in time to avoid underflow.

Bug: 28286313
Change-Id: I056768dbd860993c41d44c33a4846d62c50db933

Some HID devices will accept both write request and write command.
Prefer write commands when possible.

This is due to bug in some hardware, that will not send write response
when too many commands are queued.

Bug: 29184976
Change-Id: I5bafc067f044a48f152c60b8e17722067bc1453e

Id field is equal to attribute handle, so it must be 16bits.
Otherwise only piece of GATT database can be used.

Bug: 29253825
Change-Id: Ia88e3ad3eb96ca8e97349aa5e852e66c57072b49

GATT Service is contained between start and end handle. If
characteristic definition is at end handle, and it's value definition is
after end handle, it will not be properly discovered. That's because we
use value_handle instead of attribute_handle to identify
characteristics.

As a workaround, increase service boundary if value is defined after
it's definition.

Bug: 29253825
Change-Id: Ib145aea4f5cf38a1fbb977c301136e16f8f900f7

If there are any HID operations scheduled after disconnect, mark the
device as no longer executing after establishing connection. Otherwise
we'll be unable to send any HID commands to it.

Change-Id: Ie1bb622388a3bcb574dec3dde76ef4518ee1102f

If there are any GATT operations interrupted by a disconnect, mark the
device as no longer executing. Otherwise we'll stil receive data from
the device, but we'll be unable to send any HID commands to it.

Bug: 29184976
Change-Id: I489f41c970abad0ff1145005d3250beafa83d27e

Also fixed AVRC_PDU_GET_PLAY_STATUS timeout response.

Bug: 28177785
Change-Id: Icde64a5c1806453850abe101f9707fff61566566
(cherry picked from commit 1c512b85)

Use Case: Disable time out when atleast one active ACL connections

Steps:
1. Turn ON Bluetooth
2. Connect to any Headset
3. Turn OFF bluetooth.

Failure: Turn OFF bluetooth is failed when atleast one active ACL connections

Root Cause:i If we have one or more active ACL links, so it is not notitying
BTA_DM_DISABLE_EVT to upper layer.

Fix: Set disable timer in case of any active ACL connection

Bug: 29217385

Change-Id: Ica47717119f1ab9fa9f4f5244e8483c408276a2a
(cherry picked from commit 3e4f01ad)

Payload size for GATT Read by type request for UUID32 and UUID128
are incorrectly allocated. This leads to memory overflow when
Read by type request is sent for them and eventually results
in crash during free. This change makes sure that the payload
can accomodate upto 128 bit UUID.

This crash is observed while running TC_GAR_CL_BV_03_C Qual test
case.

Bug: 29011042
Change-Id: Ib2b41b769b394670099f4549f204e2972f7df876
(cherry picked from commit 7da98621)

Steps:
Execute TC_LE_REJ_BI_02_C

Failure:
PTS stuck after sending INFO RSP, ECHO RSP on the LE signaling channel

Root Cause:
The Bluetooth stack is not rejecting INFO RSP, ECHO RSP even though
these commands are invalid for the LE signaling Channel.

Fix:
Reject the INFO RSP, ECHO RSP with error L2CAP_CMD_REJ_NOT_UNDERSTOOD.

Bug: 27852645
Change-Id: Ie09f658d606206cbdbea928e8cb6d10332743cf4
(cherry picked from commit d35bb40d)

The spec only allows one result per event.
Abort processing and return if the spec is not followed.

Bug: 29141745
Change-Id: Ic414b28adadec274779b0d953d12d499a619aff6
Signed-off-by: Myles Watson <mylesgw@google.com>