Recovering files off an H10 Optane w/ NAND SSD Module

Recovering files off an H10 Optane w/ NAND SSD Module

May 9, 2022

Opening the disk image in hexedit.

Opening the disk image in hexedit.

Intro #

A few weeks ago, my year and a half old Dell Inspiron 2-in-1 laptop decided to just stop working. I came back from the kitchen, where I made some tea, but my screen was blank and the computer was unresponsive. I tried turning it off and on again, but all I got was a keyboard blinking every 5 or 3 seconds and the charging LED blinking white once every 7 seconds. I couldn’t find anything to help me diagnose what was going wrong, but it seemed like it was a logic board failure.

I still had files on the computer I wanted off. I had been putting off sorting thorugh duplicate files and merging my old organization system with my new one, but it turns out, that is not a good reason to not make a back-up. Fortunately, I know enough about the low-level details of what I found to figure out how to get my files off.

To start, I bought two external M.2 NVME SSD drive enclosures and a new 2 TB SSD. I needed to create a backup disk image. I also needed space to put the files I wanted saved after I successfully retreived them. I also needed a live distro of Linux to use. I just so happened to use Kali Linux, but there’s plenty of live distros to choose from.

Inspect the drive #

Make the image of the drive:

dd if=/sdc of=/media/kali/F7FD-F507/disk0.img # physical block sizes are typically 512KB, which is the default.

Next, we can see how the disk looks to various utils:

gparted /media/kali/F7FD-F507/disk0.img

Hmm, I didn’t have a raid configured.

blkid /media/kali/F7FD-F507/disk0.img

Really? This must have something to do with the H10 Optane’s setup. It must be using the Intel RAID controller in the BIOS to split the four channels into two each between the Optane portion and the 3D NAND portion of the module. Let’s see what the original disk looks like in dmraid (because I couldn’t see how to do this with the disk image file):

sudo dmraid -ay # this should find it and if it's valid it'll load it

That’s not good. Let’s see what else we can get out of this by ignoring the RAID superblock:

blkid --probe --usages noraid /media/kali/F7FD-F507/disk0.img

Protective MBR, eh? Well, that doesn’t tell us much. Perhaps TestDisk will help us more?

No, none of that looks right. Looks like we’re going to have to look at this in a hex editor! Let’s open up hexedit, but let’s also be careful not to edit it:

hexedit /media/kali/F7FD-F507/disk0.img

Alright, now we’re getting somewhere. This looks like a GPT style drive with some issues. Let’s decode this hexadecimal data and see what is happening. Immediately, we can see it’s missing the first GPT header that should say “EFI PART”, but it’s not missing the first volume table. Here’s the data prefixed by the offset in hexadecimal:

  • 0x00x200 – Protective MBR (this is ignored)
    • 0x1BE0x00 (non-bootable)
    • 0x1BF0x000200 (begin second sector)
    • 0x1C20xEE (GPT partition)
    • 0x1C30xFEBFD6
      • 0b11111110 10 111111 11010110 =>
      • 0b11111110 (Head: 254) 0b111111 (Sector: 63) 0b1011010110 (Cyclinder: 726)
      • (((726 * 255) + 254) * 63) + 63 = 11679255 sectors, 5.9 GB, 5.56 GiB
    • 0x1C60x01000000 (Preceeded by one relative sector, this one)
    • 0x1CA0xFFFFFFF (Ending sector, 2.2 TB, 1.99 TiB)
    • 0x1FE0x55AA (MBR Magic)
    • 0x200Where ye be? Coming up all zeroes.
  • Partition Table
    • 0x4000x480
      • C12A7328-F81F-11D2-BA4B-00A0C93EC93B – EFI System Partition – ESP
        • 0x28732AC11FF8D211BA4B00A0C93EC93B
      • 0x410 GUID: 0xBEB0DC7C3ECD4942988EC41374820D63
      • 0x420 Starting LBA: 0x0008 – offset 0x8000*0x200 = 0x1000000
      • 0x428 Ending LBA: 0xFFB704 – offset 0x4B7FF*0x200 = 0x96FFE00
      • 0x438 Label: EFI system partition
    • 0x4800x500
      • E3C9E316-0B5C-4DB8-817D-F92DF00215AE – Microsoft Reserved Partition – MSP
        • 0x16E3C9E35C0BB84D817DF92DF00215AE
      • 0x490 GUID: 0xA614A3694E39B94DAB59F5E477613D8C
      • 0x4A0 Starting LBA: 0x00B804 – offset 0x4B800 * 0x200 = 0x9700000
      • 0x4A8 Ending LBA: 0xFFB708 – offset 0x8B7FF * 0x200 = 0x116FFE00
      • 0x4B8 Label: Microsoft reserved partition
    • 0x5000x580
      • EBD0A0A2-B9E5-4433-87C0-68B6B72699C7 – Basic Data Partition – BDP
        • 0xA2A0D0EBE5B9334487C068B6B72699C7
      • 0x510 GUID: 0x6B28C785D0CEBC4CAC54A65CDC5DFF29
        • 85C7286B-CED0-4CBC-AC54-A65CDC5DFF29
      • 0x520 Starting LBA: 0x00B808 – offset 0x8B800 * 0x200 = 0x11700000
      • 0x528 Ending LBA: 0xFF7F6739 – offset 0x39677FFF * 0x200 = 0x72CEFFFE00 (this tracks with the size I know the volume to be)
      • 0x538 Label: Basic data partition
    • 0x5800x600
      • DE94BBA4-06D1-4D40-A16A-BFD50179D6AC – Windows Recovery Environment
        • 0xA4BB94DED106404DA16ABFD50179D6AC
      • 0x590 GUID: 0x065030C5029ED247A509427B34214010
        • C5305006-9E02-47D2-A509-427B34214010
      • 0x5A0 Starting LBA: 0x00806739 – offset 0x39678000 * 0x200 = 0x72CF000000
      • 0x5A8 Ending LBA: 0xFF6F8639 – offset 0x39866FFF * 0x200 = 0x730CDFFE00
    • 0x6000x680
      • DE94BBA4-06D1-4D40-A16A-BFD50179D6AC – Windows Recovery Environment
        • 0xA4BB94DED106404DA16ABFD50179D6AC
      • 0x610 GUID: 0xFE7A7B4441DE2246ABEBD01A11A66F4B
        • 447B7AFE-DE41-4622-ABEB-D01A11A66F4B
      • 0x620 Starting LBA: 0x00708639 – offset 0x39867000 * 0x200 = 0x730CE00000
      • 0x628 Ending LBA: 0xFF97723B – offset 0x3B7297FF * 0x200 = 0x76E52FFE00
    • 0x6800x700
      • DE94BBA4-06D1-4D40-A16A-BFD50179D6AC – Windows Recovery Environment
        • 0xA4BB94DED106404DA16ABFD50179D6AC
      • 0x690 GUID: 0x0D29D810B3D56D4092ED3F3B29FC2A7F
        • 10D8290D-D5B3-406D-92ED-3F3B29FC2A7F
      • 0x6A0 Starting LBA: 0x00A0723B = 0x3B72A000 * 0x200 = 0x76E5400000
      • 0x6A8 Ending LBA: 0xFFA79D3B = 0x3B9DA7FF * 0x200 = 0x773B4FFE00
  • 0x773C019E000x773C01A0AC – backup volume array (this matches the array above, but I’ve abbreviated it by partition type)
    • 0x773C019E000x773C019E80 – ESP
    • 0x773C019E800x773C019F00 – MSP
    • 0x773C019F000x773C019F80 – BDP
    • 0x773C019F800x773C01A000 – Recovery
    • 0x773C01A0000x773C01A080 – Recovery
    • 0x773C01A0800x773C01A100 – Recovery
  • 0x773C01DE000x773C01E000 – GPT Header (enitre raw data below w/o offset)
    • 0x4546492050415254 // EFI PART
    • 0x00000100 // Revision
    • 0x5C000000 // GPT header size
    • 0x14CC2471 // CRC32 of header (zeroed for calculation)
    • 0x00000000 // Reserved (zeroes)
    • 0xEF009E3B00000000
      • Current LBA: 0x3B9E00EF * 0x200 = offset 0x773C01DE00
    • 0x0100000000000000
      • Backup/original LBA: 0x1 * 0x200 = offset 0x200 (welp, that is missing)
    • 0x2200000000000000
      • First Usable LBA: 0x22 * 0x200 = offset 0x4400
    • 0xCE009E3B00000000
      • Last usable LBA: 0x3B9E00CE * 0x200 = offset 0x773C019C00
    • 0x2E85CB8C6AEB8B4D8635A53B968A0197
      • GUID: 8CCB852E-EB6A-4D8B-8635-A53B968A0197
    • 0xCF009E3B00000000
      • Starting LBA of array of entries: 0x3B9E00CF * 0x200 = offset 0x773C019E00 (which checks out with the backup table)
    • 0x80000000
      • Number of partition entries in array
    • 0x80000000
      • Size of a partition entry
    • 0x94580226
      • CRC32 of entries
  • 0x773C01E0000x773C01E196VolPort (?)
  • 0x773C241E000x773C242000Intel Raid ISM Error Log Sig (?)
  • 0x773C2420000x773C243000 – Lots of 0xFFFFFF00 followed by 0x265ABD5FDE57D801 – Test data? There’s some of this before this, too.
  • 0x773C255C000x773C256000Intel Raid ISM Cfg Sig.

Retrieving the data #

Now that we’ve decoded all the relevant records and partition tables, verified our partitions look right, and they are on the disk where the backup partition table says the should be, we can copy off the partition we want:

dd if=/media/kali/F7FD-F507/disk0.img of=/media/kali/F7FD-F507/bitlocker.dd bs=512 skip=571392 count=962512895
blkid /media/kali/F7FD-F507/bitlocker.dd

That looks right. It’s detecting a bitlocker partition. Next, we can use dislocker to decrypt the disk using the recovery key and check to make sure it worked and there’s an NTFS volume:

sudo mkdir -p /media/bitlocker/disk
sudo dislocker -V /media/kali/F7FD-F507/bitlocker.dd -p -- -o allow_other /media/bitlocker/disk
blkid /media/bitlocker/disk/dislocker-file

That also looks correct. Let’s go ahead and mount this NTFS volume:

udisksctl loop-setup -f /media/bitlocker/disk/dislocker-file
udisksctl mount -b /dev/loop1

Then lets check to see if the files are there:

ll /media/kali/OS/Users/jill/

This is looking great! Now let’s copy over the files to our backup disk:

# Copy the files
cp -ar /media/kali/OS/Users/jill /media/kali/F7FD-F507/
# Copy the WSL subsystem image
cp -ar /media/kali/OS/Users/jill/AppData/Local/Packages/CanonicalGroupLimited.UbuntuonWindows_79rhkp1fndgsc/LocalState/ext4.vhdx /media/kali/F&FD-F507/
# Cleanup
udisksctl unmount -b /dev/loop1
udisksctl loop-delete -b /dev/loop1
sudo umount /media/bitlocker/disk

If there are some files in the Windows Subsystem for Linux that we want to backup, we can get those files off the container image located somewhere under /Users/jill/AppData/Local/Packages/*/LocalState/*.vmdx:

# Let's get the files off the vhdx image
sudo apt-get install libguestfs-tools
sudo mkdir /media/vhdx
sudo guestmount --add /media/kali/F7FD-F507/OS/Users/jill/AppData/Local/Packages/CanonicalGroupLimited.UbuntuonWindows_79rhkp1fndgsc/LocalState/ext4.vhdx -i --ro /media/vhdx
sudo cp -ar /media/vhdx/home/adaburrows/workspace /media/kali/F7FD-F507/
sudo guestunmount /media/vhdx

Success! Thankfully no parts of the partition had been mangled during the logic circuit malfunction. I’m sure there’s other ways this could have failed which would make it much harder to ge the data off the chip on the SSD board.

Possibilities #

If there’s demand for this sort of thing, I could write a program that takes care of fixing up a disk (or disk image) like this. The basic steps it would go through are:

  • Check for the isw_raid_member using libblkid.
  • Check for the structure of the with MBR and GPT partition tables (maybe see if there’s any useful info in the newer isw metadata).
  • If one of the GPT partition tables is present and correct (seems to match what is on the disk),
    • then write the correct GPT header and partition tables,
    • else scan the disk and try to determine the correct partition table and the write the correct headers and partition tables.
  • Remove the isw_raid_member sector.
  • Remove the protective MBR partition that prevents the computer from booting off the disk.

That should be enough to get a non-damaged disk up a running again. For my purposes, I just needed this data and not a fully working disk.

Further Reading #