Allow them to be static.
This is required for a couple of reasons:
- enabling HIDL passthrough in recovery
- enabling VTS tests to be static blobs
Bug: 32920003
Bug: 64040096
Test: update-all-google-makefiles.sh
Change-Id: I1b2401fb273ab80819e3870aa71fe742269674ba
The byte offsets for Packet Boundary (PB) and Broadcast (BC) flags
in the HCI ACL Header is swapped. This will fixed it so that it matches
the Bluetooth Specifications.
Bug: 63241720
Test: Run manual VTS tests for Bluetooth
Change-Id: Ifc1f4d4f054a1dab13458c563e47bfb884f39d92
* Add libbluetooth-types - library containing types implementation, that
is common between stystem/bt and packages/apps/Bluetooth. It must be
included in every project using btif interface.
* Put Raw Address implementation into libbluetooth-types
* Unify all "to/from string" helper methods into ToString and FromString
* bd_addr_empty -> RawAddress::kEmpty
* bd_addr_any -> RawAddress::kAny
Also fix leaks in jni str2addr by adding ReleaseStringUTFChars
Test: types_unittest
Change-Id: I7dca82ec2293ae5338a9500dd6aa43b59d25657a
Merged-In: I7dca82ec2293ae5338a9500dd6aa43b59d25657a
If to send type and data separately for one HCI packet,
it will cause two system call context switch to kernel space,
which will introduce software overhead on data path.
Plus, if vendor does not use pure UART interface, it causes different
data behavior on BUS and may not adapt to all vendors as legacy HAL did.
Considering backward-compatibility, to use writev to send
type and data together once as legacy BT HAL did.
Test: H4 UTTest, BT VTS test, Bluetooth on/off
Change-Id: I2d93085fe0c01b48d0e3729a3fa85b5b27335b2c
android.hidl.base@1.0 and android.hidlmanager@1.0 are built into libhidltransport.
Test: links
Bug: 33276472
Change-Id: I424371932af129bba9d58dcf8652c64297f35f05
It is not used by bluetooth and nfc services.
Bug: 33241851
Bug: 37305530
Test: built and tested on sailfish
(cherry picked from commit 593eeb7a4b)
Change-Id: I599073296baa5eee72a1fce8caa4eb8352f66218
Merged-In: I599073296baa5eee72a1fce8caa4eb8352f66218
The test is still built in data/nativetest64, but it must be marked as
vendor since it depends on vendor modules.
Also proprietary -> vendor.
Test: build with BOARD_VNDK_VERSION := current
Test: run test on internal marlin, passes
Bug: 33241851
(cherry picked from commit 2cfae52a2e)
Merged-In: I856838fe208a4cd038ea16d47f03f6278224b2ee
Change-Id: I856838fe208a4cd038ea16d47f03f6278224b2ee
Helper libraries also now on vendor partition.
Test: links
Test: bluetooth works on internal Marlin
Bug: 33241851
Change-Id: Ie99faf9146ce2ce3df83a2b46acfaa36a54c16c2
EAS (Energy-Aware Scheduler) tends to cluster tasks together unless
they're marked to prefer an idle CPU (prefer-idle) during placement.
This is to save energy however it increases latencies. This is
noticeable when the Binder driver demotes the RT task to CFS and when
we try to do load balance in this path, the selection path still selects
a busy CPU. Add Bluetooth HAL service to the foreground schedtune group
to fix this as its critical for tasks like bluetooth audio playback.
Test: Boot and verify bluetooth HAL is in the foreground stune group
Bug: 36867985
Change-Id: Ib3c43cca4a2049af3700978b58a05989cc20b6ac
Signed-off-by: Joel Fernandes <joelaf@google.com>
VtsHalBluetoothV1_0TargetTest uses VtsHalHidlCallbackBase.
Callback class extends VtsHalHidlCallbackBase and use waitForCallback, notifyFromCallback, and setWaitTimeout
The original Bluetooth test contains the following use cases:
* Wait for specific callback
* Set timeout for specific callback waiting
The modified version supports the above use cases.
Bug: 36036059
Test: Test pass locally
Change-Id: I961baaf7f1b09fa9c134629cc7280565a2664ed5
Log if the Bluetooth service has died.
Also check the return status for Bluetooth service binder callbacks to
avoid DEAD_OBJECT exception when com.android.bluetooth crashes.
Bug: 36026072
Test: unit tests; Bluetooth off/on/off; kill -6 com.android.bluetooth
Change-Id: Ia28e6777ee7b78f06db8d5214324e696441d00f7
The vendor open interface command returns the total number
of FDs opened. (2 or 4 if the interface is MCT)
However, the list of FDs is always 4 in either case. If the
open command returns 2, the FDs are shared cmd/event and
data in/out.
2 FDs example
[41, 41, 42, 42]
4 FDs example
[41, 42, 43, 44]
If we condition the registration of ACL_IN on the total
number of FDs, then 2 FD MCT interfaces will not get data.
Fixes: 36067612, 36035039, 36188793
Test: manual, pairing is quick and data does transfer
Change-Id: Icc8728239ba81426d5fc0e678c5c5480fd9b5081
The power mode should be unimportant if power off succeeds.
Test: Switch users (Toggles Bluetooth under CPU load)
Bug: 36026072
Change-Id: I09fafe4d4f9a466a7e12e617925f98ef3805c738
- Hold the mutex during callbacks
- Disable power mode before deleting hci_
Bug: 36084266
Test: Switch users repeatedly (toggles Bluetooth)
Change-Id: I7e4d405e9028c3363c112a8b09c03b8aea652b7a
Some devices still use MCT as their UART protocol, so
we need to bring it forward to the fresh new HAL.
Also, adding tests for H4 while I'm here.
Test: new unit tests pass and a device using MCT now
boots BT again \o/
Fixes: 34992730
Change-Id: Idb8e536a2779929ad8a0d4bac492c3011995cd79
Bug: 35328775
Test: works in both binderized and passthrough modes
Merged-In: I18f21b53c2282e27342a58716475ae213a2837d4
Change-Id: I18f21b53c2282e27342a58716475ae213a2837d4
