If you don't want to end up loosing some work (as I did), make sure you don't have any machines with their state saved (ie. make sure you shut all machines down) before you upgrade.

All in all I am very pleased with the setup, as it behaves consistantly now (knowing the needed tweaks to get it running). I believe it performs as well as running an AP with the madwifi-ng karma kernel patches from digininja. Well, except for one thing. When changing the MAC address of the wifi card it doesn't really do such a good job anymore. For some reason most clients fail to associate, some still do but theres a considerable drop in the number of connecting clients.

I have tried to change the MAC using both ifconfig and simply specifying the -a option to airbase-ng. Sniffing the traffic using another wifi card shows a lot of probe requests and responses being re-sent, but no requests to associate. From my initial checks I do not see any difference in the requests or responses apart from the "Frame check sequence" and the MAC of-course. But then again, I haven't put that much time into it.

ANY insight into why this happens, and if I am missing something REALLY obvious would be great. I will file a bugreport in the aircrack-ng TRAC system once I collect and clean the relevant packet traces.

I am running the latest PC Suite 7.0.8.2 on Windows XP in Virtualbox 1.6.4. My current workaround is to connect to the phone over bluetooth from the same Virtual XP installation using a USB Bluetooth dongle. I must say that this part is a bit of a pain in the ..... Well, I guess you can't have it all.

The issue has been reported at both Virtualbox and Nokia forums so hopefully there will be a fix soon.

What remains in order to get a good solid environment running is a better wifi card with a lot better transmit power, which I've ordered from the UK today. Hopefully it will be here by monday so me and my friend can do some more tests over a beer or two.

For starters I (once more) ripped apart my USB adapter and de-soldered my external antenna cable and found a better spot for it. The new re-fitted antenna works a lot better now and actually picks up quite a few new networks (and clients of course).

Another change I made was to switch the wireless drivers to the ones available from the git kernel sources which for the moment seem to work better than the previous ones.

I still believe in the airbase-ng concept of a user-mode AP which works with many different chipsets and not that dependent of driver patches. However, in order to switch away from the madwifi-patches it has to work atleast as good as they do and with a USB wifi-interface.

Once I get things running I'll post my setup here. If someone already has done so, please drop me a line with your setup including hardware

As a friend told me about the airbase-ng project I though I would give it a shot...

Having tried airbase-ng with both the built-in Atheros chipset and with numerous USB rt73 based wifi-interfaces I was still not satisfied with the result. Even though I got airbase-ng working (at least sometimes) it would not let clients associate to the wifi-network to the extent that the "karma" madwifi-ng driver patches do. In spots were the patched madwifi drivers were picking up and letting 10 clients associate, airbase-ng would pick up and associate one and only list direct and broadcast probes from the others. Having limited time to debug the software in order to understand the reason for this I stuck to the madwifi-ng "karma" patches.

A good set of "karma" madwifi-ng patches are available from www.digininja.org and can be downloaded here:
http://www.digininja.org/files/karma-madwifi-0.9.4-3379.patch

These patches make it possible to activate and inactivate the "karma" behaviour by using iwpriv and therefore do not interfere as much with the driver as other patches have been doing in the past.

In order to install them to the eee a subversion snapshot of the madwifi-ng drivers, supporting the chipset in the eee must be downloaded. A subversion snasphot of the madwifi-ng driver can be downloaded from here:
http://snapshots.madwifi.org/madwifi-hal-0.10.5.6/madwifi-hal-0.10.5.6-r3835-20080801.tar.gz

As the patches come for another subversion version than the one were downloading a couple of hunks will fail when applying the patch. In order to address these failed hunks I am supplying an additional patch which has to be applied after this first one to correct the failed hunks. This patch can be found here:
http://www.cqure.net/files/001-madwifi-hal-0.10.5.6-r3835-20080801-digininja-fixup.patch

These are the steps i took to install the patched driver to the eee PC:

wget http://snapshots.madwifi.org/madwifi-hal-0.10.5.6/madwifi-hal-0.10.5.6-r3835-20080801.tar.gz
tar xvzf madwifi-hal-0.10.5.6-r3835-20080801.tar.gz
cd madwifi-hal-0.10.5.6-r3835-20080801/
wget http://www.digininja.org/files/karma-madwifi-0.9.4-3379.patch
patch -p1 < karma-madwifi-0.9.4-3379.patch
wget http://www.cqure.net/files/001-madwifi-hal-0.10.5.6-r3835-20080801-digininja-fixup.patch
patch -p1 < 001-madwifi-hal-0.10.5.6-r3835-20080801-digininja-fixup.patch
make
sudo make install

"Personal use is when you install the product on one or more PCs yourself and you make use of it (or even your friend, sister and grandmother). It doesn't matter whether you just use it for fun or run your multi-million euro business with it. Also, if you install it on your work PC at some large company, this is still personal use. However, if you are an administrator and want to deploy it to the 500 desktops in your company, this would not qualify as personal use. Well, you could ask each of your 500 employees to install VirtualBox but don't you think we deserve some money in this case? We'd even assist you with any issue you might have."

Virtualbox comes in both an Open Source Edition (OSE) and a closed source edition. Apart from not supporting USB the OSE release would be sufficient for my needs and has many of the features available in VMware. For details on the differences between the OSE and closed source edition look here.

My current setup is an Ubuntu host system (Hardy) on which I am virtualizing a number of different boxes used for development, testing, hacking and office stuff. The hosts I have been testing on Virtualbox all run Windows. As Hardy (at the time of the writing) had an outdated version of Virtualbox I went for a Ubuntu packaged version available from the vendor (Sun's) web site.

The last time I tried Virtualbox it didn't go so good as I was to lazy to reinstall my Windows boxes and migration was not as easy as now. This time this was a lot smoother as Virtualbox supports the VMWare virtual disk format. So basically I copied the disks to a new folder and created a XP configuration pointing the disks to the copied ones. Upon first boot Windows XP SP2 blue screened. This brought me to this article: http://www.virtualbox.org/wiki/Migrate_Windows

I followed the easy steps in the "How to migrate existing Windows installations to VirtualBox", involving running MergeIDE on the system when booted in VMware and disabling the agp440.sys by booting into Windows recovery console. Finally, checking the IO APIC option (as I was to lazy to change my hal) in the Virtualbox configuration I was good to go. The system booted perfectly and ones running I installed the Guest tools to get better hardware and graphic support.

Something that I noted directly and which I can confirm after several weeks of use is that Virtualbox at least "feels" considerably faster. I do not have any hard figures as I did not benchmark it. However, my Vista systems are quite fast and I can now actually do some real work on them. Another thing that works a LOT faster is suspending and resuming guest systems, and as I use this feature a lot. In Virtualbox this takes seconds rather than minutes, even with systems having a gig or more of memory.

Another nice thing with Virtualbox is its support for something called Seamless Mode. This mode allows for windows from virtual machines to better blend in with windows from the host OS. In this way you can have Windows applications from the guest side by side with your host applications.

Bridging network interfaces is one of the things that isn't as smooth as in VMware. In VMware it's a piece of cake and works painlessly (except for some wireless configurations). In Virtualbox it involves using the hosts OS bridging capabilities which makes it slightly more painful. My first attempt on bridging my wireless-interface failed miserably and I haven't tried since. The NAT networking however is roughly the same. Both applications support virtualizing the interface as either the AMD PCNet or the Intel Pro adapter.

When switching from the OSE release to the closed source 1.6.2 release that I am running now I experienced some troubles with the USB support. Each time I configured a guest machine I received the following warning "Could not load the Host USB Proxy Service (VERR_FILE_NOT_FOUND). The service might be not installed on the host computer.". The consequences of the error, as it suggests, was that obviously USB was not working. This was because Virtualbox looks in the old /proc/bus/usb location rather /dev/bus/usb for details regarding USB. In order to fix this in Ubuntu one needs to edit the /etc/init.d/mountdevsubfs.sh file.

sudo gedit /etc/init.d/mountdevsubfs.sh

Look for the following section in this file:

#
# Magic to make /proc/bus/usb work
#
#mkdir -p /dev/bus/usb/.usbfs
#domount usbfs "" /dev/bus/usb/.usbfs -obusmode=0700,devmode=0600,listmode=0644
#ln -s .usbfs/devices /dev/bus/usb/devices
#mount --rbind /dev/bus/usb /proc/bus/usb


Then uncomment the last four lines so it looks like this:

#
# Magic to make /proc/bus/usb work
#
mkdir -p /dev/bus/usb/.usbfs
domount usbfs "" /dev/bus/usb/.usbfs -obusmode=0700,devmode=0600,listmode=0644
ln -s .usbfs/devices /dev/bus/usb/devices
mount --rbind /dev/bus/usb /proc/bus/usb


Finally, even though Virtualbox looks very promising it has crashed 3 or four times during the last month, something that has not happened for me once running VMware for the last couple of years. This was obviously VERY annoying and unacceptable. I don't know what to blame so far my however I would gladly blame my crappy HP6910p notebook. However, that would probably not be fair as I am running VMware on this same hardware.

So, I still have some more parameters to compare, investigate and try on my part before making a decision on what platform to use in the future.

I realized that I was missing a dependency (libevolution3.0-cil) needed to get the Evolution plugin running. Simply installing it after Do was in place resolved my troubles:

sudo aptitude install libevolution3.0-cil