 | Level: Introductory Lewin Edwards (sysadm@zws.com), Author, Freelance
10 May 2005 The Mac mini is an ideal low-cost, high-performance PowerPC development platform for numerous applications. Learn how to install and configure Linux on the mini. Future articles will add the software required to make it into a stand-alone multimedia appliance. This short series of articles shows you how to take a conveniently
inexpensive, high-end PowerPC® platform (specifically, an Apple Mac mini)
and build it into a home multimedia appliance using Linux™. At the end
of the series, you'll have a stand-alone device that can play slide shows
of images, audio, and movies, and that is controlled and administered
from another machine using a standard Web browser.
The PowerPC platform is very well-suited to this type of multimedia
application, and the G4 with AltiVec used in the Mac mini is an
exceptionally powerful and flexible choice. This first article introduces you to the hardware's capabilities and walks you through installing and configuring Yellow Dog Linux so you can delve into some application code in the next article.
Licenses, patents, and law
Before getting into the actual project details, however, I need to say a
few words about an important intellectual property issue. You need to be
aware that data compression, encryption, and other technologies related
to digital multimedia are a minefield of patents, unfriendly copyright
laws, and intellectual property squabbles. This series of articles, and
its accompanying source code, are educational and research materials, so I
can ignore many of the legal issues. If you're following along and
experimenting, for your own edification, with the materials I provide,
you can largely skip over the legal worries, too. However, if you plan to
distribute a product based on this or similar code, you have significant
homework to do. The source code accompanying these articles is licensed to you
under the GNU GPL; you can sell derivative works or give them away as
you please (as long as you comply with the terms of the GPL), and I don't
want a dime from you. However, I don't own, and hence can't license to
you, any patented algorithms this code happens to implement. You need to
do an audit of which patents might need to be licensed, and which
authority governs their licensing.
This warning is mostly applicable to movie playback. For example, there
are numerous open source MPEG-2 video playback programs. Standard DVD
discs use MPEG-2 encoding. However, if you start selling a device that
uses one of these open source player programs to play DVDs (even if you
ignore the thorny issues surrounding DVD encryption and only support
unprotected disks), you'll soon be receiving letters demanding license
fees for each unit sold. An interesting data point I read recently is
that the US$39.95 DVD players you commonly see at chain stores contain
almost US$20 of patent license fees. (This is one reason why China is
trying to promote the homegrown EVD standard in preference to DVD for
domestic video content; EVD was specifically designed to avoid Western
patent portfolios). Hidden legal bear traps like this are waiting almost
everywhere for the unwary. For example, Fraunhofer IIS requires a
(small) per-unit payment for every device sold containing an MP3
decoder. There is even a patent claim on an aspect of the common JPEG
still-image compression algorithm; it's widely regarded as meritless, to
date it has not been successfully litigated by the holder, and it does
expire fairly soon, but it's still something you might need to think
about.
One safe route, by the way: Build your application around a media player
that has already negotiated licenses for the appropriate patents. As far
as I know, the only company actively working on a patent-licensed media player
for Linux at the moment is Linspire (see Resources). They have a licensed
version of xine for sale; it's intended for use
with their particular flavor of
Linux, but when I've inquired about this in the past they have indicated
a willingness to issue serial numbers for use with any Linux variant,
for a fairly reasonable per-unit fee. The only problem in this
series is that their product is x86-only, though they might be persuaded
to build it for PowerPC if asked nicely.
Why the Mac mini?
If you're still reading after all that legal waffle, take a look at
the important attributes of the Mac mini. This system makes a good choice for
this project for the following reasons:
- Price. At US$499, the entry-level Mac mini is unquestionably the
best-value, high-end PowerPC evaluation platform money can buy.
It is about one quarter the cost of an equivalent evaluation
board from industrial single-board computer (SBC) vendors, AND
the price of the Mac mini includes a hard disk, RAM, optical
drive, and housing and power supply, all of which would be add-on
costs for the industrial SBC.
- Physical size. Even without tearing the electronics out of the
housing, the Mac mini is small and light.
- Orientation-neutral design. The optical drive is slot-loading
and is equally usable at any angle to the horizontal. This means
that you can simply glue it to the back of a monitor and presto!
-- instant multimedia appliance!
The following is a brief overview of the Mac mini hardware capabilities:
- PowerPC G4 microprocessor. The motherboard is jumper-configurable from 1.25 to 1.58GHz (the jumpers are soldered),
with the standard speeds from Apple being 1.25 or 1.42GHz. Find a link in Resources to a site with all the overclocking
details you might want, if that sort of thing appeals to you.
Overclocking is totally unnecessary for this application and I
don't recommend it. The Mac mini is already frisky enough for
everything you need to do.
- Built-in wired Ethernet
- One Firewire port
- Two USB 2.0 ports
- Internal (mono) speaker and stereo audio output. At the time of
writing this article, the specific audio chip used in the Mac
mini is not supported by the Linux kernel, but this appears to
be a very minor chip ID (revision level) issue. There is no audio
input connector.
- DVI video output. This also provides analog VGA signals through an
adapter shipped with the computer. (It can even composite video
and S-video signals through an optional adapter.)
- One internal IDE bus connected to a 2.5" IDE hard disk (either
40Gb or 80Gb according to model).
- A slot-loading laptop-style
optical drive; either a combo DVD/CD-RW drive or a
"Superdrive" (DVD+/-R, CD-RW).
- A proprietary internal mezzanine card holding an optional
802.11g wireless LAN adapter (unfortunately not supportable by
Linux due to lack of disclosure by Broadcom, the chipset
manufacturer) and/or a Bluetooth module.
- An internal 56Kbps modem. I don't have an analog phone line
(we're a cell phone and broadband family) so I don't know
anything about this peripheral.
Why Linux?
This series uses the Linux operating system for a couple of
reasons. The foremost reason is that it is a design goal for
this project to be entirely open source. This goal cannot be fulfilled
on top of Mac OS. The other primary reasons for selecting Linux are that
it is modularly configurable to a relatively fine granularity, and it
offers excellent API-level compatibility amongst different
platforms that are popular for high-end (32-bit and larger) embedded
cores: PPC, ARM, x86, and MIPS.
I experimented with a few distributions of Linux for PowerPC platforms. Debian (available in
Resources) weighs in at a svelte nine gigabyte download, or
fourteen CD images, although you don't actually need all of those
disks. I eventually chose Yellow Dog Linux 4.0.1 because the
standard disk set is a smaller download (four CD images), and, more
importantly, this distribution resembles Fedora very closely. I happen to have spent most of my Linux life working with Red
Hat-based products, so this was a comfort choice for me. I had no
compelling technical reason to choose Yellow Dog, and if you're more
comfortable with another distribution, feel free to use it as you follow
along with this series. The
application you'll build in this series will run happily on
practically any distribution.
Installing Yellow Dog
Linux
To install Yellow Dog, first download and burn the four ISO CD images.
Power up the Mac mini, insert CD #1 in the drive (quickly) and hold C to
force CD startup. You'll be taken to the yaboot startup screen, at which
you should simply type install to run the default graphical installer.
You can use this disk for emergency startup as well as installation, and
it has some alternative images on it also, such as a G5-optimized
version. You can get a list of available images by pressing Tab at the
boot: prompt, but for now let's move on with the installation.
When you press Return after typing install,
there will be a brief pause
while the kernel and initial RAM disk are loaded off the CD. You then
see the familiar Linux kernel startup messages (observe that it's
running in framebuffer mode, by the way), and finally you reach the
X-based graphical installer. If you've ever installed Red Hat or Fedora
Core, the graphical installer will be completely familiar to you; if not,
it'll be unfamiliar but still very easy. I'll
briefly walk you through the screens so that you wind up with a
configuration that contains everything you need and not too much
additional fluff:
- Welcome to Yellow Dog Linux -- Click Next.
- Language -- Choose English and click Next.
- Keyboard Configuration -- Choose US English and click Next.
- Installation Type -- Select Custom and click Next.
- Disk Partitioning Setup -- Select Manually partition with Disk
Druid and click Next.
- Disk Setup -- You'll be installing the simple case: a Linux-only
system (as opposed to a dual-boot Mac OS + Linux environment). In
Disk Druid, first delete any partitions you see. Next, create a
1Mb partition of type Apple Bootstrap. Next, create a swap
partition at least as large as the amount of physical RAM in
your Mac mini. Finally, create a root partition of type ext3,
and select Fill to maximum allowable size so it covers the
remainder of the disk. This is not necessarily the best
file system arrangement, particularly for an embedded appliance,
but I'm configuring things this way for simplicity's sake. By
the way, you'll notice that this version of Disk Druid, unlike
the version used on x86 distributions, does not sort partitions;
it puts them on the disk in the order you request them.
- Firewall Configuration -- For this simple application, choose No
firewall, and set Enable SELinux to Disabled.
- Additional Language Support -- Leave this at its default
settings; just click Next.
- Time Zone Selection -- Choose your physical location as
appropriate.
- Set Root Password -- select a password that's easy for you to
remember.
The last thing you have to set up is the Package Group Selection
dialog. First, go down the list of package groups unchecking everything.
Then, check the following package groups ONLY, go into the details for
these groups, and verify that only the items listed below are checked:
- In the X Window System group, check
Xautoconfig,
xorg-x11-twm and xterm.
- In the Editors group, check whichever editor(s) you prefer to
use; I use
emacs; you might like something else.
- In the Text-based Internet group, select
elinks -- it's very
useful to have a simple Web browser available in case you need
it.
- In Development Tools, the default package options are pretty
good. I personally prefer to deselect
gcc-g77,
gcc-gnat, and gcc-java,
since I know I never use these languages, but feel free to
leave them checked if you want to keep your options open.
- In Kernel Development, for some reason, GTK+ and Qt library
materials are listed. Deselect all the optional packages in this
group; you only need the kernel source.
- In X Software Development, keep the default package set.
All the hard work is now done. After you confirm your desire to install
the operating system, the installer informs you which CDs you're going
to need -- if you followed my installation choices exactly, you'll need
CDs 1, 2, and 3. You then merely need to change disks as prompted.
When the installation is complete, the last disk will eject and
you'll be prompted to reboot. The startup process, if left unmolested,
will run through all the familiar Red Hat-style screens and scripts, and
you'll wind up at a text-mode login prompt. If Kudzu (RedHat's hardware
probing code; see Resources) runs the first time you
start the machine up, let it install or remove hardware configurations
however it wants. Log in as root, with the
password you specified earlier in the installation process.
Configuring the system
You now need to do a little tweaking and polishing to get the system
into a useful state. The first thing you need to know is that you won't
have audio facilities -- at least not right away. The reason for this is
that currently the audio chip in the Mac mini is not supported by the Linux kernel. However, it will almost certainly be supported soon (in
fact, a patch might well be available by the time you read this).
Alternatively, you could pick up a supported USB audio interface, such
as the Griffin Technologies iMic (see Resources).
Enabling sound
To test the audio features I'll mention in a later
article, and to verify that my code really works, I secretly built myself a
second development system (a first-generation Bondi blue iMac, if you're
interested) with the exact same software configuration that I discuss
here. Audio wasn't detected or enabled by default after installing on
the iMac, so I had to set it up manually. At the end
of /etc/rc.d/rc.sysinit (immediately before the
comment that reads "# Let rhgb know that we're leaving rc.sysinit"), I added the lines:
Listing 1. Configuring sound
modprobe snd-powermac
modprobe snd-pcm-oss
aumix-minimal -v85 -l85 -c85 -w85 -W85
|
I also added the line alias sound-slot-0
snd-powermac to /etc/modprobe.conf,
although this isn't strictly necessary.
You can set your own audio volume levels, of course; those defaults just
happen to work for my environment. You can use the playwave command with
any convenient .wav file, or the cdp program with an audio CD, to
test your current volume settings; use the aumix
user interface to
adjust the various mixer settings. After you set the audio to a
comfortable level, run aumix again and press s (note the lowercase).
This will save the current mixer settings to a file called ~/.aumixrc --
simply examine this file to determine the exact numbers to put in the
script. Note that you cannot use the aumix -L
(load settings) option
from within startup scripts; it almost always segfaults. That's why you
use aumix-minimal with command-line options.
You should make all the audio-related configuration changes described
above to your Mac mini, in preparation for the availability of
compatible drivers. Think of it as a vaccination against having to come
back and reread this article later!
Enabling the network
The next thing to do is get the network connectivity
operational. Although Linux supports the Ethernet MAC in the Mac mini, unfortunately it doesn't appear to be auto-detected. Make sure that the line alias eth0 sungem appears
in /etc/modprobe.conf. You then need to create a configuration file
for the Ethernet adapter. This file must be saved
as /etc/sysconfig/network-scripts/ifcfg-eth0,
and it should contain the following lines:
Listing 2. Configuring the network
DEVICE=eth0
BOOTPROTO=dhcp
ONBOOT=yes
TYPE=Ethernet
|
Note that some people report that Kudzu automatically set all of this up
for them on first boot. This didn't happen for me; it could have
something to do with the rather austere set of install options I
selected, or it could simply be the unreality field that surrounds me.
An FYI note: If you're really getting into the swing of this
article, and you're setting up an ancient iMac like mine as your
secondary system, note that the Ethernet driver is called bmac on this
old machine -- so change the modprobe.conf line to
alias eth0 bmac. The
other setup details are all the same.
At this point you should be able to bring up the Ethernet interface with
ifup eth0. Assuming you're connected to a network and a
functional DHCP server is within reach, you should be able to run some
basic network tests like pinging your default router. Use ifconfig to
find out your machine's IP address, and write it down -- you'll need it
to do some testing in a moment.
Installing a Web server
Next, install the thttpd Web server,
which will provide a user interface to control your appliance-to-be. You can download the source code for thttpd from the link in
Resources. Extract the tarball, cd to the directory thus created, and
run ./configure to create the makefile. Now execute
make all ; make install to build and install the daemon. During the install process,
you'll see an error message about a nonexistent www group. You can safely ignore this error, since it relates to a feature of thttpd that you
won't use.
Now that you've built and installed the server, create a minimal
configuration file, using the following lines:
Listing 3. Configuring thtppd
dir=/web
cgipat=/cgi-bin/*
logfile=/dev/null
pidfile=/var/run/thttpd.pid
|
Save the above lines as /etc/thttpd.conf, and
edit /etc/rc.d/rc.local to
add the line /usr/local/sbin/thttpd -C
/etc/thttpd.conf. It doesn't
matter where you put this line. By the way, while you're debugging your
system, it might be useful to direct the log output to a real file
rather than /dev/null. For a production
system, however, you should
disable logging unless you have some methodology in process to rotate or
truncate the log file. Otherwise, you're setting the stage for an
out-of-disk-space condition at some unpredictable time in the future.
|
Security unconscious
Those of you who are downright razor-sharp, or who simply like to notice
these sorts of things, will observe that most consumer electronics
manufacturers practically ignore the concept of security for network-
accessible appliances. (The only regular exception to that rule is the
exceedingly limited case of digital rights management (DRM) - and that's
not security for you, that's (illusory) security for the manufacturer,
who can often also be a provider of paid-for media).
You'll notice that I've specifically excluded the topic of security from
this article. This is not intended to encourage the above-mentioned
cavalier dismissal of security as a design goal.
The main reason I've left it out is that this article is about the fun
and games of multimedia playback. The network infrastructure built here
is strictly for demonstration purposes, and is not intended to be
production grade. In other words, I didn't want to spend all my word
budget on scaffolding - I wanted to focus on the application of
interest.
A compelling secondary reason to leave it out is, as mentioned in the
body text of this article, that network security, and particularly Web
server security, is a specialized field. In other words, I didn't want
to give a five-minute dime tour of a field that takes years of
experience to understand adequately.
However, I'm happy to accommodate my loyal readers - if you're
interested in seeing one or two additional articles at the end of this
series, which talk about some aspects of how to design the network
interface to be more secure, just put a comment in the discussion
forum or drop me an email.
|
|
Before continuing, make certain that the /web directory exists on your
system, and create it if it doesn't. Also ensure that it is world writable
with chmod 777 /web. When you have done this, run
thttpd -C /etc/thttpd.conf to start the thttpd daemon manually. On another
machine on the same network, open a Web browser and navigate to
http://a.b.c.d/, where a.b.c.d is the IP address of your Macintosh. If
everything is working properly, and assuming you didn't put anything in
the /web directory, you'll see a directory listing, on a lime-green
background, headed "Index of /".
Installing an FTP server
Next, you need an easy way for people to get pictures and other
multimedia files onto the unit's hard disk. I use ProFTPD
for this, because it's one of the best-known open source FTP servers
around. It's flexible, reasonably secure, and available at an unbeatable
price. The current version at the time of writing is 1.2.10, and you can find it from the link in Resources. Again, to build and install it,
simply extract the tarball, change to the directory thus created,
run ./configure then make all ;
make install.
Configuring ProFTPD is substantially more complicated than configuring
thttpd, because the FTP server tries to interact with the operating
system a lot for authentication purposes. The default configuration
file, which gets installed to
/usr/local/etc/proftpd.conf, contains a
bunch of assumptions that aren't valid for what you want to do. Here's an
abbreviated configuration file that should meet your needs:
Listing 4. Configuring ProFTPD
ServerName "developerWorks Multimedia Appliance"
ServerType standalone
DefaultServer on
Port 21
Umask 022
MaxInstances 30
User root
Group root
AllowOverwrite on
<Anonymous /web>
User ftp
Group ftp
UserAlias anonymous ftp
MaxClients 10
</Anonymous>
|
You also need to edit /etc/rc.d/rc.local
again and add the line
/usr/local/sbin/in.proftpd; again, it doesn't matter where you put
this line. At this point you should be able to
run /usr/local/sbin/in.proftpd to start the server manually, and then
test connecting from another machine (log in as "anonymous" with any
password).
The above discussion assumes that your network is inherently
secure against people that are going to want to do Bad Things to your
photos. If you want to develop a more secure system, please refer to the
ProFTPD documentation (see Resources). The complication here is that you want to provide
remote access to the same directory that thttpd uses for the Web serving
root; you need to be extremely careful that you don't inadvertently make
it impossible for thttpd (and the CGI programs it's going to call) to
read files that were uploaded over the Web interface.
One more useful network daemon runs by default in the Yellow
Dog setup you installed, and that's sshd. Leave this process
running, because it's handy to be able to connect to the box over the
command line. (You might also want to do your development remotely). Because
you haven't set up any secondary users on the Mac mini, to connect through
ssh on your remote computer, use the command
ssh -l root a.b.c.d, where
a.b.c.d is the IP address of the Mac mini. The first time you do this,
you will need to confirm the identity of the remote system. Remember
to type yes and not just y.
Summary
This article showed you how to install Linux,
configure the networking and audio subsystems, and install and
configure a simple Web server. You also set up an FTP server that
will let you get multimedia content onto the hard disk remotely. At this
point you can start doing your development pretty much entirely from a
remote PC, too.
Resources - See the other articles in this series.
- My Linux distribution of choice is Yellow Dog. Besides being freely downloadable, you can also buy this distribution in various cute packaging options.
- You can
buy
a Mac with Yellow Dog Linux pre-installed on it
from Terra Soft Solutions.
- Another Linux option you might want to consider is Debian. Start with
the latest
PPC Debian snapshots,
found on their Web site. If you're familiar with Debian
on x86, this is probably a better choice. Although I was initially
considering this distribution, I decided that the Red-Hat-ness of Yellow
Dog made it a more comfortable platform.
- If you're going with Debian, you'll probably want to read this
guide to
installing Debian on the Mac mini.
Actually, it's an interesting piece of background reading
regardless of which distribution you intend to use.
- Yellow Dog Linux may not support the mini's built-in audio, but it
supports the Griffin
iMic USB audio interface.
- The Web server used in this article is the slender, high-performance
thttpd. I've spoken about this
elsewhere, so I won't repeat myself. Just download it, already!
- I also use the ProFTPD FTP
server. This is a
very full-featured package; this series only uses a tiny section of it.
- Linspire sells a custom Linux distribution for PCs.
The interesting thing about this company is that they have taken an
active interest in acquiring multimedia codec licenses, and they sell a
Linux media player which is fully licensed for playing DVDs and other
media. If you were planning to do a commercial product of the kind we're
talking about here, you might want to contact these people.
- This FAQ about DVDs
goes into more detail about licensing issues.
- Another developerWorks article offers a different
view of the Mac mini as an embedded development platform.
- IBM maintains substantial documentation on its strong
commitment to Linux.
- Mini too cheap? Maybe you should run your Linux
application on big iron.
- If you enjoy living on the bleeding edge, you can overclock
your Mac mini with the information found here.
- Find more information about the Kudzu hardware probing library
and related code on RedHat's site.
About the author | | | Lewin A.R.W. Edwards works for a Fortune 50 company as a wireless security/fire safety device design engineer. Prior to that, he spent five years developing x86, ARM and PA-RISC-based networked multimedia appliances at Digi-Frame Inc. He has extensive experience in encryption and security software and is the author of two books on embedded systems development. He can be reached at sysadm@zws.com. |
Rate this page
| |
