This field indicates the frame size and is filled by
the HAL module, so that the client does not have
to calculate it.
As a side effect, in the VTS code, a StreamContext can
now be created solely from a StreamDescriptor.
Added unit tests for the functions from Utils.
Bug: 205884982
Test: atest libaudioaidlcommon_test
Test: atest VtsHalAudioCoreTargetTest
Merged-In: Ief836b8b2d35bacb1f9778e2462d540554149d7f
Change-Id: Ief836b8b2d35bacb1f9778e2462d540554149d7f
(cherry picked from commit 5862c1e3bc)
This patch implements audio I/O for the synchronous, non-MMAP
case.
Updated the StreamDescriptor structure to make it usable.
Clarified comments on the expectations for the client and
the HAL module.
Bug: 205884982
Test: atest VtsHalAudioCoreTargetTest
Merged-In: I09651c6e80a397c80870622ac19234b4d4a38cbb
Change-Id: I09651c6e80a397c80870622ac19234b4d4a38cbb
(cherry picked from commit 01803d454a)
Fix the mistake of making StreamWorker to inherit
from the part which provides actual thread logic (Impl).
The lifetime of the logic object must be longer
than the lifetime of the StreamWorker's thread.
Otherwise, the thread could still have running while
the logic has already been destroyed (consider
the order of destructors in C++ class inheritance).
With this fix, the StreamWorker class does not have
to be a template anymore, thus reorganize the code
to move big methods into a .cpp file.
Bug: 205884982
Test: atest libaudioaidlcommon_test --iterations
Merged-In: I5bc2c8fd9d78a0fbc9fddab67456cc5214584045
Change-Id: I5bc2c8fd9d78a0fbc9fddab67456cc5214584045
(cherry picked from commmit 84024eccee)
Change-Id: I70958f437657b574cda6480c3216a0b1ea252433
Enhance the return type of the 'workerCycle' to allow it
exiting without inducing an error on the controller side.
Also, put StreamWorker into a namespace.
Bug: 205884982
Test: atest libaudioaidlcommon_test --iterations
Merged-In: I3b27028b10f80f27985040cae8f8b0e6ab63ddad
Change-Id: I3b27028b10f80f27985040cae8f8b0e6ab63ddad
(cherry picked from commit 5021df71c7)
Replace 'platform_apis: true' with 'sdk_version: module_current'.
The former causes compiling against SDK source code,
this is not needed for the audio HAL.
Bug: 205884982
Test: m android.hardware.audio.service-aidl.example
Change-Id: I9b9c32707ac7276f5cd077681bd1f8122bc20feb
Merged-In: I9b9c32707ac7276f5cd077681bd1f8122bc20feb
(cherry picked from commit 714e959c89)
The Light HAL implements a function getLights that returns a vector
of HwLight objects. This vector is called "lights" in Lights.cpp,
but it is called "types" in Lights.h. Lights.h has been changed to
call the vector "lights".
Test: VtsHalLightTargetTest
Change-Id: I1e567d1454d42e3efc0186cde8340c32e7f3456e
This patch adds necessary structures and prepares for implementing
data flow for audio I/O.
Also in this patch we clarify the relationship between audio patches
and buffer size for audio I/O, and between buffer size and latency.
Bug: 205884982
Bug: 233816848
Test: atest VtsHalAudioCoreTargetTest
Merged-In: I8522632607d4cf50a112225c19b5dd5ad8848591
Change-Id: I8522632607d4cf50a112225c19b5dd5ad8848591
(cherry picked from commit 68bee70442)
Audio threads need to be able to set their priority.
Also, traditionally these worker threads set a custom
thread name. Implement this functionality in StreamWorker.
Since initialization steps can fail, implement simple
error reporting via a string field `mError`. The state
of the string field replaces the dedicated `ERROR` worker
state.
Bug: 205884982
Test: atest libaudioaidlcommon_test --iterations
Merged-In: Ie9ab94922d47f277a4993a90b478a2fa76657923
Change-Id: Ie9ab94922d47f277a4993a90b478a2fa76657923
(cherry picked from commit e9e0f7c0f5)
Avoid taking a lock in high priority worker threads
without a real need.
Bug: 205884982
Test: atest libaudioaidlcommon_test --iterations
Merged-In: I8cc0f5cb58752b7b7d413a9f4e46093c39445892
Change-Id: I8cc0f5cb58752b7b7d413a9f4e46093c39445892
(cherry picked from commit d989a4b669)
This utility class has been copied from HIDL VTS.
It will be used both for the default implementation
and AIDL VTS, and might need modifications.
Bug: 205884982
Test: atest libaudioaidlcommon_test
Merged-In: I43b35b0c23ae45305dca66e15b60820cad19635e
Change-Id: I43b35b0c23ae45305dca66e15b60820cad19635e
(cherry picked from commit c17f0484bc)
Certain mutation testing -- mutateOperandLifeTimeTest and
mutateOperandInputOutputTest -- can introduce potentially very large
CONSTANT_COPY operands, which can in turn create potentially very
large Models which must be passed across binder. To avoid overflowing
the binder buffer, we estimate the size of the mutated Model, and skip
the test if that size is too high. The old logic recognizes that our
tests only have a single active binder transaction at a time, and
assumes that there are no other clients using the same service at the
same time, and so we should have the binder buffer to ourselves; to be
conservative, we reject any Model whose estimated size exceeds half
the binder buffer size. Unfortunately, sometimes the binder buffer
still overflows, because it unexpectedly contains an allocation from
some other transaction: It appears that binder buffer memory
management is not serialized with respect to transactions from our
tests, and therefore depending on scheduler behavior, there may be a
sizeable allocation still in the buffer when we attempt to pass the
large Model. To fix this problem we become even more conservative,
and instead of limiting the Model to half the binder buffer size, we
limit it to half IBinder.MAX_IPC_SIZE (the recommended transaction
size limit). To confirm that this change does not exclude too many
tests, I checked how may times the size filter function
exceedsBinderSizeLimit is called, how many times it rejects a model
under the new logic (modelsExceedHalfMaxIPCSize), and how many times
it rejects a model under the old logic (modelsExceedHalfMaxIPCSize).
Test: VtsHalNeuralnetworksV1_0TargetTest --gtest_filter=TestGenerated/ValidationTest.Test/*-*dsp*
Test: # models = 3592, modelsExceedHalfMaxIPCSize = 212, modelsExceedHalfBufferSize = 18
Test: VtsHalNeuralnetworksV1_1TargetTest --gtest_filter=TestGenerated/ValidationTest.Test/*-*dsp*
Test: # models = 7228, modelsExceedHalfMaxIPCSize = 330, modelsExceedHalfBufferSize = 28
Test: VtsHalNeuralnetworksV1_2TargetTest --gtest_filter=TestGenerated/ValidationTest.Test/*-*dsp*
Test: # models = 52072, modelsExceedHalfMaxIPCSize = 506, modelsExceedHalfBufferSize = 28
Test: VtsHalNeuralnetworksV1_3TargetTest --gtest_filter=TestGenerated/ValidationTest.Test/*-*dsp*
Test: # models = 73342, modelsExceedHalfMaxIPCSize = 568, modelsExceedHalfBufferSize = 28
Test: VtsHalNeuralnetworksTargetTest
Bug: 227719657
Bug: 227719752
Bug: 231928847
Bug: 238777741
Bug: 242271308
Change-Id: I3f81d71ca3c0ad4c639096b1dc034a8909bc8971
* changes:
audio: Add support for compressed offload
audio: Improve debug logging in the AIDL version, fix bugs
audio: Disable the C++ backend for android.hardware.audio.core
- Add compressed offload mix port into default implementation.
- Require AudioOffloadInfo to be passed to IModule.openOutputStream
for compressed offload port configs.
- Update VTS to handle compressed offload.
Bug: 205884982
Test: atest VtsHalAudioCoreTargetTest
Merged-In: I118b2c04bff12b64a7cac4dc2c88217a6a270046
Change-Id: I118b2c04bff12b64a7cac4dc2c88217a6a270046
(cherry picked from commit 975ea3ae89)
- Make sure the AIDL default implementation has debug output enabled.
- Log additional info in the AIDL VTS to facilitate debugging.
- Make resource handler classes move-only types.
Bug: 205884982
Test: atest VtsHalAudioCoreTargetTest
Merged-In: I111b72aaf12962f00b4d31b8ac87186bca5eb853
Change-Id: I111b72aaf12962f00b4d31b8ac87186bca5eb853
(cherry picked from commit f82fc6476d)
android.hardware.audio.core will need to use FMQ,
which does not support the C++ backend in the AIDL version.
Migrate VTS tests from the C++ backend to the NDK backend.
Bug: 205884982
Test: atest VtsHalAudioCoreTargetTest
Merged-In: Ia5d29126afdb4bb97dd063a032b96ab83b9ce36e
Change-Id: Ia5d29126afdb4bb97dd063a032b96ab83b9ce36e
(cherry picked from commit 8fcc442416)
Add methods 'IModule.connect/disconnectExternalDevice' which inform
audio HAL about connection / disconnection of an external
non-attached device. Add method 'getAudioRoutesForPort' to
retrieve only routes that include the specified port.
Update the behavior of 'getAudioPorts' and 'getAudioRoutes'
indicating that the result may change due to instantiation
of new device ports for connected external devices.
Clarify behavior of 'IModule.setAudioPortConfig' that it can not
work with device ports with no profiles.
Add debug flags structure 'ModuleDebug' and method
'IModule.setModuleDebug' to control the debugging aspects. VTS
tests use these flags to test HAL behavior which would otherwise
require human intervention.
Update the default implementation and VTS for the AIDL changes.
Bug: 205884982
Test: atest VtsHalAudioCoreTargetTest
Merged-In: Iad5f7009e283729206f88b6278c8992f7f8a92a2
Change-Id: Iad5f7009e283729206f88b6278c8992f7f8a92a2
Using service_fuzzer_defaults in
- android.hardware.health-service.aidl_fuzzer
- android.hardware.vibrator-service.example_fuzzer
Bug: 240318902
Test: m android.hardware.health-service.aidl_fuzzer
Test: m android.hardware.vibrator-service.example_fuzzer
Change-Id: I3fd1cf37a5835ebef14658ed40332132cec07e49
... for the same reason as cc489264af
Bug: 225941299
Test: m
Merged-In: I269d61c8c441b3db0832f5450f5e5a84a5989a99
Change-Id: I269d61c8c441b3db0832f5450f5e5a84a5989a99
(cherry picked from commit 446519a370)
Background:
android.hardware.audio.common was frozen in the upstream tm-dev.
Previously, the interface had only V1 (as the in-development version)
and it was in the VNDK. However, after the freeze, it will have V1 and
V2, and only V1 will remain in the VNDK, because we can't have multiple
versions of the same AIDL interface in VNDK (it's just waste of memory).
android.hardware.audio.core is a new AIDL interface that doesn't exist
in TM. It is also a member of VNDK and imports the audio.common
interface. So far, the import has been done without an explicit version
number (i.e. `imports: ["android.hardware.audio.common"]`). As a result,
V1 was imported before the freeze, and V2 (the latest) will be imported
after the freeze.
Problem:
audio.core depending on audio.common-V2 is NOT allowed since the former
is a member of VNDK, but the latter isn't. VNDK should form a closure.
We are actually hitting this:
```
> error: hardware/interfaces/audio/aidl/Android.bp:71:1: module
"android.hardware.audio.core-V1-ndk" variant
"android_vendor.Tiramisu_arm64_armv8-a_static" (created by module
"android.hardware.audio.core-V1-ndk-generator" (created by module
"android.hardware.audio.core_interface")): (native:vendor:vndk) should
not link to "android.hardware.audio.common-V2-ndk" (native:vendor):
VNDK-core must only depend on VNDK-core or VNDK-SP
```
To fix this, make the dependency to audio.common to be explicit about
the version. This change can't be uploaded to tm-dev as audio.core
doesn't exist there, and must be submitted before the AIDL APIs are
frozen there.
Bug: 225941299
Test: m
Merged-In: Iadc96d558d5c7431b0a2b0cbd28b99319a24ee44
Change-Id: Iadc96d558d5c7431b0a2b0cbd28b99319a24ee44
(cherry picked from commit cc489264af)
This adds basic support for enumerating capabilities
of an audio module, creation of patches, and opening of
I/O streams.
Bug: 205884982
Test: atest VtsHalAudioCoreTargetTest
Merged-In: Ib1ef13c95d338b6ef1bde7d87dcd675fc1d35e85
Change-Id: Ib1ef13c95d338b6ef1bde7d87dcd675fc1d35e85
(cherry picked from commit ee27478264)
VtsHalEvsTargetTest does not verify any behavior depending on
a frame buffer's color format.
Bug: 240685726
Test: atest VtsHalEvsTargetTest
Change-Id: If8b8c595cfc7404a8122728593f778528b45af15
In the reference implementation of the identity function, there are two places where the memory requested in the openssl algorithm is not released. This memory should be freed.
Test: Vts/Cts
Bug: 242927524
Change-Id: I88ffba39cb6ec887f395122e4670bf9f1a2d8e12
Previously, the build rules assumed that the target supports both 64 and
32 bit architectures, which isn't true for 32-bit-only targets. That
caused build breakage on such targets. Fixing that by building the
modules only for the first architecture (64 on 64-only or 64/32 target,
and 32 on 32-only target).
Bug: 241916304
Test: build the 32-bit-only target
Merged-In: I5353ab1561622eae12e2dff87eb0f17b47c3773d
Change-Id: I5353ab1561622eae12e2dff87eb0f17b47c3773d
(cherry picked from commit 3ab5c90104)
