(Update 01/09/2009 - All of the below is now not necessary as the latest firmware has an option from the web interface to turn on remote root ssh shell access, YAY for Western Digital!)
I recently bought a Western Digital ShareSpace. It’s a larger version of the My Book World range of Products which have quite a following at WDMBWE. Mine is a 4TB version which has four 1TB drives in a raid 5 configuration giving 2.7T usable. They all run Linux, the Share Space ones seem to have some sort of u-boot and have the root file system on the first partition of the disks. Unfortunately getting a shell on these boxes is a bit harder than the original devices.
Martin Hinner’s method at his site does not work because the firmware got a major overhall. The original My Book World’s used perl (/auth/firmware_upgrade.pl) but the newer Share Space devices use php (/admin/system_firmware_manual.php). My first thought was to implement a similar hack that Martin used, amazingly western digital put a link to the main code on there website. This gives all the juicy details about how the upgrade process works.
The automatic update process gives the western digital webpage its version number and it returns if there is an update or not along with a link to the firmware blob. This believe it or not is tar.gz with the 1st and 16th block swapped where blocks are 5k. Due to the way the upgrade process works there is no easy to to hook into upgrade process. If a manual firmware image is selected it gets downloaded and a script on the device installs the update. If you could trick the upgrade process into thinking it it a service pack update it will run a pre-install script inside the firmware blob. However the service pack install file (/etc/sp) does not exist so I abandoned this method.
The easiest way I found was to add a telnet service. To do this I removed the first disk (bottom) and connected it via a sata to usb converter to my laptop. I did see an existing wiki page about how to do this, but it missed out some large details. It is important to edit the filesystem as a broken raid1 mirror, or else on the next boot the root filesystem mirrors will not be in sync. The next time the device boots it will notice one disk is more uptodate and resync the other three and as its only 200M its very quick to sync. Something like this worked for me:
sudo mdadm --assemble /dev/md0 /dev/sdb1 sudo mdadm --run /dev/md0 sudo mount /dev/md0 /mnt
Next edit /mnt/etc/init.d/rcS and find the “Start Network” comment and add the telnet lines in so it looks like this:
# Start Network && HTTPD /etc/init.d/S40network restart /etc/init.d/S55mini_httpd restart echo "telnet stream tcp nowait root /usr/sbin/telnetd /usr/sbin/telnetd" >>/etc/inetd.conf /etc/init.d/S60inetd start FACTORY_DEFAULT_FLAG=/etc/.factory_restore GENERAL_RESTORE_FLAG=/etc/.general_restore
Now you will have a working telnet deamon but no user to login as, the default user admin does not have shell set. I just blanked the root users password by editing /mnt/etc/passwd and making the root line look like:
Then unmount it and stop the md device with “umount /mnt” and “mdadm -S /dev/md0”. Pop the disk back in and as soon as it starts pinging you should be able to login as root. At which point I set a new root password with passwd. Then ipkg can be installed in /opt and once a reliable ssh deamon is working the above hack in rcS can be removed.