Stub output stream is like a /dev/null audio device, however
for synchronous transfers it must block for the duration of
the audio chunk being transferred. Implement more accurate
accounting for the blocking time using the same approach as
employed by the remote submix stream implementation.
Bug: 356719263
Test: atest CtsNativeMediaAAudioTestCases
on aosp_cf_x86_64_auto target
Change-Id: I137aed397246bcf2b5ef6789aa4d2d27ead64467
Move the cleanup of the stream worker thread from '~StreamCommonImpl'
up to concrete stream implementations. This is because when
the worker thread is stopping, it calls 'DriverInterface::shutdown'
method of the stream. At the time when '~StreamCommonImpl' is
running, the concrete stream class has already been destroyed.
The cleanup actually only happens in the case when the client
did not close the stream properly via 'IStreamCommon.close', or
when the stream creation has failed in the middle.
Bug: 355804294
Test: atest VtsHalAudioCoreTargetTest
Change-Id: Ie86f682af202976ed48d24338b2dffcfd20d9a76
Calling 'stop' can cause the worker thread to terminate
before it handles the 'halReservedExit' command. This should
be avoided because a proper exit sequence may do cleanups.
Since all stream workers must handle the 'halReservedExit'
command, use of 'stop' should not be needed (if the thread
code gets stuck on a call to drivers, calling 'stop' will
not interrupt this), thus it is being replaced by 'join'.
Bug: 344482249
Test: atest CtsMediaAudioTestCases
Test: atest VtsHalAudioCoreTargetTest
Change-Id: If13f7239423657b80091239ff67e7fe350957e2e
Merged-In: If13f7239423657b80091239ff67e7fe350957e2e
1. add createMmapBuffer() for vendor to override and create mmap fd.
2. add refineMmapPosition() for vendor to override and update
latency in mmap case.
3. fix testcases position check in mmap case.
Bug: 274456992
Bug: 345591089
Test: atest VtsHalAudioCoreTargetTest
Change-Id: Ie63fdd47c0ddc563d84699dfdf6d4e9b72b5af43
The HIDL implementation blocked for the audio buffer duration
when the output side of the remote submix pipe does not have
the sink connected. This behavior was accidentally removed
when fixing b/327220024.
Also, limit the amount of the debug messages displayed when
the sink is shutdown to avoid spamming the syslog.
Bug: 328347445
Test: repro steps from the bug
Test: atest VtsHalAudioCoreTargetTest
Test: atest --test-filter=".*AudioPlaybackCaptureTest.*" CtsMediaAudioTestCases
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:6507376fb940198e8c39ad6d5345d132b1dd7219)
Merged-In: Ia66cb7b4567d42a41bf4715b7d725d36510ac50c
Change-Id: Ia66cb7b4567d42a41bf4715b7d725d36510ac50c
When calculating buffer size, use powers of two
for "large" buffers only on SRs >= 44100 Hz.
With lower SRs, sizes rounded up to a multiple
of 2 give too high latency.
Bug: 328432688
Test: repro steps from b/328432688#comment3
Test: atest VtsHalAudioCoreTargetTest
Change-Id: I605ef8507fef030d349e85c33ed0229c9566ddb5
Align behavior with the HIDL implementation. This helps
to prevent glitches in the scenarios that use remote
submix for sending audio externally: Auto and Cast.
Bug: 322247032
Bug: 327220024
Test: Repro steps in b/327220024#comment14
Change-Id: If2e4ebd7145375a268ee5f0e4ab656bb3748b2fe
Merged-In: If2e4ebd7145375a268ee5f0e4ab656bb3748b2fe
When the framework uses mixers, it rounds up the buffer
size to the frame count which is a multiple of 16.
In the HIDL implementation, this adjustment was done after
the framework thread was already created, and FMQ was
created on the first transfer, using adjusted size. In
the AIDL implementation, the FMQ is created together with
the stream, using the size suggested by the HAL.
Bug: 321233946
Test: verify audio playback over S/W A2DP
(cherry picked from https://googleplex-android-review.googlesource.com/q/commit:f2f9ae08c2ee4e396bbe4b8c7e955bad725ce545)
Merged-In: I35a5479bfc87a290aff09d51415381948857a146
Change-Id: I35a5479bfc87a290aff09d51415381948857a146
A follow-up for aosp/2929441. The interval between retries
of connecting to BT proxy must be reduced significantly.
Otherwise, in situations when the BT is not connected
(that happens during tests), the resulting long delay causes
failures at the frameowork level due to resulting long
timeout.
Bug: 322820830
Test: atest audiosystem_tests --iterations 100
Change-Id: Idd84cb22b28d109151f98bb5883dca18a9ae928d
When ModuleBluetooth::createProxy is invoked on BT device
connection, the BT stack may not be fully ready yet,
and port registration can fail. This is an intermittent
state and registration should succeed after retrying.
Bug: 320838889
Test: atest pts-bot:A2DP/SRC/REL/BV-02-I
Change-Id: I0c7cf7c1c6a8ee03ef55b004f89139e3b56ee9cd
In case when the command was sent by the HAL itself (from another
thread), the worker thread must not post a reply. The only case
when a reply needs to be posted is in the case when the command
was sent from a VTS test. This case is identified by the fact that
the cookie has value '0'.
Bug: 300181540
Test: atest VtsHalAudioCoreTargetTest
Change-Id: Ifeb0722b5cf7346a694c5a938f6b324f5fa825f1
Implement IModule::dump to display the current state in
the audioflinger dump.
Throttle repetitive logging when there is nothing to read.
Remove stale comment for already fixed b/307586684.
Bug: 307586684
Test: adb shell dumpsys media.audio_flinger
Change-Id: I1f1f6e1658d035d46af3a933a825b20a78c7f297
On 32-bit builds, the 'long' type is 32 bit. Thus, it can not
be used to store the value returned from 'uptimeNanos'.
Bug: 318055805
Test: atest CtsMediaAudioTestCases
Change-Id: I4abf72415b7241728d3ddb1d11043dfd59a8d08a
In order to align with legacy behavior, when opening a stream,
the module must suggest the current configuration of the BT session.
For that to work, the BT device proxy must be opened prior
to creating a stream, code moved to ModuleBluetooth.
Fix minor inconsistencies and bugs found during testing.
Bug: 301213930
Bug: 316027906
Test: atest pts-bot
Change-Id: I04ddaf73be82f872a3f32a789563c3cbd648eb61
When there is a pipe established for a remote submix
device, suggest the configuration of the peer when opening
the other end.
Refactor SubmixRoute management to move it out from
StreamRemoteSubmix so that ModuleRemoteSubmix could also
use it.
Bug: 294962274
Test: atest audiorouting_tests
Change-Id: Ib31a662e7b65b92c614dc441a01160cae3485f3a
In 'Module::setAudioPortConfig' allow subclasses to provide
their own suggested configuration.
In 'Module::populateConnectedDevicePort' provide the ID of
the device port instance that will be created as a result
of connecting an external device.
Also, expose 'ModuleDebug' flags to subclasses so that they
can know when device connections are simulated.
Bug: 264712385
Test: atest VtsHalAudioCoreTargetTest
Change-Id: Iceb2bdeb61c3470554b592fe197efa54d3b9e578
When external device is about to disconnect, the audio framework will
notify the HAL module about the coming device disconnection so that the
HAL module could abort any active read/write operations on drivers to
avoid problems with the HW interfaces.
Bug: 279824103
Test: atest VtsHalAudioCoreModuleTargetTest
Change-Id: I9f960b8ae5df11a764e70bd63f98c0f8b8386c34
Merged-In: I9f960b8ae5df11a764e70bd63f98c0f8b8386c34
When external device is about to disconnect, the audio framework will
notify the HAL module about the coming device disconnection so that the
HAL module could abort any active read/write operations on drivers to
avoid problems with the HW interfaces.
Bug: 279824103
Test: atest VtsHalAudioCoreModuleTargetTest
Change-Id: I9f960b8ae5df11a764e70bd63f98c0f8b8386c34
Similar to the primary module implementation, align the time
spent in the transfer operation with the duration of audio.
Change the read operation to ingest as much data as possible
during the audio burst duration.
Ensure that checking the existence of a SubmixRoute and adding
a new one is an atomic operation.
Minor improvements to avoid extra synchronization.
In the configuration, change the limit of max open streams to 10
to match the legacy implementation.
Bug: 302132812
Test: atest CtsMediaAudioTestCases --test-filter=".*AudioPlaybackCaptureTest#testPlaybackCaptureDoS"
Change-Id: Iccb6aaac46c039551c3d5f7760b1459168d9cfe5
android.hardware.bluetooth.audio should really be registered in its own
process, but this is a temporariy solution that is similar to the
old HIDL example.
Test: launch_cvd
Bug: 312265159
Change-Id: Ib60e2654d3b751bedf256d90de72921210174fe4
The default implementation of the primary HAL is intended to
work with the AVD. Current implementation of virtualized audio
I/O on AVD falls behind actual hardware in terms of I/O timing.
This is not tolerated by CTS tests which are tailored for
real hardware.
Make the primary HAL implementation more resilient to irregular
ALSA read/write behavior on AVD by prolonging I/O time, or
skipping operations when falling behind expected time.
Bug: 302132812
Bug: 302587331
Test: atest CtsMediaAudioTestCases
Change-Id: Ia290d9541a8a0e22d28024f7930ef554396d63c6
Fix the issue with retrieving BluetoothA2dp and BluetoothLe
instances access.
In 'Module::connectExternalDevice', always call
'populateConnectedDevicePort' to allow the module implementation
to cancel device connection.
Move frame count calculation into Utils.
Clean up includes and 'using' directives.
Bug: 302132812
Test: atest VtsHalAudioCoreTargetTest
Change-Id: I22f65b8bcdcdd9bcd67a8290520eb4a465d4c720
The latency figure depends on the module implementation.
Instead of using a hardcoded value, each module should be
able to specify its own value. This value is then used
for calculating the minimum buffer size.
Set the nominal latency of the primary (CF) module to a high
value since the virtual device implementation fails CTS tests
if it attempts to pretend that it provides low latency.
Bug: 302132812
Test: atest CtsMediaAudioTestCases --test-filter=".*AudioTrackTest.*"
Test: atest CtsMediaAudioTestCases --test-filter=".*AudioRecordTest.*"
Change-Id: I8ce9f230378eea787c9b3c7ce3660c1e4e7bc895
The implementation had duplicated code in 'transfer', which already
present in 'outWrite'.
Cleaned up delay calculations and logging.
Fixed the VTS to send 'prepareToClose' before attempting to join
the worker. Otherwise, the worker could be stuck on a blocking
operation due to inactivity of the other party and join would never
happen.
Bug: 302132812
Test: atest VtsHalAudioCoreTargetTest --test-filter="*AudioModuleRemoteSubmix*"
Change-Id: Id8455eb12d1d2999dc0bc7b64f0d70a61a177598
** Partial upstream of ag/24854732. Only the interface part is
** included.
This should enable the sound dose gts on cuttlefish devices. The sound
dose HAL uses the internal MelProcessor to compute the MELs which are
reported to the framework.
Test: atest GtsAudioTestCases:SoundDoseTest
Bug: 301527435
Change-Id: Ifc505a0171bc8b4d3f5cf65d950fa5c0f812087f
Merged-In: Ifc505a0171bc8b4d3f5cf65d950fa5c0f812087f
