As oneway HIDL calls generate methods that return 'Result<void>',
the returned value must be checked to avoid triggering an
assertion if the result was an error (e.g. due to the client death).
Bug: 151900655
Test: N/A
Change-Id: I4721240b5959d6f11cb4447b30928bf4c83da058
The callback is targeted for events related to an output stream.
Currently, there is one callback event defined, which is codec
format changed event.
Bug: 133526565
Test: manual
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: I73a4914c1ffc30e1c88b8fedd61a031e24a069f6
Test: no regression with offloaded Play Music
Bug: 133526565
Merged-In: Ie655a96503be5a4ad1660d4b2183b01e514452fd
Change-Id: Ie655a96503be5a4ad1660d4b2183b01e514452fd
(cherry picked from commit c91b679ac1)
Playback rate parameters allow changing the playback speed
of audio streams optionally preserving the pitch (this is
often referred to as "timestretch"). HALs can apply these
parameters when decoding of streams is done in hardware.
Bug: 133526565
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: Ie78cd924bd034b57bf5b6a74affd62641ffc0aba
These properties are mainly used on TV platforms.
Bug: 133526565
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: I5dd2fe9cb987bf8435b8e5da96f8590288e5707c
Add method 'updateAudioPatch' which should be used when
an existing patch needs to be updated with new routing.
Use of this method allows audio HAL to avoid disrupting
audio stream while changing routing.
Bug: 79248321
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: I6c87f67fa4f2463ba9e8f0272a3232f5c9c55714
Add a method to IDevice interface allowing the attachement of an
audio effect to an audio device.
This is used when an audio effect is implemented below the HAL (e.g by an audio DSP)
and is attached/enabled when a particular sink(e.g speaker) or source(e.g mic) device
is selected.
Bug: 136294538
Test: make
Change-Id: I73d78c4f234fd80443a1cb3772c2d65457968652
Some methods of legacy HALs are optional. The legacy HAL
can avoid implementing them either by setting the function
entry in the interface table to NULL or by setting it
to a stub function returning -ENOSYS. In the latter case
the default implementation was spamming syslog with
warnings. Specify that ENOSYS returned from optional legacy
HAL methods must be ignored.
Bug: 141989952
Test: check logcat
Change-Id: Idbdff8d88e3303a0181cb5629ab82485148e6d8e
Add documentation and check that 'minSizeFrames' parameter
is a positive value. Previously a similar check was added
to Pixel HAL.
Bug: 141989952
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: I526431fc214f7f5bc424f0c9e4c890cf918ca2b2
As audio device type can not be used as bit mask any more, refactoring
audio device type usages in conversion from/to HAL. Use a set of audio
device types instead of bit mask.
Bug: 135621476
Test: atest VtsHalAudioV5_0TargetTest
Change-Id: I1a0f574744f855bb1684cd28613571399781abbc
Merged-In: I1a0f574744f855bb1684cd28613571399781abbc
It is suggested to do so according to Vector.h and SortedVector.h.
At framework side, there are audio containers such as FormatVector,
ChannelMaskSet, SampleRateSet to replace the usage of Vector
SortedVector for audio stuff. In default hal implementation, we can
use these audio containers to replace Vector and SortedVector.
Bug: 135621476
Test: play/record audio
Change-Id: I4d985327fb76cd06afe241860d8b592abcbfe4f2
Merged-In: I4d985327fb76cd06afe241860d8b592abcbfe4f2
As audio device type can not be used as bit mask any more, refactoring
audio device type usages in conversion from/to HAL. Use a set of audio
device types instead of bit mask.
Bug: 135621476
Test: atest VtsHalAudioV5_0TargetTest
Change-Id: I1a0f574744f855bb1684cd28613571399781abbc
IDevice.close must not proceed if there are streams
that are currently opened on this device.
Bug: 114451103
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: I61d81bc0333098c341d5d551bf59331e49fcf682
Fixed behavior of IStream|IEffect.close to release the underlying
HAL resource synchronously. This is to avoid adding artificial
delays in VTS that become totally unpractical in V6.
Added clarification about expected client behavior for
IStream|IEffect.close w.r.t. audio data transfer.
Added IDevice.close method which releases HAL device resource.
Updated VTS tests to remove delays in V6.
Bug: 114451103
Bug: 141989952
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: I439f0f923c091af2ab234d15ca847cfade341f25
Merged-In: I439f0f923c091af2ab234d15ca847cfade341f25
Fixed behavior of IStream|IEffect.close to release the underlying
HAL resource synchronously. This is to avoid adding artificial
delays in VTS that become totally unpractical in V6.
Added clarification about expected client behavior for
IStream|IEffect.close w.r.t. audio data transfer.
Added IDevice.close method which releases HAL device resource.
Updated VTS tests to remove delays in V6.
Bug: 114451103
Bug: 141989952
Test: atest VtsHalAudioV6_0TargetTest
Change-Id: I439f0f923c091af2ab234d15ca847cfade341f25
Implement parsing of AudioPolicyManager config for finding
out supported format configurations of streams. This only applies
when running tests for HAL V6. Previously format configurations
mandated by CDD were used for testing, this does not work well
for non-primary modules.
Fix the following issues found while running the tests
for "r_submix" and "msd" modules:
- IStream::getSupportedFormats must return a status
to indicate that this capability is not supported by HAL;
- it is allowed for IStream::setDevices to return
NOT_SUPPORTED status.
Other changes:
- Factor out helper functions for generating format
configurations;
- Fix generation of the channel mask component in the names
of tests that use AudioConfig, add sampling rate to test
names.
Squashed with the following commit to avoid breaking compilation:
audio vts: Remove explicit dependency on the new types
Avoid using the new ChannelMaskSet and SampleRateSet types
directly to simplify upstreaming.
Bug: 141989952
Bug: 141847510
Test: atest VtsHalAudioV5_0TargetTest
atest VtsHalAudioV6_0TargetTest
also, run modified V5_0 test using generators for V6_0
Change-Id: If0d330881901908e546baab89f63d3333003e355
Merged-In: If0d330881901908e546baab89f63d3333003e355
Generate with ./newHal.sh 5.0
+ some typo fix in the .hal
+ some clang-tidy run
This new HAL is an exact copy of the V5.
It will be modified in following patches while R is developed.
Test: push V6 to device and check audio works fine
Bug: 134940862
Change-Id: Ic25d30175032e76e0d3208fb876451de9ae6b7b4
Merged-In: Ic25d30175032e76e0d3208fb876451de9ae6b7b4
Signed-off-by: Kevin Rocard <krocard@google.com>
Implement parsing of AudioPolicyManager config for finding
out supported format configurations of streams. This only applies
when running tests for HAL V6. Previously format configurations
mandated by CDD were used for testing, this does not work well
for non-primary modules.
Fix the following issues found while running the tests
for "r_submix" and "msd" modules:
- IStream::getSupportedFormats must return a status
to indicate that this capability is not supported by HAL;
- it is allowed for IStream::setDevices to return
NOT_SUPPORTED status.
Other changes:
- Factor out helper functions for generating format
configurations;
- Fix generation of the channel mask component in the names
of tests that use AudioConfig, add sampling rate to test
names.
Bug: 141989952
Bug: 141847510
Test: atest VtsHalAudioV5_0TargetTest
atest VtsHalAudioV6_0TargetTest
also, run modified V5_0 test using generators for V6_0
Change-Id: If0d330881901908e546baab89f63d3333003e355
Generate with ./newHal.sh 5.0
+ some typo fix in the .hal
+ some clang-tidy run
This new HAL is an exact copy of the V5.
It will be modified in following patches while R is developed.
Test: push V6 to device and check audio works fine
Bug: 134940862
Change-Id: Ic25d30175032e76e0d3208fb876451de9ae6b7b4
Signed-off-by: Kevin Rocard <krocard@google.com>
During the transition to Treble, "device connect" / "disconnect"
parameter was erroneously attributed to audio streams. This
issue got fixed in Audio HAL V4.0, but AudioParameter file
was still using incorrect name for the constant.
Test: make
Change-Id: I18e1e3fccad3d41372f3c5c046af4f6ffd83328b
Merged-In: I18e1e3fccad3d41372f3c5c046af4f6ffd83328b
Since these were combined into libhidlbase.
Bug: 135686713
Test: build only (libhwbinder/libhidltransport are empty)
Change-Id: I075670b64eebbbbd6a6ae0e84ad51bf1c6f5ba36
It is suggested to do so according to Vector.h and SortedVector.h.
At framework side, there are audio containers such as FormatVector,
ChannelMaskSet, SampleRateSet to replace the usage of Vector
SortedVector for audio stuff. In default hal implementation, we can
use these audio containers to replace Vector and SortedVector.
Bug: 135621476
Test: play/record audio
Change-Id: I4d985327fb76cd06afe241860d8b592abcbfe4f2
During the transition to Treble, "device connect" / "disconnect"
parameter was erroneously attributed to audio streams. This
issue got fixed in Audio HAL V4.0, but AudioParameter file
was still using incorrect name for the constant.
Test: make
Change-Id: I18e1e3fccad3d41372f3c5c046af4f6ffd83328b
Input value were not checked, resulting in out of range input accepted.
This was checked by test but never implemented in the default HAL.
Bug: 133105753
Test: atest VtsHalAudioV5_0Target
Change-Id: Ie6dae638b60daff6923668dc9637067f29e48b21
Signed-off-by: Kevin Rocard <krocard@google.com>
Since open input / output stream functions in legacy audio
HAL API do not have parameters for source / sink metadata,
make a call to update metadata right after opening a stream
in the default implementation.
Bug: 120859615
Test: audio still works
Change-Id: Ie81f861a024261a1f13fb6aa23976a92658e679c
Complement imports of audio::CPP_VERSION with audio::common::CPP_VERSION
for core.
Import namespaces in VTS tests.
Remove "using namespace" from .cpp files where there
are already imports in the .h file.
Test: try moving types between audio/types.hal and common/types.hal.
Change-Id: I49c2e2466c8175f2516c2273f5c26d125afd4801
Instead of importing every single type individually ("using <type>"),
import entire namespaces of
::android::hardware::audio::CPP_VERSION and
::android::hardware::audio::common::CPP_VERSION
inside the "implementation" namespace in the default implementation.
This simplifies modifying types, as it is not needed anymore
to surround "using <type>" directives with #if MAJOR_VERSION ...
Note that the contents of the namespaces are imported inside
the "implementation" namespace, which reduces risks of name collision
Also, fixed the namespace for HidlUtils--it needs to be in
'implementation'.
Test: make
Change-Id: I81bbb3074b5763717ae72e747a075fd46dc36d18
So that it can be used by the Bluetooth HAL.
Test: compile
Change-Id: I2e2b1e12e1f837183c4edbd1d330a13a67d1ecaf
Signed-off-by: Kevin Rocard <krocard@google.com>
find -name 4.0 | xargs -I@ cp -r @ @/../5.0
find 5.0 */5.0 -type f | xargs sed -Ei 's/V4/V5/;s#(@|/)4.0#\15.0#'
find -name *cpp -o -name *.h | xargs sed -i 's/VERSION == 4/VERSION >= 4/g'
mv {4.1,5.0}/config/audio_policy_configuration.xsd
Then a manual update to Android.bp to add V5 support.
Then a manual update to service.cpp to add support for loading 5.0.
If someone knows a way to avoid copying the .hal, it would be great.
They will be 99% identical between V4 and V5.
Bug: 118203066
Test: vts-tradefed run commandAndExit vts --module VtsHalAudioEffectV2_0Target
vts-tradefed run commandAndExit vts --module VtsHalAudioV2_0Target
vts-tradefed run commandAndExit vts --module VtsHalAudioEffectV4_0Target
vts-tradefed run commandAndExit vts --module VtsHalAudioV4_0Target
Change-Id: If99a5645d19c9780019704ea4f51f8114d83ee8f
Signed-off-by: Kevin Rocard <krocard@google.com>
Lots of Android.bp content was duplicated per version.
Bug: 118203066
Test: compile
Change-Id: I4072314f1e1278c6d876fc681ba20363728f151b
Signed-off-by: Kevin Rocard <krocard@google.com>
Use the same technique developed for the simpler libaudiohal to
factorize the audio HAL impl.
Generated with script:
set -euo pipefail
cd hardware/interfaces/audio
# Make 2.0 and 4.0 files mostly identical by removing all 2.0, V4.0...
# and replacing it by CPP_VERSION and FILE_VERSION
# This allow both implementation to be mostly identical, except for code difference
# Also remove IncludeGuard.h as it will be included from compiler flag
# Also update license to 2018
find -name *.cpp -o -name *h |
grep -v -e service.cpp |
xargs sed -Ei \
-e 's/(#.*)V[42]_0_(.*_H_?)$/\1\2/' \
-e 's/(LOG_TAG.*)[24]_0/\1/' \
-e '/#ifndef AUDIO_HAL_VERSION/,/#endif/d' \
-e '/^#.*def.*AUDIO_HAL_VERSION\b/d' \
-e's/V[42]_0/CPP_VERSION/' \
-e 's/(#include ).(android.*)[24].0(.*).$/\1PATH(\2FILE_VERSION\3)/' \
-e 's|(#include) .(.*)[24].0/(default/.*).$|\1 "\2\3"|' \
-e 's/\bAUDIO_HAL_VERSION\b/CPP_VERSION/' \
-e '/IncludeGuard.h/d' \
-e 's/(Copyright .C. 201)./\18/'
# set MAJOR_VERSION instead of having a different macro for each version
# this allow to do preprocessor arithmetic (eg: MAJOR_VERSION >= 4)
for v in 2 4; do
find -ipath '*/'$v'.0/*/Android.bp' |
xargs sed -i -e '/cflags:/,/^$/d' -e '/vndk/{:a;/}/!{N;ba};d}' -e '$i\
cflags: [\
"-DMAJOR_VERSION='$v'",\
"-DMINOR_VERSION=0",\
"-include common/all-versions/VersionMacro.h",\
]'; done
# replace# ifdef VERSION_X by #if MAJOR_VERSION == X
find -path *all-versions/*.h |
xargs sed -Ei -e 's/def.*AUDIO_HAL_VERSION_([24])_0.*/ MAJOR_VERSION == \1/' \
-e 'T;s/ +/ /'
# copy all-versions impl in each version impl
find -path '*/[24].0/*.[hc]*' |
grep -ve all-versions -e test |
xargs -P99 -n4 sed -i -Ee '/include <.*all-versions.default/!b' \
-e 's#.*<#find -path */#' -e 's/>$/|xargs tail -n +16/' -e e
# remove all-versions impl
rm -r {core,effect}/all-versions/
# merge version impl into a single all-version
for dir in core/*.0/vts/functional/*.0/ ;do
dest=$(echo $dir | sed 's#/..0/#/all-versions/#')
mkdir -p $dest
mv -T $dir $dest
done
find -mindepth 3 -path '*/2.0/*' -a \( -name '*cpp' -o -name '*h' \) |
grep -v 'all-versions' |
sed -E 'h;s/2/4/g;H;s/4.0/all-versions/;s/4.0//;H;g;s/\n/ /g;'|
xargs -P99 -L1 sh -c '
set -euo pipefail
mkdir -p $(dirname $2);
diff --old-group-format="#if MAJOR_VERSION == 2
%<#endif
" \
--new-group-format="#if MAJOR_VERSION == 4
%>#endif
" $0 $1 > $2 || true;
rm $0 $1'
# merge the X.0 Android.bp & OWNERS in the all-versions
for dir in common effect core; do
for sub in default vts/functional; do
test -f $dir/2.0/$sub/Android.bp || continue
awk 1 $dir/*.0/$sub/Android.bp >> $dir/all-versions/$sub/Android.bp
# delete licenses except for the first one and add whitelines
sed -i -e '/^}$/{N;/^}\n$/!s/\n/&\n/}' $dir/all-versions/$sub/Android.bp
sed -i -e '1,17b;/^\/\//,/^$/{d}' $dir/all-versions/$sub/Android.bp
done
for sub in default vts; do
test -d $dir/2.0/$sub || continue
test -f $dir/2.0/$sub/OWNERS &&
awk 1 $dir/*.0/$sub/OWNERS | sort -u > $dir/all-versions/$sub/OWNERS ||
true
rm -r $dir/*.0/$sub
done
done
# delete all-versions-impl dependencies
find -name 'Android.bp' | xargs sed -i -e '/all-versions-impl/d'
# cleanup unused files
rm common/all-versions/default/include/common/all-versions/default/HidlUtils*
rm common/all-versions/util/include/common/all-versions/IncludeGuard.h
find -depth -type d -empty -delete
# Clamp consecutive number of empty lines to 2
find -name *.cpp -o -name *h | xargs sed -Ei ':a;/^\n*$/{N;ba};s/\n\n+/\n\n/'
# transform #endif\n#if to #elif
find -name *.cpp -o -name *h | xargs sed -i '/^#endif/{N;s/.*\n#if/#elif/}'
# remove leftover include guard in cpp
find -name *.cpp |xargs sed -Ei '/^#.*_H_?$/d'
# apply clang-format
find -name *cpp -o -name *h |
xargs ../../../prebuilts/clang/host/linux-x86/clang-stable/bin/clang-format --style file -i
# clang format breaks PATH(a/b) to PATH(a / b), remove the space surrounding /
find -name *cpp -o -name *h | xargs sed -i "/#include PATH/s# / #/#g"
Test: compile
Bug: 118203066
Change-Id: I3692a444307afc5f71064fe0b9e6b8af3c9ff1dd
Signed-off-by: Kevin Rocard <krocard@google.com>
Rename files so that, if they are in a X.0 folder, their name does
not contain that version number. That allows for easier copy paste
when creating a new version.
Conversion.h was included from the all-versions instead of the X.0
headers. This in incoherent with the other headers.
Add VersionMacro.h that will be used in the follow up patch.
Also remove shim from VNDK as only stable ABI&API libraries should be
part of it.
Test: Compile
Bug: 118203066
Change-Id: I9cd77b3df5ccede5ae51c3cc94fab5939c51d9ef
Signed-off-by: Kevin Rocard <krocard@google.com>
That will permit minimal change on future refactors.
Bug: 118203066
Test: compile
Change-Id: Ibed095d3821e2f7b6177faa07775bde9821459c7
Signed-off-by: Kevin Rocard <krocard@google.com>
hidl_memory size is not used on the framework side but is checked by the
hidl framework to be a sensible value.
Due to a hack in O and P, the value can be negative, this leads to
errors in HIDL.
This patch make sure the hidl_memory size is always positive and thus
not affected by the hack.
Test: adb shell data/nativetest64/write_sine/write_sine -pl -m3 -x
Bug: 38118159
Change-Id: Ie53c46c558e8042d74ee32b55219195da82e4bcc
Signed-off-by: Kevin Rocard <krocard@google.com>
The HAL has a negotiation mechanism where if an open output/input
stream is made with unsupported parameters, it will return
a valid config and UNSUPPORTED_PARAMETRS.
This negotiation was confusingly logged as error.
Test: "adb logcat -s DeviceHAL" show no error on successful
playback/capture
Change-Id: I5383cdcfdf8a0ca66768ffc106a1b651b9d8b2d1
Signed-off-by: Kevin Rocard <krocard@google.com>
