Virtual Dev Environment For Chumby

As an alternative to cross-compiling software for the Chumby platform with the help of a toolchain, you can also install a virtual environment compiled to run for the ARM platform.

This will allow you to compile Perl and other tools you might find useful on a Chumby.

Description

These are the steps needed to set up a complete dev environment for ARM.

We use QEMU on Debian to run a Debian instance emulated for ARM.

The only system where this works reliably is Debian.

You have to:

- Get a Debian vmware machine from the appliance marketplace.

or

- Download and install the latest Debian (40.r3)

I tried to set the environment below on other linuxes, with mixed results:

- on Redhat Fedora 8, the ARM emulator crashes throughout the installation, it was impossible to get anything installed consistently. Even with a pre-configured environment it was not possible to get a stable ARM system working.

- on Ubuntu and Suse you can get the install to complete, but at first boot you run into issues with the framebuffer support. The kernel would have to be eventually recompiled with different parameters to get this working.

- on (Free|Open)BSD there is not only the framebuffer issue, but also it is rather difficult to get a bridged networking working.

- on Windows XP, setting up a bridged networking with the recommended TAP driver never worked right, I gave up after many tries.

Building the ARM emulator image

Build the environment in /root/arm. These are the files you should have to start:

 initrd.img-2.6.18-6-versatile
 linux-image-2.6.18-6-versatile_2.6.18.dfsg.1-18etch1+versatile_arm.deb
 vmlinuz-2.6.18-6-versatile

To get the files above, as well as to set up your system properly, follow closely the instructions to build the ARM environment at:

 http://www.aurel32.net/info/debian_arm_qemu.php

If QEMU cannot be installed in the Debian host with apt-get, download QEMU from

 http://packages.debian.org/qemu

It is recommended to use the packaged QEMU that came with your Debian, sometimes this will be 0.8.x, which is the highest version that is stable on that platform. The versions before 0.9 do not have qcow2 support, so for the sake of portability, do not create anything above the qcow version.

I found it eventually more reliable to actually install everything on a raw file, then convert the raw to qcow. For example, neither qcow nor qcow2 were stable on other linuxes, but once installed on raw and converted back to qcow, the system booted fine.

QEMU has a tool named qemu-img that you can use to convert images between the supported formats. Do not use the compressed format, it causes serious issues when trying to update existing packages.

To start the system, use a shell file (i.e. startQemu.sh) with the following contents:

 #!/bin/bash
 qemu-system-arm -M versatilepb -kernel vmlinuz-2.6.18-6-versatile \
   -initrd initrd.img-2.6.18-6-versatile \
   -hda hda.img -append "root=/dev/sda1" \
   -net nic,macaddr=00:16:3e:00:00:01 -net tap

You should change the macaddr to a different value, to allow for uniqueness of mac addresses on the network. On some versions of QEMU, removing the macaddr parameter is required, otherwise the -net nic does not bind properly.

Add -m256 to the command to make 256MB available to the ARM environment, if your virtual host has enough memory. This is also the top limit, the default is 128MB.

With the way the ARM is supported, there is no way to get the system time propagate correctly with QEMU. The system will always atart with the unix 0 date (1969), and this will cause serious issues when trying to install/update packages. Therefore, the first priority was to set the time right from the beginning.

After booting the arm environment, if the system date was not set, set it like:

 date -s "03/21/2008 +12:00:00"

It is recommended to put the above statement in an rc2.d linked script that will set the system time early in the boot process.

Install NTP client support

For correct behavior of the emulator, the real date has to be set after boot. For that, we use NTP:

 apt-get install ntpdate

You have to create and customize the file /etc/ntp.conf:

 # /etc/ntp.conf, configuration for xntpd
 
 logfile /var/log/xntpd
 driftfile /var/lib/ntp/ntp.drift
 statsdir /var/log/ntpstats/
 
 statistics loopstats peerstats clockstats
 filegen loopstats file loopstats type day enable
 filegen peerstats file peerstats type day enable
 filegen clockstats file clockstats type day enable
 
 server 64.125.79.28
 server 192.73.48.1
 server 64.125.79.29
 server 192.73.48.2
 server 64.125.79.30
 server 131.216.22.15
 server 69.36.249.227
 server 132.236.56.250
 server 131.216.22.9
 server 192.73.48.6

Of course, you could use foll host names instead of the IPs above:

 server pubts0-sj.witime.net
 server chronos1.umt.edu
 server pubts1-sj.witime.net
 server chronos2.umt.edu
 server pubts2-sj.witime.net
 server tick.cs.unlv.edu
 server stratum2.sjc1.publicntp.net
 server ntp0.cornell.edu
 server tock.cs.unlv.edu
 server chronos3.umt.edu

You also have to create the start file /etc/init.d/ntpdate:

 #!/bin/sh
 # /etc/init.d/ntpdate
 #
 NTPDATE=/usr/sbin/ntpdate
 LOGFILE=/var/log/xntpd
 NTPCONFFILE=/etc/ntp.conf
 NTPSERVERS=
 GREP=/usr/bin/grep
 CUT=/usr/bin/cut
 
 if [ -f $NTPCONFFILE ] ; then
   echo "===========================" >> $LOGFILE
   echo "<<< `date` " >> $LOGFILE
   for i in `grep ^server $NTPCONFFILE | $CUT --fields 2 --delimiter \ `
     do
       NTPSERVERS="$NTPSERVERS $i"
     done
   $NTPDATE $NTPSERVERS  >>  $LOGFILE
   # echo $NTPDATE $NTPSERVERS $LOGFILE
   echo ">>> `date` " >> $LOGFILE
 fi
 exit 0

Link this file in the rc2.d as follows:

 cd /etc/rc2.d/
 ln -s S19ntpdate -> ../init.d/ntpdate

To make sure the date setting works one way or another, we also set the date to something in the current time, in case the NTP connectivity fails.

The best way is to edit the /etc/rc.local, and add:

 date -s 03/21/2008
 /etc/init.d/ntpdate

Install dev packages

 apt-get install build-essential

This will install GCC and all other packages needed to build stuff.

Install the SSH daemon

 apt-get install ssh
 /etc/init.d/ssh start

Issues with moving the appliance

Every time you move the appliance (or if you run it from a notebook and reconnect to a different network), it is possible the the eth0 becomes eth1, etc. In this case you will have to reconfigure the network bridge for the ARM emulator to be able to connect to the TAP device.

For that, you modify the statement in the /etc/network/interfaces to reflect the current eth card ID:

 auto br0 
 iface br0 inet dhcp 
   bridge_ports eth1 
   bridge_maxwait 0

Full documentation for QEMU can be found at:

 http://fabrice.bellard.free.fr/qemu/qemu-doc.html