Background: OEM install of Ubuntu 18.04 on a Dell Latitude. Note that this not an install of Ubuntu by the OEM, rather, Ubuntu 18.04 was installed by me using the "Install as OEM" option.

While exploring the menus the user clicked on something that upon next boot changed the OEM install into a regular install, with a username other than oem. He was prompted with the regular install screen and filled it out to the best of his ability. In no place was there a warning that data would be lost. The system was immediately shut down and I have the drive now mounted read only on my desktop.

Now that this is converted to a regular install, how might one go about recovering the data that was in /home/oem/? Note that extundelete could not find anything deleted on the disk, but I do see that the custom /opt/ directory with third party applications is still intact.

I used autopsy to scan the drive and I do see that there is a file (not a directory) called /home/oem/ but its contents do not seem useful: https://pastebin.com/XXCiu7Bb

I have also tried opening the partition with testdisk (Photorec tool) but also there I see the file /home/oem but not the directory /home/oem/ which I know to have been the home directory of the OEM user. Possibly I created this file by accident in an early stage of the recovery process while it was still in the laptop, but now I have the drive mounted read only on another computer. Yes I am aware that some data may have been lost due to the drive having been mounted read-write.

How might I recover any files that the user had in the OEM install? I would be happy with anything from the /home/oem/ directory that no longer exists, but mostly text and ODT files if possible.

  • 3
    It seems somewhat odd that you even had an oem user. The OEM install system is designed such that on first bootup, the user is prompted to create their own username/password and set up their own account. I'm assuming Dell messed something up and that your user clicked the "Prepare for shipping to end user" button (which, again, shouldn't exist). As for recovery, try tools like photorec perhaps? – Kaz Wolfe Nov 28 '19 at 5:54
  • @KazWolfe: Though this is called an OEM install, it was not installed by Dell. This machine came with some version of Windows and I installed Ubuntu on it. – dotancohen Nov 28 '19 at 21:29
  • This was the first time that I had seen the OEM install. I have been installing Ubuntu since about 2007 and Linux since September 2001, but I've never seen the OEM install option until now. I did not realize that it could be wiped out so easily. That said, there was no excuse for not backing up the user data. Computers fail be it hardware failure, software bugs, malicious intent, a mistake, or plain stupidity. – dotancohen Nov 28 '19 at 22:25
  • The only time I can think of a home user use the OEM install is when they want to sell or give their computer to someone else. You won't be happy if I sell you an Ubuntu computer and keep an ssh key in the /home/oem so that I can remotely login to your compter when I like. Neither will I be happy if you find my stash of pictures I have left in the /home/oem folder. – user68186 Nov 28 '19 at 23:52
  • 1
    I'm voting this question because it is a good reminder to people to backup the files they've created (plus configuration settings) daily. – WinEunuuchs2Unix Nov 29 '19 at 2:13

Let's quickly start with the general purpose of the OEM environment: it allows the creation and customization of a single machine or a small batch of machines at a time. This is (relatively) similar to the Windows concept of audit mode. A typical OEM workflow would look like this:

  1. An OEM boots up a device and installs OEM mode to the device.
  2. The OEM will configure the device, install any applications, load drivers, so on.
  3. After configuration, the OEM will prepare the device to the end-user for shipping. The system will shut down/restart. [I assume this was the step that the user ran, triggering this entire situation.]
  4. If the device is a "gold-master" and is part of a larger batch of devices, the device will be imaged out and cloned to other devices.
  5. A prepared device is shipped to end-users.
  6. The user receives a prepared device, turns it on, and goes through OOBE setup (username, password, etc.).

Most home or small workshops offering Ubuntu installation would use OEM installation on a per-device basis without any imaging. The end user would receive the device, and on first-boot they'd be prompted to create their account and configure everything.

So, what happens to the oem user when the device is prepared for shipping? Well, when a device is prepared for shipping, it kicks off a number of tasks that delete the OEM user and prepares the environment (uninstalling packages, updating systemd unit files, and so on). All of the scripts for preparation are located in /usr/sbin/oem-config*, and can be viewed here. The user is them dumped back into Ubiquity (the setup screens), where they go through account setup.

The bad news here is that all of these tasks are pretty write-intensive. A lot of the (recently-)freed blocks that was occupied by /home/oem will now be used for the new user's configurations and the changes made by the new system and the OOBE configurations.

With all this being said, you're fighting an uphill battle trying to recover this data. The files lost are (relatively) small, and as such are very easy to get overwritten when the new account is created. Tools like photorec/testdisk and autopsy would likely be your best bets, as you've already tried. If these tools can't recover anything, there likely isn't much more that can be done save for some very expensive and/or very powerful recovery tools that people don't usually have access to.

You did mention the existence of a /home/oem file as well as a folder. This may be causing data recovery tools to get a bit confused - a file can't have subfiles, so any attempts to piece together a file tree will fail at this point. When running a disk recovery tool, be sure to look for any files that are "orphaned" or otherwise don't belong to a specific directory.

I'm going to go back to my recommendation from the comments of using photorec. Instead of focusing on partition- or filesystem-level recovery like testdisk, it tries to pull and recover individual files while disregarding the directory or filesystem structure. This is suboptimal (paths and potentially file names are lost), but having the raw data recovered is better than having nothing recovered. I suspect autopsy has a similar "miscellaneous" output bin, but I'm not familiar enough with the tool to tell you where to look for that.

If photorec can't recover the files themselves, it's likely that they were (mostly) overwritten during the personalization phase initiated by the user. In this case, it might be best to contract the services of a data forensics expert with access to the necessary tools and knowhow to recover data like this.

  • Thank you, I will focus my efforts on "orphaned" files. – dotancohen Nov 28 '19 at 22:57
  • I would add point 6. The end user when boots the computer for the first time will then be prompted to setup their user name, password, time zone, language, keyboard, etc. so that they can user the computer on a daily basis. – user68186 Nov 28 '19 at 23:44

The good news first

The good news is if User OEM was deleted and a new user was created in it's place then the skeleton files created would hopefully be the same size as the skeleton files deleted from User OEM. Then if new user didn't create any new files besides that, the new files created by User OEM would hopefully still be on the disk in abandoned sectors. Hopefully the only OEM files lost would be OEM skeleton files simply replaced by new user skeleton files in relatively the same locations. Underline "hopefully"....

And now the "But..."

It would certainly be a lot of work to recover the data. If you are interested in the .odt files you'll first need to find a string of characters to represent an .odt file. For example:

$ ll Double*

-rw-rw-r-- 1 rick rick 192599 Mar 18  2018 Double suspend.odt

$ head "Double suspend.odt" -c 400


So I could grep my hard drive (you can't use this method though) for all files with the string "mimetypeapplication/vnd.oasis.opendocument.textPK":

$ time grepall "mimetypeapplication/vnd.oasis.opendocument.textPK"

Binary file /usr/share/cups/data/form_english_in.odt matches
Binary file /usr/share/cups/data/form_russian_in.odt matches
Binary file /usr/share/doc/fonts-sil-abyssinica/documentation/AbyssinicaSILGraphiteFontFeatures.odt matches
Binary file /usr/lib/libreoffice/share/template/common/internal/idxexample.odt matches
    (... SNIP ...)
60.47user 22.74system 2:23.48elapsed 58%CPU (0avgtext+0avgdata 20788maxresident)k
61233744inputs+0outputs (0major+8533minor)pagefaults 0swaps

real    2m25.881s
user    1m0.507s
sys     0m22.759s

Now instead of using grep you would have to use a tool that searches every byte of your hard drive whether it is part of a file or not. An example would be tools in this closed Q&A:

Then there are other tools:

Finally if you had just written the new Microsoft DOS 1.0 and lost the only copy of the assembly code you could use a data recovery service. Which you can google for. Finally consider this article:

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.