For adapdations for Android versions 4 to 10, see pulseaudio-modules-droid-jb2q
Building of droid modules is split to two packages
Linking to libdroid is not encouraged, usually only HAL functions are needed which can be accessed using the pulsecore shared API (see below).
Supported Android versions:
Headers for defining devices and strings for different droid versions are in src/common/droid-util-audio.h.
When new devices with relevant new enums appear, add enum check to configure.ac. CC_CHECK_DROID_ENUM macro will create macros HAVE_ENUM_FOO, STRING_ENTRY_IF_FOO and FANCY_ENTRY_IF_FOO if enum FOO exists in HAL audio.h.
For example:
# configure.ac:
CC_CHECK_DROID_ENUM([${DROIDHEADERS_CFLAGS}], [AUDIO_DEVICE_OUT_IP])
CC_CHECK_DROID_ENUM([${DROIDHEADERS_CFLAGS}], [AUDIO_DEVICE_OUT_OTHER_NEW])
# and then in droid-util-audio.h add macros to proper tables:
/* string_conversion_table_output_device[] */
STRING_ENTRY_IF_OUT_IP
STRING_ENTRY_IF_OUT_OTHER_NEW
/* string_conversion_table_output_device_fancy[] */
FANCY_ENTRY_IF_OUT_IP("output-ip")
FANCY_ENTRY_IF_OUT_OTHER_NEW("output-other_new")In addition to the above macros there are also now defines HAVE_ENUM_AUDIO_DEVICE_OUT_IP and HAVE_ENUM_AUDIO_DEVICE_OUT_OTHER_NEW.
The purpose of droid-modules is to "replace AudioFlinger". Many hardware adaptations use ALSA as the kernel interface, but there is no saying that someday vendor would create and use something proprietary or otherwise different from ALSA. Also the ALSA implementation in droid devices may contain funny ways to achieve things (notable example is voicecall) which might be difficult to do if interfacing directly with ALSA to replace AudioFlinger. Also using ALSA directly would mean that the whole HAL adaptation would need to be ported for each new device adaptation. With droid-modules this is much more simpler, with somewhat stable HAL (HALv3 as of now, also different vendors add their own incompatible extensions) API. In best scenarios using droid-modules with new device is just compiling against target.
The common part of PulseAudio Droid modules contains library for handling most operations towards audio HAL.
Configuration parser reads audio policy xml files.
If the configuration is in non-default location for some reason "config" module argument can be used to point to the configuration file location.
By default files are tried in following order,
/odm/etc/audio_policy_configuration.xml
/vendor/etc/audio/audio_policy_configuration.xml
/vendor/etc/audio_policy_configuration.xml
/system/etc/audio_policy_configuration.xmlIdeally only module-droid-card is loaded and then droid-card loads configuration, creates profiles and loads sinks and sources based on the selected profile.
When module-droid-card is loaded with default arguments, droid-card will create a default profile (called unsurprisingly "default"). The default profile will merge supported output and input streams to one profile, to allow use of possible low latency or deep buffer outputs.
In addition to aforementioned card profile, droid-card creates some additional virtual profiles. These virtual profiles are used when enabling voicecall routings etc. When virtual profile is enabled, possible sinks and sources previously active profile had are not removed.
As an illustration, following command line sequence enables voicecall mode and routes audio to internal handsfree (ihf - "handsfree speaker"):
pactl set-card-profile droid_card.primary voicecall
pactl set-sink-port sink.primary output-parking
pactl set-sink-port sink.primary output-speakerAfter this, when there is an active voicecall (created by ofono for example), voice audio starts to flow between modem and audio chip.
To disable voicecall and return to media audio:
pactl set-card-profile droid_card.primary default
pactl set-sink-port sink.primary output-parking
pactl set-sink-port sink.primary output-speakerWith this example sequence sinks and sources are the ones from default card profile, and they are maintained for the whole duration of the voicecall and after.
This sequence follows the droid HAL idea that when changing audio mode the mode change is done when next routing change happens. output-parking and input-parking ports are just convenience for PulseAudio, where setting already active port is a no-op (output/input-parking doesn't do any real routing changes).
Current virtual profiles are:
Communication profile is used for VoIP-like applications, to enable some voicecall related algorithms without being in voicecall. Ringtone profile should be used when ringtone is playing, to again enable possible loudness related optimizations etc. Voicecall-record profile can be enabled when voicecall profile is active.
If mix port with flag AUDIO_OUTPUT_FLAG_VOIP_RX exists when communication virtual profile is enabled additional droid-sink is created with the config defined in the mix port. Voip audio should then be played to this new sink.
Normally user should not need to load droid-sink or droid-source modules by hand, but droid-card loads appropriate modules based on the active card profile.
Changing output routing is as simple as
pactl set-sink-port sink.primary output-wired_headphoneSinks or sources do not track possible headphone/other wired accessory plugging, but this needs to be handled elsewhere and then that other entity needs to control sinks and sources. (For example in SailfishOS this entity is OHM with accessory-plugin and pulseaudio-policy-enforcement module for actually making the port switching)
As droid HAL makes assumptions on (input) routing based on what the parameters for the stream are (device, sample rate, channels, format, etc.) normal PulseAudio sources are a bit inflexible as only sample rate can change after source creation and even then there are restrictions based on alternative sample rate value.
To overcome this and to allow some more variables affecting the stream being passed to the input stream droid source is modified to reconfigure itself with the source-output that connects to it. This means, that just looking at inactive source from "pactl list" listing doesn't tell the whole story.
Droid source is always reconfigured with the last source-output that connects to it, possibly already connected source-outputs will continue to read from the source but through resampler.
For example,
1) source-output 44100Hz, stereo connects (so1) 1) source is configured with 44100Hz, stereo 2) so1 connects to the source without resampler 2) source-output 16000Hz, mono connects (so2) 1) so1 is detached from the source 2) source is configured with 16000Hz, mono 3) so2 connects to the source without resampler 4) resampler is created for so1, 16000Hz, mono -> 44100Hz stereo 5) so1 is re-attached to the source through resampler 3) source-output 16000Hz, mono connects (so3) 1) so1 and so2 are detached from the source 2) so3 connects to the source without resampler 3) so1 is re-attached to the source through resampler 4) so2 is attached to the source
Certain property values are set to all active sinks and sources based on their functionality to ease device classification.
Currently following properties are set:
If the property is set and with value "true", the sink or source should be used for the property type. If the property is not defined or contains value "false" it shouldn't be used for the property type.
For example, we might have sink.primary and sink.low_latency with following properties:
There also may be just one sink, with all the properties defined as "true" and so on.
Right now there exists only one source (input device) which will always have both properties as true.
There are some adaptations that require hacks to get things working. These hacks can be enabled or disabled with module argument "options". Some options are enabled by default with some adaptations etc. There are also some more generic options.
Currently there are following options:
Options can be enabled or disabled normally as module arguments, for example:
load-module module-droid-card hw_volume=false record_voice_16k=trueWhen voicecall virtual profile is enabled, active droid-sink is internally switched to voicecall volume control mode. What this means is changing the sink volume or volume of normal streams connected to the sink do not change active voicecall volume. Special stream is needed to control the voicecall volume level. By default this stream is identified by stream property media.role, with value "phone". This can be changed by providing module arguments voice_property_key and voice_property_value to module-droid-card.
Usually droid HAL has 6 volume levels for voicecall.
It is possible to add temporary route to sink audio routing with specific stream property. When stream with property key droid.device.additional-route connects to droid-sink, this extra route is set (if possible) as the enabled route for the duration of the stream.
For example, if droid-sink has active port output-wired_headphone:
paplay --property=droid.device.additional-route=AUDIO_DEVICE_OUT_SPEAKER a.wavAs long as the new stream is connected to droid-sink, output routing is SPEAKER.
If there is need to call HAL directly from other modules it can be done with function pointer API stored in PulseAudio shared map.
Once the function pointers are acquired when called they will work the same way as defined in Android audio.h. For example:
void *handle;
int (*set_parameters)(void *handle, const char *key_value_pairs);
char* (*get_parameters)(void *handle, const char *keys);
handle = pa_shared_get(core, "droid.handle.v1");
set_parameters = pa_shared_get(core, "droid.set_parameters.v1");
get_parameters = pa_shared_get(core, "droid.get_parameters.v1");
set_parameters(handle, "route=2;");
char *value = get_parameters(handle, "connected");