Ad Terras Per Aspera

Apparently there is no dead simple way to send audio from one computer to another in a low(er) latency way.

Can’t beat this, works for any ALSA app that you can change the output for (or just change your default in .asoundrc).

On source computer:
modprobe snd-aloop
arecord -f cd -D hw:Loopback,1,0 | netcat dest 1234
mplayer -ao alsa:device=hw=Loopback.0.0 something.mp3

On destination computer:
netcat -k -l -p 1234 | aplay

Well, it seems and the Linux kernel can’t bridge dissimilar network types, which means I can’t bridge Ethernet and Firewire (workaround until I replace my NIC in my desktop because it fried, having my laptop route for my desktop).

My laptop’s IP is 192.168.2.4
My desktop’s IP is 192.168.2.2
My router’s IP is 192.168.2.1

So, the work around seems to be this…

On the laptop, with eth0 already up:
ifconfig firewire0 up 192.168.2.4
route del -net 192.168.2.0 netmask 255.255.255.0 dev firewire0
route add -host 192.168.2.2 dev firewire0
iptables -F
iptables -P FORWARD ACCEPT
echo 1 > /proc/sys/net/ipv4/conf/all/forwarding


On the desktop:
ifconfig firewire0 up 192.168.2.2
route add default gw 192.168.2.1

To make this permanent, you would edit /etc/networking/interfaces like this…

On the laptop:
auto eth0
iface eth0 inet static
address 192.168.2.4
netmask 255.255.255.0
broadcast 192.168.2.255
gateway 192.168.2.1
post-up ifconfig firewire0 down
post-up ifconfig firewire0 up 192.168.2.4
post-up route del -net 192.168.2.0 netmask 255.255.255.0 dev firewire0
post-up route add -host 192.168.2.2 dev firewire0
post-up iptables -F
post-up iptables -P FORWARD ACCEPT
post-up echo 1 > /proc/sys/net/ipv4/conf/all/forwarding

On the desktop:
auto eth0
iface firewire0 inet static
address 192.168.2.2
netmask 255.255.255.0
broadcast 192.168.2.255
gateway 192.168.2.1


I’m not sure how many people this bug effects, but with certain setups Catalyst 10.6 will segfault on load. This effects very few people.

The fix is to edit /etc/X11/xorg.conf and add BusID "PCI:1:0:0" inside the Section "Device" that defines your card (almost all people only have one of these)

The BusID should match the PCI ID that lspci | grep Radeon says.

Using the patches detailed here, I have produced a fglrx.tar.bz2 that builds.

Just drop it in /usr/src/ (replacing the old one) and module-assistant update,build,install fglrx like usual. Patched from Debian package version 1:9-5-1.

Update: fglrx 9.6 is now packaged in sid, this should contain the updates and work on 2.6.30.

Out there, on the Internet, there is a lot of conflicting information on how to properly build a UML image on Debian. These instructions are for Debian Sid, but should work on Etch and Sarge as well. The instructions are partially based on the ones available here.

First, you want to apt-get install user-mode-linux uml-utilities bridge-utils debootstrap realpath. Then, become root either by su or sudo bash. Make an empty directory (such as ~/uml) and cd into it. Now, follow this psudeo script:

# setup environment, feel free to change DEBIAN_MIRROR to your closest mirrror, IP to an IP not already being used on your network, and GATEWAY to your LAN’s gateway (not your host machine)
export TMPDIR=`pwd`"/tmp"
export IMAGE="root_fs"
export DEBIAN_MIRROR="ftp://ftp.debian.org/debian/"
export HOSTNAME="NameThisMachine"
export IP="a.ip.to.use"
export GATEWAY="ip.to.talk.to"
mkdir $TMPDIR

# produce a 1GB image, fudge with seek to change size
dd if=/dev/zero of=$IMAGE bs=1 count=1 seek=1G

# make partition, and mount it; Debian Sid kernels can only mount ext2, ext3, cramfs, iso9660, and reiserfs, ext3 is the only one worth using
mkfs.ext3 $IMAGE
mount -o loop $IMAGE $TMPDIR

# download and install Debian into your image: change arch to the arch you’re using, and change sid to sid, etch, or sarge; this step takes awhile
debootstrap --arch i386 sid $TMPDIR $DEBIAN_MIRROR

# install a correct fstab, proc and sys are automatically loaded on boot
echo "/dev/ubd0 / ext3 defaults 0 0" > $TMPDIR/etc/fstab
echo $HOSTNAME > $TMPDIR/etc/hostname

# tweak inittab
cp $TMPDIR/etc/inittab $TMPDIR/etc/inittab.save
grep -v "getty" $TMPDIR/etc/inittab.save > $TMPDIR/etc/inittab
echo "# We launch just one console for UML:" >> $TMPDIR/etc/inittab
echo "c0:1235:respawn:/sbin/getty 38400 tty0 linux" >> $TMPDIR/etc/inittab
echo "# UML modification: use tty0 or vc/0" >> $TMPDIR/etc/securetty
echo "tty0" >> $TMPDIR/etc/securetty
echo "vc/0" >> $TMPDIR/etc/securetty

# add networking stuff
echo "auto lo" >> $TMPDIR/etc/network/interfaces
echo "iface lo inet loopback" >> $TMPDIR/etc/network/interfaces
echo "" >> $TMPDIR/etc/network/interfaces
echo "auto eth0" >> $TMPDIR/etc/network/interfaces
echo "iface eth0 inet static" >> $TMPDIR/etc/network/interfaces
echo "address $IP" >> $TMPDIR/etc/network/interfaces
echo "netmask 255.255.255.0" >> $TMPDIR/etc/network/interfaces
echo "gateway $GATEWAY" >> $TMPDIR/etc/network/interfaces

# clean up
umount $TMPDIR
rmdir $TMPDIR

You now have an image full of Debian named root_fs. To run, simply execute linux mem=64M ubd0s=root_fs. If you need more memory, increase the mem paramater. By default, there is no password for root, so just use root as your login and press enter when it asks for password.

A little aside on networking:the auto tuntap method (eth0=tuntap,,,the.ip.to.use, which basically does ARP proxying on the host machine) does not work properly as it only allows host->UML traffic. The only way I’ve been able to get UML networking to work properly is to build a full bridge setup. The following commands in /etc/network/interfaces/ sets up the bridge. You probably have something like:

auto eth0
iface eth0 inet dhcp


Comment that out with #s, and change it to:

auto br0
iface br0 inet dhcp
pre-up tunctl -u user-to-run-uml -t tap0
pre-up ifconfig eth0 0.0.0.0 promisc up
pre-up ifconfig tap0 0.0.0.0 promisc up
pre-up brctl addbr br0
pre-up brctl stp br0 off
pre-up brctl setfd br0 1
pre-up brctl sethello br0 1
pre-up brctl addif br0 eth0
pre-up brctl addif br0 tap0
post-down tunctl -d tap0


allow-hotplug eth0
iface eth0 inet manual
pre-up ifconfig eth0 0.0.0.0 promisc up
pre-up brctl addif br0 eth0
pre-down brctl delif br0 eth0
pre-down ifconfig eth0 down


Add the additional parameter to linux‘s command line: eth0=tuntap,tap0. That will make UML connect to tap0, and use the IP set above when you built the image. If you use a firewall, you need to change the interface to firewall from eth0 to br0, and to allow br0->br0 traffic.