GitLab

When adding devices to the whitelist for LE background connections, a
local variable is used as the hashmap key, preventing any successful
lookups going forward. Thus the device will repeatedly add the same
device to the whitelist and preventing successful connections going
forward.

Bug: 23423602
Change-Id: I3d2590c0a1dd66c6e1864ea53f875a713660b645

Add missing return value check when reading the HCI type byte.
This check is needed as a safeguard. For example, function
event_uart_has_bytes() could be called (indirectly)
within the run_reactor() loop not only when there are bytes to read,
but also if there is an error (e.g., EPOLLHUP | EPOLLRDHUP | EPOLLERR).

Bug: 23105107
Change-Id: Ic3b6e4d656406949e384c8106b0c607f7c221759

For security reasons, TCP sockets now listen on the loopback
IPv4 address 127.0.0.1 for incoming TCP connections.

Bug: 23272146
Change-Id: I88523f643f305f2281740575d7011b6077bf0843

AVRC response is created by copying the original received packet.
When allocating a buffer for the response, the buffer length
should be large enough to contain the response.

Bug: 22437809
Change-Id: I862d633e76d3c5221582459f19935a45e53577c7

In the document, the supported version is 96.
But, the defined supported version is 95 in the code.
The callback function btm_ble_vendor_capability_vsc_cmpl_cback
can not read number of track entries when it compare with 95
and supported version from firmware.

Bug: 22906552

Change-Id: I64e6f14f34ef3ed0ddc3fee2fad05eb03e5938f1
Signed-off-by: Jacob Lee <jacob.lee@mediatek.com>

Implemented the factory reset function to be used to reset all bluetooth
settings on device to factory default

Bug: 16161518

This change allows the host to mask the Cross key bit in
the key distribution fields of the pairing request and
response while pairing with the Dialog keyboard and mouse
to prevent the remote from rejecting the DUT's pairing request.

Bug: 22799966
Change-Id: I89420e77875957c78e915c508de311d779fd03df

Add an explicit check for AVRCP vendor data when parsing received vendor
commands or responses.

Bug: 21768387
Change-Id: I715de6fc7348d063c448971a8dae6dd1b00c7062

This is same race condition observed in btif_media_task_aa_tx_flush_req()
because btif_media_task_aa_tx_flush_req() and btif_media_task_stop_aa_req()
are called back-to-bach.

This race condition is triggered when A2DP audio is streaming on shutdown:
"btif_a2dp_on_stopped() -> btif_media_task_stop_aa_req()" is called
to stop the particular audio stream, and this happens right after
the "cleanup() -> btif_a2dp_stop_media_task()" processing during
the shutdown of the Bluetooth stack.

Bug: 22700411
Change-Id: Ia0c98d44a108cf0f57731ac8129e4d76c9934542

Closing an RFCOMM server socket does not remove the corresponding
service record from the security database. However, the RFCOMM
channel becomes free for reuse. The next RFCOMM server socket will
therefore "inherit" the service record for the closed one if it
happens to reuse the same channel.

Bug: 22880207
Change-Id: Ida3fee49e5f40667d9992dc4c4442f9289adae9e

It seems that the current implementation of btm_sec_encrypt_change()
does not handle the case when it is called with an invalid handle,
such as inside file btu_hcif.c :

  case HCI_SET_CONN_ENCRYPTION:
    /* Device refused to start encryption. ... */
    btm_sec_encrypt_change(BTM_INVALID_HCI_HANDLE, ...)

Bug: 22791224

Change-Id: Ide9404d0c82819399cf258ae3f90c25b352f1e20

Bug: 21963935
Change-Id: Id72001ce17996ea04c3eba32cebcac4dbbe317bc

btif_av_event_free_data() was not called in all states (idle state only)
leading to potential memory leaks.

Bug: 22822688
Change-Id: I40520c605c9a806e6cd5ee6e36c101d0aa8d4355

This change checks the transport type on receiving the disconnection
and resets the encryption key size only if the LE disconnection
has taken place.
This fixes the issue where read request to characteristics that
require encryption,  fails after cross key derivation due to
disconnection of BR/EDR transport.

Bug: 22515016
Change-Id: If6aad91a628eabbb5a4b7f5c22812fe94d4c5db2

Promoting AVCTP version to 1.4 from 1.2 as Bluedroid stack
already has got support for the same.
Below changes are incorporated as part of this:
 - AVCTP version in SDP entry is upgraded to 1.4 for both AVRCP
   Target and controller role.
 - Service class #1 is added in extra for AVRCP Controller SDP
   entry to be in sync with AVCTP version change.
 - Browsing support is not enabled as that needs corresponding
   profile level implementation, which is currently unavailable.

Change-Id: I72f7f67eb0a789fd321e9468f2a51bb5e9385a89

Use the correct offset_len when allocating buffers for AVRC
packets that need to be fragmented.

Bug: 22156175

Change-Id: I7db12474c84edacb4f0739d50a43e8cebdcca676

The uhid_event function used to expect that the return value of each
read function call to match with the size of the struct uhid_event.
Since the header file doesn't match the kernel driver, these two size
don't always match. The exact size check is replaced with expecting
the minimum required size.

Bug: 20108348
Change-Id: Ib61537092b109296f8290d802b68fc2efe78888c

When unparing HID devices, the attr_mask was not cleared.
By not clearing this data, some HID device will never be able
to connect again. (E.g. the Apple Magic Mouse)

Bug: 15566403
Change-Id: Ic80909dcecdf48d967c1a936b31554653761fa42
Signed-off-by: Casper Bonde <c.bonde@samsung.com>

Previously, the logic for stopping the timers didn't take
into account whether each timer was already running.

Bug: 22666419
Change-Id: Ia99bf8be917e9ea69f478a954085336fc899040a

- Prevent possible endless loop if |nb_frame| is too high
- Remove off-by-one calculation before sending frames
- Remove log spam and add better debug info

Bug: 22658329
Change-Id: I374ee980aec48763beb49b4f6f8b076124cadf40

Fix an earlier commit that prevents from customizing some of
the Bluetooth sniff parameters - those cannot be overwritten
anymore in the bdroid_buildcfg.h file.

Also, fixed the indexing in tables bta_dm_pm_cfg[]
and bta_dm_pm_spec[]

Bug: 22676670
Change-Id: I3a7074b9a9c91d312dc5d4314b7c304baf4ae20d

This race condition is triggered when A2DP audio is streaming on shutdown:
"btif_a2dp_on_stopped() -> btif_media_task_aa_tx_flush_req()" is called
to stop the particular audio stream, and this happens right after
the "cleanup() -> btif_a2dp_stop_media_task()" processing during
the shutdown of the Bluetooth stack.

Bug: 22602117
Change-Id: I5de6a8f15b6a2771dde2e299a5b60554063696a2

This patch checks for the NVRAM data when inquiry result
is received from the remote device. In case the device is
marked as LE only or BR/EDR only, and inquiry result is
received from the alternate transport, the device type is
marked as Dual mode in the NVRAM

Bug: 22604450
Change-Id: Id925e8bad152a33c2bd3c371ca42a6f9c694e3b0

In case of pairing to an already paired device (in an instance
where the central remote has removed the keys), the change
makes sure that the correct address is used when LTK key
request occurs at the peripheral.

Bug: 22605510
Change-Id: I959003f39f70281ff1e6af8d4c4549138bc1682c

This change ensures that the completion of SMP pairing for LE
devices specifically requests remote device discovery based on
transport type set to LE to prevent initiation of BR/EDR
connection due to incorrect transport type info

Bug: 22515456
Change-Id: Id1e5603d3cc53ca3dff427b93059a00f8d9150a7

Bug: 22634292
Change-Id: I64745671f3b69fc7e6913213aeec55267e9bc49b

This change enables the enhanced connection complete event in order
to receive the local rpa address while connection is completed
with a paired peripheral. This is required if central tries to
repair or key upgrade an already paired device so that the confirm
or the DHKey Check values match during the pairing process

Bug: 22515703
Change-Id: If27f3b22bc568df6f081f8ad13dfc6783a83ae47

* Removed btif_storage_is_device_bonded(), because it is not needed,
  and it was giving the wrong answer in use cases like Smart Setup
  with BR/EDR connections.

* Added a call to btif_storage_remove_ble_bonding_keys()
  within btif_storage_remove_bonded_device() so the bonded device
  state is properly removed.

* Don't save the BLE bonding keys if it is temporary bonding

Bug: 22233299
Change-Id: I33d9f76a124acc60173f0acaa517bc29ee6603e8

Bug: 22473630
Change-Id: I1168444a6969532377e1bbcb2b1058d768099c20

This patch fixes a bug that caused all multi-advertising instances to be
initialized with Instance ID 0 (which is not allowed by the stack), if LE
privacy is not available. The problem was that the internal data structures that
represent advertising instances were not getting their |inst_id| field
initialized where it's supposed to.

Although far from clean, this code worked before. The culprit for the regression
is "f9fdf890 Random address does not get written properly", which moved the
initialization of instance IDs from the loop in BTM_BleEnableAdvInstance to
btm_ble_multi_adv_enb_privacy. The latter never gets called if privacy is not
available, which leads to partially initialized structures.

Obviously both of these places were wrong to begin with. I saw the word
"init" in a function called btm_ble_multi_adv_init. I figured this might have
something to do with initializing, so I moved the logic there.

Bug: 21267281

Change-Id: I38b9a2a71cd3f45feb267a13bf29f93564ee6075

This change allows the host to mask the Cross key bit in
the key distribution fields of the pairing request and
response while pairing with the Moto key link. This is
needed since this specific remote device performs
calculations of the pairing confirm after masking the
cross key bit in the key distribtuin field which results
in a mismatch of the pairing confirm calcualted by the host
DUT and the remote.

Bug: 22539807
Change-Id: I7d758b05023cf346c97939883edda9d3adb08e1a

This patch fixes an issue that is reproducible in highly noisy environments
(massive deployment of BLE beacons), through repeated connect/disconnect
attempts on a remote peripheral while scanning for beacons in the background.
The state machine in bta/gatt has a special control flow for handling disconnect
requests during discovery, which in this case failed to resolve the original
request by issuing an HCI_Disconnect command. This is now fixed by always
explicitly triggering the connection close sequence once the discovery state has
been cleaned up.

This patch also includes a fix for a crash that occurred as a side-effect of the
scenario described above.

Bug: 22350508
Change-Id: Ie9cbd3c8f54239b142bfb8dde80d9581ae70ed43

This function is used for PM sniff mode timers - at least until we can
re-factor those not to rely on this function.

Bug: 22040710
Change-Id: Ibe6f49440228732b2bd8242db4a44e481b00b62e

This patch ignores the input parameter and used the own
address type from address management block while setting
the batch scan parameters.

Bug: 22227689
Change-Id: I23fc80b68cfbd91d718c1094ef2483eaccda3ae1

Allow PM to support multiple delay timer for different profiles and
power mode requests. And set correct connectivity mode.

Bug: 22040710
Change-Id: Idabd9ea944f0c5a89ce542d85db9f103fa7d1816

Use case: PTS test case for cross transport link key
derivation

Steps to reproduce:
1. Start Test case GAP TP/LEP/DM/BV-15 in PTSv6.0

Failure: PTS test case TP/LEP/DM/BV-15 fails as the link
key is not stored when the key is derived from other
transport key. So it will again go for pairing even though
it has link key but not stored.

Root cause: Link key is stored only when bond type is
presistent, but it sets the bond type to persistent only
when it receives SSP request but not link key is derived.

Fix: So making the solution generic to store always when the
link key type is authenticated combination key with P-256.

Bug: 22486860
Change-Id: I3eb201262f72c86a9438351cad14a6698065fa90

As per feedback from manufacturer.

Bug: 21817410
Change-Id: I4b5b3d8d99f9299754221968a5e1ca5f21cb9478

Bug: 21817410
Change-Id: I5c5f6da9e8326e4c9424907337331b39fb67b444

Change-Id: I7177aaa0e2d0c3b0fa79ae655c0171f0703c54f3