Audio

From postmarketOS
Note If you're trying out anything and learn something that would have helped, please extend this wiki page.

Audio on Android

Mixer configuration

This provides a brief overview of how Android audio HALs configure the ALSA mixer.

Android devices can have an assortment mixer configurations such as headphones, speakers, or microphones. For a particular device the audio route configurations are defined in mixer_paths.xml. For an example see the mixer path XML for Nexus 5 (hammerhead).

Android audio HALs make use of tinyalsa and audio_route. Tinyalsa is a small standalone library to interface with ALSA in the Linux kernel. Audio HALs call the audio_route library which loads audio routes from an XML and controls the mixer through tinyalsa.

Format of mixer_paths.xml

  • ALSA controls are defined by ctl elements.
  • Audio routes are defined by path elements. A path element contains a collection of ctl elements as well as other path elements.
  • The path name attribute is used to select an audio route.

During initialisation, all the mixer controls specified at the top of the file will be applied. To apply a particular path (e.g. "speaker") the controls in that path will be applied, switching to say, "headphones", involved resetting the controls back to their values before the speaker path was applied, in reverse order, and then applying the new path.

Headphone jack

Android kernels use switch_class kernel driver to notify the Android frameworks of headphone jack events. The driver source code can be located at: drivers/switch/switch_class.c or drivers/staging/android/switch/switch_class.c. Note that the switch class driver is an Android specific driver. It has not been upstreamed to the mainline kernel.

In userspace the WiredAccessoryManager monitors the /sys/class/switch/h2w for uevents when the headphone is (un)plugged.

Udev rules can be used to detect if headphones are plugged in, example for this is from xiaomi-santoni

Getting Audio to work on non-Android userspace

Bring up Android DSP subsystem

Note This is most likely for downstream kernels, if you're using a mainline kernel, this is pointless.
Note This was only tested on Qualcomm MSM8937, this might also work on other chipset too, but this is untested.

ADSP needs to be powered on before audio could work on downstream kernels. To do that, you'll have to supply firmware files that is available on your internal MMC.

On a shell prompt inside postmarketOS, type fdisk -l /dev/mmcblk0 (replace mmcblk0 with the correct device name for internal MMC). If that doesn't display the partition labels, you can also try ls -l /dev/disk/by-partlabel or ls -l /dev/block/platform/*/by-name/. At this point you should get a familiar partition layout of a Android device.

Now, the only thing that is important is the partition with MODEM label on it, which on my device it's mmcblk0p1. Mount the partition as read-only to /mnt using:

# mount -o ro /dev/mmcblk0p1 /mnt

Inside /mnt, there is a folder named image which has a bunch of firmware files for different parts of the device, like wcnss (for wlan), DRM contents and so on, the only thing that needs to be extracted is all the files that named adsp.*:

// Create /lib/firmware/postmarketos if it doesn't exist
# mkdir -p /lib/firmware/postmarketos

// Copy all adsp.* to the postmarketOS firmware directory
# cp -v /mnt/image/adsp.* /lib/firmware/postmarketos

After that, you should be able to bring up the ADSP subsystem by running:

# echo 1 > /dev/subsys_adsp

You should check dmesg and /var/log/firmwareload.log to make sure there aren't firmwares failing to load.

If it goes fine, you should see a bunch of playback devices by typing aplay -l

You also want to install adsp-audio from pmOS repo to load the subsystem on automatically on bootup.

Convert XML audio to UCM

Note This section is incomplete. It requires a reference to more complete documentation of the device config XML format.

The audio routes in Linux userspace can be defined in ALSA UCM configuration files. There is a tool called xml2ucm which will convert the Android mixer_paths.xml and a device config XML to UCM files. The xml2ucm tool introduces the device config XML which provides necessary routing information that was hardcoded in the audio HAL on Android.

The mixer_path.xml is commonly found in the Android device repository. (e.g. android_device_lge_hammerhead). You must write the device config XML for your device.

Format of device config XML

A device config XML describes a sound card with use cases and sets of audio routes. In the examples available at the moment there is only one use case named HiFi. Under each <use-case> there are multiple <device> elements. In this context a <device> means audio route(s) which comprise a logical audio device (e.g. Headphone, Headset, Speaker).

Android audio HALs can and do select multiple audio routes at once. For example a headset can have a headphone output route and a microphone input route. To handle this a <device> element (the headset) can contain multiple <path> elements (audio routes). Example device config XML files are provided in the xml2ucm repository.

The ctl-device must be the name of the sound card which can be obtained using aplay. In the following example from the Samsung p4wifi the ctl-device is the sound card named wm8994aif1.

$ aplay -L
null
    Discard all samples (playback) or generate zero samples (capture)
default:CARD=wm8994aif1
    wm8994-aif1, 
    Default Audio Device
sysdefault:CARD=wm8994aif1
    wm8994-aif1, 
    Default Audio Device

Once the mixer_path.xml and device config XML are converted to UCM format then an audio route can be set by using alsaucm. Example:

$ alsaucm -c wm8994aif1 set _verb HiFi set _enadev Headphones

libhybris

Note Might not work on some devices, which in this case providing logs to the maintainer is required, and debugging is also needed too.

In general in Hybris environment just install pulseaudio-modules-droid package and sound will work.

Make sure you have the following line in your /etc/pulse/default.pa:

load-module module-droid-card

Loading droid-card module is enough, other modules will be loaded automatically. Also it would not hurt to comment out module-udev-detect, module-detect lines to lower the amount of error messages in pulseauduio log (these modules don't work anyway), like that:

### Automatically load driver modules depending on the hardware available
#.ifexists module-udev-detect.so
# load-module module-udev-detect
#.else
### Use the static hardware detection module (for systems that lack udev support)
# load-module module-detect
#.endif

Start pulseaudio in verbose mode (/usr/bin/pulseaudio --log-target=stderr --log-level=debug -v) to make sure you can get to see if module-droid-card loads successfully. If it does then you should be able to play sound files using paplay from pulseaudio-utils package.

If you hear sounds coming from it, it works perfectly fine. If not, check dmesg, pulseaudio log and logcat.

You might also want to install alsa-plugins-pulse for ALSA backend on PulseAudio.

Audio cutting out when multiple streams start playing

See here

alsactl method

Cascardo reported having success with the following method on Debian, this should work for postmarketOS as well:

I ran a debian chroot on a cyanogenmod, set the volume up, then ran alsactl save on the chroot. booted into debian, then after alsactl restore was run, I had audio working

so, in summary, it's hard to know which mixer settings will make it work, so my strategy was running some android stack that correctly set the mixer, then saving the state with a chroot or any other tool that would allow you to save, copy and restore that state later on a gnu stack

This method may work for non-Android devices as well, as long as you see an audio device in alsamixer. Just boot the normal OS, save the volume settings somehow, then restore them.

amixer method

You can also skip the chroot step and run amixer on Android directly; then, you'll have to copy over the other settings manually. This is a bit more time-intensive, but should work well for most devices.

  • Root your device
  • Get a copy of amixer that runs under Android (I got mine by installing this Magisk module; sadly it installed without execute permissions, I had to move /system/xbin/alsa_amixer to /data (/sdcard wouldn't let me make it executable) and run it from there)
  • Get into a terminal and run amixer > /sdcard/amixer-output
    • You will have do this twice - once with a headset plugged in, and once without; save them as separate files
  • Copy the files onto your computer
  • On your mainline device, run amixer, then diff the output with the downstream output
    • You might want to get the diff using diff -u, then remove the "Capability" lines using grep -v "Capability" \<path-to-diff\> since it's fine if they're different
  • Open alsamixer on the Linux device and tweak the values until they match
    • Repeat diffing process as many times as necessary
  • Then save the output of amixer again, or run alsactl store

strace method

It is possible to get audio working by stracing Android's audioserver and replicating the mixer configuration it performs. See here.

The positive side of such porting is that the amount of manual work is reduced to the minimum, another positive thing is that it may also work for non-Android proprietary userspace.

Spreadtrum

Audio support for Spreadtrum SoCs is provided in soc-sprd-audio and its subpackages. To get audio support for the device, add the package related to the audio device to dependencies of device package or install it manually, like

apk add soc-sprd-audio-sc8830
Note SC8830 audio device is available for a huge range of SoCs. Check contents of Android's tiny_hw.xml to find which device is used.

WM1811

This audio chip is present in several Samsung devices that are based on the Exynos 4412 platform. There is a configuration for the Samsung Galaxy Note 10.1 available at the opensgn project. Have a look at the files in opensgn/data/n8000/usr/share/alsa/ucm/Midas_WM1811/. In opensgn/data/n8000/etc/rc.local there are the commands to use the configuration:

# restore alsa sinitial tate
/usr/sbin/alsactl restore -f /usr/share/alsa/ucm/Midas_WM1811/init.state

# enable speaker for output and builtin mic for input
/usr/bin/alsaucm -c Midas_WM1811 set _verb HiFi set _enadev Speaker
/usr/bin/alsaucm -c Midas_WM1811 set _verb HiFi set _enadev Mic

For other devices the configuration might need to be adapted. On the Samsung Galaxy Note 8.0 for example, the audio device is named "Kona_WM1811". Replacing all occurrences in file names and file contents of Midas by Kona worked.

In case the existing UCM configs don't work, it's possible to get the ALSA mixer settings by running amixer on Android. This also works with the mainline kernel, as the settings there are nearly identical.

In mainline

The WM1811 is supported in mainline, as part of the Galaxy S3 (midas) support. Portions of the midas code can be reused between other Exynos devices using the same SoC and sound chip.

Mainline

Troubleshooting audio input

A nice way to troubleshoot audio input is to run arecord in a terminal to get realtime audio levels of the default alsa input:

$ arecord -vv --format=dat /dev/null
Recording WAVE '/dev/null' : Signed 16 bit Little Endian, Rate 48000 Hz, Stereo
ALSA <-> PulseAudio PCM I/O Plugin
Its setup is:
  stream       : CAPTURE
  access       : RW_INTERLEAVED
  format       : S16_LE
  subformat    : STD
  channels     : 2
  rate         : 48000
  exact rate   : 48000 (48000/1)
  msbits       : 16
  buffer_size  : 24000
  period_size  : 6000
  period_time  : 125000
  tstamp_mode  : NONE
  tstamp_type  : GETTIMEOFDAY
  period_step  : 1
  avail_min    : 6000
  period_event : 0
  start_threshold  : 1
  stop_threshold   : 24000
  silence_threshold: 0
  silence_size : 0
  boundary     : 6755399441055744000
#+                                                 | 00%

See also