TerminalMage dot NET

USB audio peripherals still don’t “just work” in Linux, but they’re starting to get a little easier to use.

The last time I tried using PulseAudio, a little over two years ago, I experienced no end of frustration. This probably had as much to do with a lack of understanding as anything else, but regardless, I have been using nothing but ALSA since then. However, ALSA’s USB support is not very good. Changing the default audio device requires an application to be restarted to recognize the new default device. So, I figured I’d give PulseAudio another try and see if I could make my USB headset “just work” this time around. To my surprise, it was quite easy to do so, and in this post I will detail the steps I took. I did this on Arch Linux, but the steps should apply in most Linux distributions.

Before we start, we need to ensure that PulseAudio is A) installed, and B) running. Most distributions will take care of both of these for you, but due to the do-it-yourself nature of Arch Linux, in my case it needed to be installed:

# pacman -S pulseaudio pulseaudio-alsa pavucontrol

Notice that I also installed two other packages, pulseaudio-alsa and pavucontrol. The former contains the ALSA plugin for PulseAudio, the latter is a helpful application for configuring PulseAudio and visualizing your PulseAudio configuration. Most Linux distributions will include the ALSA plugin (and may name the package differently), but you may need to manually install pavucontrol.

So, now that I have PulseAudio installed, it needs to be running. If you are using a desktop environment like KDE, Gnome, etc., PulseAudio should be automatically started when you login. To see if PulseAudio is running, check for it in the process table.

$ ps aux | grep pulseaudio | grep -q start && echo yes || echo no
yes

OK, PulseAudio is both installed and running. Open pavucontrol and click on the Output Devices tab. In the upper-right corner of each device will be a few buttons. The green one with a check mark on it (outlined in red in the example to the right, click for a full-size image) shows which device is the default output device. The Input Devices tab contains its own default as well, but for now we will focus on the output devices, for the sake of simplicity.

Start playing something, preferably in a PulseAudio-compatible application. For this example, I started a movie file in mplayer from the command-line, forcing PulseAudio output using the -ao (audio output) option:

$ mplayer -ao pulse foo.mp4

With something now generating audio output, plug in the USB headset, and you’ll notice a new device show up in the Output Devices tab. But you’ll also notice that the audio isn’t going to the headset yet.

This can be fixed by clicking the green button next to the headset device, but this would have to be repeated every time the headset is plugged in, which would be an inconvenience to say the very least.

The solution to this is to write a udev rule which will detect when this device is plugged in, and then run a script to set the headset as the default input and output device. I’ll explain how to do this in a bit, but first we need to understand what it is that we want to do with this script.

PulseAudio has a command-line tool called pacmd which allows you to view and change the PulseAudio configuration. Plug in the USB headset, and then run pacmd dump to view the current configuration. There is quite a bit of output, so I have cut out all but the relevant parts:

$ pacmd dump
Welcome to PulseAudio! Use "help" for usage information.
>>> ### Configuration dump generated at Wed Nov 16 23:04:59 2011

...

suspend-sink alsa_output.usb-Logitech_Logitech_USB_Headset-00-Headset.analog-stereo no

...

suspend-source alsa_input.usb-Logitech_Logitech_USB_Headset-00-Headset.analog-mono no

...

set-default-sink alsa_output.pci-0000_00_1b.0.analog-stereo
set-default-source alsa_input.pci-0000_00_1b.0.analog-stereo

### EOF
>>>

PulseAudio calls output devices “sinks”, and input devices “sources”. We can see from this output that the default sink and source belong to the onboard audio card. To set the headset as the default device, we can use pacmd to set the default sink and source to the ones for the headset. This would be done with the following two commands:

$ pacmd set-default-sink alsa_output.usb-Logitech_Logitech_USB_Headset-00-Headset.analog-stereo
$ pacmd set-default-source alsa_input.usb-Logitech_Logitech_USB_Headset-00-Headset.analog-mono

Try runnng these commands yourself (replacing the sink/source names with the ones you observed when running pacmd dump), and you will notice that the headset is now the default input and output device if you check the pavucontrol window.

Now that we know how to set the headset as the default device via the command line, we can write a script that will do this for us. However, because this script will eventually be run as root by udev, we need to use the su command to run it as the user. The following script will check the process table and then, for each user who is running PulseAudio, will run the two commands needed to set the headset as the default input and output device.

#!/bin/bash
#
# Name:        usb-headset-set-default
# Description: invokes pacmd to set my Logitech usb headset as the default
#              audio input and output device.
#

# Sleep a little to allow PulseAudio to notice the headset
sleep 1

# Check process table for users running PulseAudio
#
# ejohnson@tardis:~%  ps axc -o user,command | grep pulseaudio
# ejohnson pulseaudio
for user in `ps axc -o user,command | grep pulseaudio | cut -f1 -d' ' | sort | uniq`;
do
	su $user -c "pacmd set-default-sink alsa_output.usb-Logitech_Logitech_USB_Headset-00-Headset.analog-stereo >/dev/null 2>&1"
	su $user -c "pacmd set-default-source alsa_input.usb-Logitech_Logitech_USB_Headset-00-Headset.analog-mono >/dev/null 2>&1"
	#su $user -c "pacmd dump >>/tmp/debug.log"
done

#echo matched at `date` >>/tmp/debug.log

Save this script to /usr/local/bin/usb-headset-set-default. It should be executable, and owned by root. Of course, you should also make sure to replace the sink and source in the script with the ones you observed when you ran pacmd dump, if they differ. Unplug the headset and plug it back in, then run the script as root, and in pavucontrol you should see that it has been set as the default input and output device. Note that there are a couple commands commented out in the script. These can be uncommented for debugging later on, if needed.

Next, we need a way to detect when the headset is plugged in. We can do this using udev.

Udev is what handles setting up device nodes for Linux. When you insert a USB flash drive, and it gets assigned a device name (like /dev/sdb), udev is what takes care of allocating this device name. Plugging in the headset will generate a number of device nodes. We can use udevadm to see each of them. Unplug the USB headset, then run the command below. After running it, plug in the USB headset and you should see some output appear.

# udevadm monitor --environment | fgrep 'DEVNAME=/dev'
DEVNAME=/dev/bus/usb/002/118
DEVNAME=/dev/snd/pcmC1D0c
DEVNAME=/dev/hidraw0
DEVNAME=/dev/snd/pcmC1D0p
DEVNAME=/dev/input/event9
DEVNAME=/dev/mixer1
DEVNAME=/dev/dsp1
DEVNAME=/dev/audio1
DEVNAME=/dev/snd/controlC1

Note: udevadm must be run as root.

Using another udevadm command, we can find the udev attributes for one of these device nodes. I have chosen /dev/audio1.

# udevadm info --attribute-walk --name /dev/audio1

Udevadm info starts with the device specified by the devpath and then
walks up the chain of parent devices. It prints for every device
found, all possible attributes in the udev rules key format.
A rule to match, can be composed by the attributes of the device
and the attributes from one single parent device.

  looking at device '/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1:1.0/sound/card1/audio1':
    KERNEL=="audio1"
    SUBSYSTEM=="sound"
    DRIVER==""

  looking at parent device '/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1:1.0/sound/card1':
    KERNELS=="card1"
    SUBSYSTEMS=="sound"
    DRIVERS==""
    ATTRS{id}=="Headset"
    ATTRS{number}=="1"

  looking at parent device '/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1:1.0':
    KERNELS=="2-1:1.0"
    SUBSYSTEMS=="usb"
    DRIVERS=="snd-usb-audio"
    ATTRS{bInterfaceNumber}=="00"
    ATTRS{bAlternateSetting}==" 0"
    ATTRS{bNumEndpoints}=="00"
    ATTRS{bInterfaceClass}=="01"
    ATTRS{bInterfaceSubClass}=="01"
    ATTRS{bInterfaceProtocol}=="00"
    ATTRS{supports_autosuspend}=="1"

  looking at parent device '/devices/pci0000:00/0000:00:1d.0/usb2/2-1':
    KERNELS=="2-1"
    SUBSYSTEMS=="usb"
    DRIVERS=="usb"
    ATTRS{configuration}=="G8 v3.0.0.0"
    ATTRS{bNumInterfaces}==" 4"
    ATTRS{bConfigurationValue}=="1"
    ATTRS{bmAttributes}=="80"
    ATTRS{bMaxPower}=="100mA"
    ATTRS{urbnum}=="663416"
    ATTRS{idVendor}=="046d"
    ATTRS{idProduct}=="0a0b"
    ATTRS{bcdDevice}=="1013"
    ATTRS{bDeviceClass}=="00"
    ATTRS{bDeviceSubClass}=="00"
    ATTRS{bDeviceProtocol}=="00"
    ATTRS{bNumConfigurations}=="1"
    ATTRS{bMaxPacketSize0}=="8"
    ATTRS{speed}=="12"
    ATTRS{busnum}=="2"
    ATTRS{devnum}=="118"
    ATTRS{devpath}=="1"
    ATTRS{version}==" 2.00"
    ATTRS{maxchild}=="0"
    ATTRS{quirks}=="0x0"
    ATTRS{avoid_reset_quirk}=="0"
    ATTRS{authorized}=="1"
    ATTRS{manufacturer}=="Logitech"
    ATTRS{product}=="Logitech USB Headset"

  looking at parent device '/devices/pci0000:00/0000:00:1d.0/usb2':
    KERNELS=="usb2"
    SUBSYSTEMS=="usb"
    DRIVERS=="usb"
    ATTRS{configuration}==""
    ATTRS{bNumInterfaces}==" 1"
    ATTRS{bConfigurationValue}=="1"
    ATTRS{bmAttributes}=="e0"
    ATTRS{bMaxPower}=="  0mA"
    ATTRS{urbnum}=="2640"
    ATTRS{idVendor}=="1d6b"
    ATTRS{idProduct}=="0001"
    ATTRS{bcdDevice}=="0301"
    ATTRS{bDeviceClass}=="09"
    ATTRS{bDeviceSubClass}=="00"
    ATTRS{bDeviceProtocol}=="00"
    ATTRS{bNumConfigurations}=="1"
    ATTRS{bMaxPacketSize0}=="64"
    ATTRS{speed}=="12"
    ATTRS{busnum}=="2"
    ATTRS{devnum}=="1"
    ATTRS{devpath}=="0"
    ATTRS{version}==" 1.10"
    ATTRS{maxchild}=="2"
    ATTRS{quirks}=="0x0"
    ATTRS{avoid_reset_quirk}=="0"
    ATTRS{authorized}=="1"
    ATTRS{manufacturer}=="Linux 3.1.0-4-ARCH uhci_hcd"
    ATTRS{product}=="UHCI Host Controller"
    ATTRS{serial}=="0000:00:1d.0"
    ATTRS{authorized_default}=="1"

  looking at parent device '/devices/pci0000:00/0000:00:1d.0':
    KERNELS=="0000:00:1d.0"
    SUBSYSTEMS=="pci"
    DRIVERS=="uhci_hcd"
    ATTRS{vendor}=="0x8086"
    ATTRS{device}=="0x27c8"
    ATTRS{subsystem_vendor}=="0x1028"
    ATTRS{subsystem_device}=="0x02f4"
    ATTRS{class}=="0x0c0300"
    ATTRS{irq}=="23"
    ATTRS{local_cpus}=="ff"
    ATTRS{local_cpulist}=="0-7"
    ATTRS{dma_mask_bits}=="32"
    ATTRS{consistent_dma_mask_bits}=="32"
    ATTRS{enable}=="11"
    ATTRS{broken_parity_status}=="0"
    ATTRS{msi_bus}==""

  looking at parent device '/devices/pci0000:00':
    KERNELS=="pci0000:00"
    SUBSYSTEMS==""
    DRIVERS==""

Wow, that’s a lot of output. Before proceeding to write a udev rule, we need to know how to utilize the attributes above to come up with a unique rule. This page contains an excellent walkthrough on how to write a udev rule, so go ahead and read it before continuing. The information about the udev commands is a little outdated (for instance, most of the commands referenced are now part of udevadm), but the information on rule syntax is very good.

So, we need to come up with a set of attributes that will match, but we also would like to find a set of attributes that are unique. Multiple matches will result in the script being run multiple times. In this case, where we’re just running a couple commands to set the default audio devices, there aren’t really any consequences to running the script multiple times, but it’s still a good idea to come up with a unique set of attributes. The rule I came up with was:

# Set the USB headset as default sink/source when it is plugged in
KERNEL=="audio?", SUBSYSTEM=="sound", SUBSYSTEMS=="usb", ATTRS{idVendor}=="046d", ATTRS{idProduct}=="0a0b", ACTION=="add", RUN+="/usr/local/bin/run-script-in-background /usr/local/bin/usb-headset-set-default"

I saved this rule to /etc/udev/rules.d/85-usb-headset.rules. Notice that the RUN parameter is set to /usr/local/bin/run-script-in-background /usr/local/bin/usb-headset-set-default. In the process of debugging this rule, I noticed that the script was running, but the headset wasn’t being set as the default sink/source. This is where the commented-out debugging lines come in. With the debugging commands uncommented, I noticed in the log file that the headset wasn’t showing up in the pacmd dump output. It turns out that PulseAudio doesn’t know anything about the headset until udev completes setting up all the device nodes. This makes perfect sense. However, if udev has matched the rule and is running the script, this suspends further udev processing until the script completes. So, what will happen is that the script tries to set devices that don’t yet exist, and in effect nothing happens. To get around this, we need a way for the command executed by udev to exit quickly, but allow for a short delay so that udev can finish doing it’s thing before pacmd is invoked. So, I created the following script and saved it to /usr/local/bin/run-script-in-background:

#!/bin/bash
#
# Name:        run-script-in-background
# Description: takes all command-line arguments and runs them as a background command
#

$@ &

This script simply takes the arguments fed to it and runs them as a separate process, in the background. This allows run-script-in-background to quickly exit so that udev can finish.

With the rule and scripts now in place, plug in the USB headset again and you should notice in pavucontrol that the headset is now the default device. If this is not the case, then use the debugging lines in the usb-headset-set-default script while tailing the log file to determine the problem. If the headset doesn’t show up in the pacmd dump output, then you might need to increase the sleep time at the beginning of the script.

I’ve been using this udev rule and script for several days now and the headset is working flawlessly. Hopefully, my instructions can help you get it working for yourself. All in all, I am very impressed with how well PulseAudio is working for me, given the nightmares I experienced in the past.

For several years, I have been using MediaTomb as a streaming media server. Functionally, MediaTomb works fantastically, but a few things about it have bugged me for some time:

1. The arrangement of media files. MediaTomb creates two hierarchies of file listings, called PC Directory and Folders. The former lets you browse from the root of your filesystem, and is a bit useless when the directory you are sharing is several levels down. Browsing through those levels each time you want to watch something is annoying. So I prefer the Folders listing, which places your shared folders at the top level and lets you drill down to subdirectories from there. Here’s where we get to the problem: While timed folder rescans and inotify allow you to add new files to the media browser simply by copying/moving those files to a watched directory, new files added in this fashion will show up in their own top-level directory under the Folders view, rather than staying organized. The only way around this is to use the web interface to delete the watched folder and re-add it, forcing a full re-scan of the media directory. It’s hard to tell whether this is a MediaTomb issue or a problem with the PS3′s DLNA client.
2. Configuration Syntax. The configuration file for MediaTomb is in XML format. I have never been a fan of XML syntax in a config file.
3. Streaming to a PlayStation 3 is not supported “out-of-box”. It’s simple enough to enable (in fact the option is present in the default config file, commented out), but this is an annoyance nonetheless.
4. MediaTomb is not actively maintained. Releases are few and far between, the last of them coming over a year and a half ago.

That is not to say that that MediaTomb is without redeeming features. For one, it allows you to write scripts to transcode media types which it cannot handle. You won’t be able to seek or pause when viewing a transcoded file, but this is actually not a MediaTomb limitation. The very nature of transcoded media is that it is being converted on-the-fly, so seeking just isn’t possible. Another nice feature of MediaTomb is that its folder layout is scriptable, allowing you to control how the media files are presented in the file browser. Perhaps this would allow me to come up with a solution to my file arrangement problem, but I just don’t have the time or inclination to try to fix it.

A few days ago, while browsing Reddit, I saw a recommendation to use MiniDLNA to stream media. It has turned out to be a very simple to use alternative, and I have now officially switched to it from MediaTomb.

MiniDLNA is written and maintained by a Netgear employee and is used in their ReadyNAS line of storage devices. It is both actively maintained and easy to use.

MiniDLNA is not as feature-rich as MediaTomb. It is not scriptable, and it cannot transcode media. But what you sacrifice in features, you gain in sheer ease of use. You need only set one configuration variable in /etc/mediatomb.conf to get it up and running:

media_dir=/path/to/media

Additional media_dir lines can be added to share other directories. Streaming to the PS3 is supported without any special configuration needing to be done.

And that’s it! There are, of course, other things you should set up. First, you can change the server name shown to clients by setting the friendly_name value. Also, while not required, enabling MiniDLNA’s cache (which will keep it from needing to re-scan everything when it is restarted), as well as logging, are both good ideas. You’ll need to set up directories for both, which are writable by the nobody user (the user under which MiniDLNA runs). This can be accomplished by running the following two commands as root:

mkdir /var/{cache,log}/minidlna
chown nobody:nobody /var/{cache,log}/minidlna

Then configure the following two parameters in the config file:

db_dir=/var/cache/minidlna
log_dir=/var/log/minidlna

MiniDLNA has proven to be perfect for me. I don’t do any transcoding or scripting with MediaTomb, so I am not missing those features. MiniDLNA supports inotify to add new files to the media browser automatically, and does a better job of organizing files added to the watched folders. They stay organized in the the proper folder, but sometimes new files appear at the bottom of that folder’s file list rather than in place alphabetically. However, this doesn’t seem to be a MiniDLNA bug, but rather a bug in the PS3′s DLNA client since the files appear in the correct order the next time the PS3 is started up, or after telling the PS3 to do a new search for media servers.

All in all, if you want a simple alternative to MediaTomb, then I highly recommend MiniDLNA.