Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 15 Sep 2018 21:10:30 +0200
From:      Polytropon <>
To:        Victor <>
Cc:        FreeBSD <>
Subject:   Re: Recover Fat32 files from a usb hd overwritten with a Freebsd image
Message-ID:  <>
In-Reply-To: <>
References:  <>

Next in thread | Previous in thread | Raw E-Mail | Index | Archive | Help
On Sat, 15 Sep 2018 19:00:44 +0200, Victor wrote:
> I'm risking to be hanged by the neck by my daughter!
> Well the fact is simple and dramatic. My daughter had a
> usb external hd, formatted as Fat32 filesystem,  where
> she was keping all her jpeg photos. 
> Accidentally I used this same usb drive to copy the
> FreeBSD-11.2-RELEASE-amd64-memstick.img using the classical
> sudo dd if=FreeBSD-12.0-RELEASE-amd64-memstick.img of=/dev/da0 bs=1m
> At the end of the process I became aware of the terrible
> mistake and I've never used the usb hd to start Freebsd.
> I wonder if I can try to recover the jpeg files from the
> 'extinct' fat32 filesystem. Surfing the net I found many
> solutions and I'm somewhat confused (photorec?).

Before I will repeat my "famous list of magical recovery tools",
let me tell you that the chances to recover files are nearly
zero. The reason: Data that has been overwritten does not
exist anymore. Overwriting data is very differerent from just
erasing the FAT, or anything at the beginning of a disk. But
depending on how the files are scattered across the disk, it
might still be possible to recover something outside of the
area where the FreeBSD image has been written to. FAT is known
for fragmentation, so having files "at the end of the disk"
is nothing strange.

By the way, are you sure it's FAT? Who uses FAT today? In
"Windows" land, external disks tend to be formatted with
NTFS, so in addition to the following information, you
might consider checking if tools from the "ntfsprogs"
package (here: ntfscat, ntfsinfo, ntfsclone, ntfsfix,
or ntfsundelete) could be usable.

> In order to avoid the hanging, is there anyone of you experts
> able to give me a useful hint for this specific case?

Because the FAT is gone, you need to use recovery tools that
process the disk as-is (read: block-wise, byte-wise, not
caring for file information that isn't there anymore).

Before you start, make sure you _never_ write to the disk
you're trying to recover from. If you can, make an 1:1 image
of the disk, and work with the image. But it's not a problem
if you carefully follow the instructions, as none of them
will write to the source for recovery.

The tools I would suggest you make yourself familiar with
will be marked with an arrow. Please understand that there is
no "one size fits all egg-laying wool-milk sow one-click GUI
program to undo all imaginable mistakes". Make sure that you
understand what you're doing. Yes, this takes time, and you
will learn a lot about file systems, extending and advancing
your skillset. This is what people mean when they tell you
to "learn from your mistakes". ;-)

	dd				<--- Make copy if needed.
	fetch -rR <device>

	magicrescue			<--- Try this!
	testdisk			<--- Maybe this!
	The Sleuth Kit:			<--- In worst case.
	recoverjpeg			<--- Or this.
	photorec			<--- Or this.
	fatback				<--- Probably won't work.

So here's an example procedure that you could adopt to the
correct device names etc. - please read carefully and decide
which steps you need. There are a few assumptions, but you
will clearly recognize their value as placeholders. Also
make sure you have sufficient disk space (usually 2 * disk
size), where both the disk image and the rescued files will
be stored; less is needed if you don't make a copy.

And always apply: "First think, then act." ;-)

This is how you can use MagicRescue:

	# cd /scratch
	# mkdir out
	# dd if=/dev/da0 of=disk.dd bs=1m
	# magicrescue -r /usr/local/share/magicrescue/recipes -d out/ disk.dd

Or to directly read from the disk (which is no problem if
the disk isn't mechanically defective, so there is no urgent
need for a 1:1 copy):

	# magicrescue -r /usr/local/share/magicrescue/recipes -d out/ /dev/da0

Programs like TestDisk and PhotoRec are text-based dialog
programs which explain the decisions and selections to be
made. As always, consult the manpages if needed.

Always remember: Even if you get the file data back, the
FILE NAMES ARE GONE. If you need a tool that postprocesses
the images and turns them into something timestamped (if
the images contained EXIF data that could be used to get
the timestamp!), contact me offlist to receive a script
which does exactly that.

As you can guess by now, I've been hanged several times,
and now I know a little bit about data recovery. ;-)


I will finally attach the content of the file "ref_fs.txt"
that was once part of TSK before they (in my opinion) stupidly
decided to remove it from their locally installed documentation
and put it into some online Wiki. There you will find the
instructions needed to use TSK, the tool that - in worst case -
will help you avoid the hanging. ;-)

                    File System Analysis Techniques
                     Sleuth Kit Reference Document 

                            Brian Carrier
                        Last Updated: July 2005

Currently, evidence is most frequently found in the file system.
This is because it is non-volatile and remnants of deleted files
can typically be found.  This file will help one to use the low-level
tools in The Sleuth Kit for a forensic analysis.

This document is organized into small scenarios, which provide
examples of how to use The Sleuth Kit.  Most of these functions
are automated with Autopsy, but they are here for reference and

The techniques used here apply to both UNIX and Windows file systems.

The steps from the timeline Sleuth Kit Implementation Notes are
followed (using both ils and fls) and you notice some interesting
activity from unallocated inodes, namely MFT Entry 5035 from image
c_drive.dd.  To display the contents of this file, use "icat":

    # icat images/c_drive.dd 5035 | less

NOTE: To prevent your terminal from getting messed up, pipe all
output of "icat" through a pager like "less".

In this scenario, we will search the unallocated space of the
"wd0e.dd" image for the string "abcdefg".  The first step is to
extract the unallocated disk units using the "dls" tool (as this
is an FFS image, the addressable units are fragments).

    # dls images/wd0e.dd > output/wd0e.dls

Next, use the UNIX strings(1) utility to extract all of the ASCII
strings in the file of unallocated data.  If we are only going to be
searching for one string, we may not need to do this.  If we are going
to be searching for many strings, then this is faster.  Use the '-t d'
flags with "strings" to print the byte offset that the string was found.

    # strings -t d output/wd0e.dls > output/wd0e.dls.str

Use the UNIX grep(1) utility to search the strings file.  

    # grep "abcdefg" output/wd0e.dls.str | less
    10389739: abcdefg

We notice that the string is located at byte 10389739.  Next,
determine what fragment.  To do this, we use the 'fsstat' tool:

    # fsstat openbsd images/wd0e.dd
    Fragment Range: 0 - 266079
    Block Size: 8192
    Fragment Size: 1024

This shows us that each fragment is 1024 bytes long.  Using a
calculator, we find that byte 10389739 divided by 1024 is 10146
(and change).  This means that the string "abcdefg" is located in
fragment 10146 of the "dls" generated file.  This does not really
help us because the dls image is not a real file system.  To view
the full fragment from the dls image, we can use dd:

    # dd if=images/wd0e.dd bs=1024 skip=10146 count=1 | less

Next, we will identify where this fragment is in the original image.
The "dcalc" tool will be used for this.  "dcalc" will return the
"address" in the original image when given the "address" in the
dls generated image.  (NOTE, this is currently kind of slow).  The
'-u' flag shows that we are giving it an dls address.  If the '-d'
flag is given, then we are giving it a dd address and it will
identify the dls address.

    # dcalc -u 10146 images/wd0e.dd

Therefore, the string "abcdefg" is located in fragment 59382.  To view
the contents of this fragment, we can use "dcat".

    # dcat images/wd0e.dd 59382 | less

To make more sense of this, let us identify if there is a meta data
structure that still has a pointer to this fragment.  This is achieved
using "ifind".  The '-a' argument means to find all occurrences.

  # ifind -a images/wd0e.dd 59382

Inode 493 has a pointer to fragment 59382.  Let us get more information
about inode 493, using "istat".

    # istat images/wd0e.dd 493
	inode: 493
	Not Allocated
	uid / gid: 1000 / 1000
	mode: rw-------
	size: 92
	num of links: 1
	Modified:       08.10.2001 17:09:49     (GMT+0)
	Accessed:       08.10.2001 17:09:58     (GMT+0)
	Changed:        08.10.2001 17:09:49     (GMT+0)
	Direct Blocks:

Next, let us find out if there is a file that is still associated with
this (unallocated) inode.  This is done using "ffind".

    # ffind -a images/wd0e.dd 493
	* /dev/.123456

The leading '*' identifies the file as deleted.  Therefore, at one point,
the file '/dev/.123456' allocated inode 493, which allocated fragment
59382, which contained the string "abcdefg".

If "ffind" returned with more than file that had allocated inode 493,
it means that either both were hard-links to the same file or that one
file (chicken) allocated the inode, it was deleted, a second file (egg)
allocated it, and then it was deleted.  The string belongs to the second
file, but it is difficult to determine which came first.  On the other
hand, if "ffind" returns with two entries where one deleted and one not,
then the string belongs to the non-deleted file.

As previously mentioned, Autopsy will do all of this for you when
you do a keyword search of unallocated space.  

To view all of the deleted file names in an image, use the "fls" tool.
For all deleted files, use the '-r' flag for recursive and '-d' flag
for deleted.

    # fls -rd images/hda9.dd | less
    d/d * 232: 	/TEMP-823450
    r/d * 293: 	/TEMP-131100
This shows us the full path that the deleted files are located.  On some
systems, such as Windows NTFS, the file content may be recovered
(depending on how much system activity has occurred).  On other
systems, such as Solaris UFS and Linux Ext3, deleted files can not
be easily recovered.   The number at the beginning of the line is
the inode number.  The '*' shows that it is deleted and the 'd' and
'r' show the type (directory and file).  The first letter identifies
the directory entry type value (which does not exist in all file
system types) and the second letter is the type according to the
inode.  In most cases these should be the same, but it may not for
deleted files if the inode has been reallocated to a file of a
different type.  If we do an "istat" on the directory (232) we will
notice that the size is 0.

    # istat images/hda9.dd 232
	inode: 232
	Not Allocated
	uid / gid: 0 / 0
	mode: rwxr-xr-x
	size: 0
	num of links: 0
	Modified:       08.23.2001 21:52:33     (GMT+0)
	Accessed:       08.23.2001 23:05:39     (GMT+0)
	Changed:        08.23.2001 21:52:33     (GMT+0)
	Deleted:        08.23.2001 23:05:39     (GMT+0)
	Direct Blocks:

Linux does this to all of its deleted directories.  It should also
be observed that no block addresses are shown in the "istat" output.
This is because the size is 0 and the program thinks that the address
is bogus.  Using the '-b' option of "istat", we can force it to
output the block address.  With Linux Ext3, the block pointers would
be 0, but Linux Ext2 kept the old addresses.

    # istat -b 2 images/hda9.dd 232
	inode: 232
	Not Allocated
	uid / gid: 0 / 0
	mode: rwxr-xr-x
	size: 0
	num of links: 0
	Modified:       08.23.2001 21:52:33     (GMT+0)
	Accessed:       08.23.2001 23:05:39     (GMT+0)
	Changed:        08.23.2001 21:52:33     (GMT+0)
	Deleted:        08.23.2001 23:05:39     (GMT+0)
	Direct Blocks:
	  388 0

Now we can examine the contents of block 388 and see the file
names that were in that directory:

    # dcat -h images/hda9.dd 388 | less

A UFS/FFS or EXT2FS/EXT3FS file system is organized into groups.
Each group has its own inodes and blocks to store data in.  When
a new file is created, it is given an inode in the same group that
the parent directory inode is in (if there are still inodes
available).  When a new directory is created, it is given an inode
in a new group.  An inode allocates blocks from the same group that
its inode is in.

When recovering a file from one UFS or EXTxFS, the group layout
can be used.  When a deleted file is found with 'fls', notice the
inode of the parent directory:

    # fls -r images/hda1.dd
	d/d 30789:      doc
    + r/r * 0:    doc/.a/ssh.tar
    + r/r 30792:    doc/.a/install

We want to recover the 'ssh.tar' file and notice that the parent
directory is 30789 and the deleted file has a cleared inode pointer.
To identify the group that it is in, the 'fsstat' tool is used:

    # fsstat images/hda1.dd
    File System Type: EXT3FS

    Group: 0:
      Inode Range: 1 - 15392
      Block Range: 0 - 32767
        Super Block: 0 - 0
        Group Descriptor Table: 1 - 1
        Data bitmap: 2 - 2
        Inode bitmap: 3 - 3
        Inode Table: 4 - 484
        Data Blocks: 485 - 32767

    Group: 1:
      Inode Range: 15393 - 30784
      Block Range: 32768 - 65535
        Super Block: 32768 - 32768
        Group Descriptor Table: 32769 - 32769
        Data bitmap: 32770 - 32770
        Inode bitmap: 32771 - 32771
        Inode Table: 32772 - 33252
        Data Blocks: 33253 - 65535

    Group: 2:
      Inode Range: 30785 - 46176
      Block Range: 65536 - 98303
        Data bitmap: 65536 - 65536
        Inode bitmap: 65537 - 65537
        Inode Table: 65540 - 66020
        Data Blocks: 65538 - 65539, 66021 - 98303


The inode is in the range of inode addresses for group 1.  To search
for the deleted file, we extract the unallocated space using 'dls':

    # dls images/hda1.dd 32768-65535 > output/hda1-grp1.dls

If we wanted to extract all of the data for the group, we could
use 'dd':

    # dd if=images/hda1.dd of=output/hda1-grp1.dd bs=4096 skip=32768 \

Where, the fragment size is 4096 (which can also be found in the
'fsstat' output).  Either of these images can then be analyze for
keywords or using other data carving tools such as 'foremost'.
This process allows one to reduce the amount of data that must be

Copyright (c) 2002-2005 by Brian Carrier.  All Rights Reserved
CVS Date: $Date: 2007/12/18 22:43:30 $

Magdeburg, Germany
Happy FreeBSD user since 4.0
Andra moi ennepe, Mousa, ...

Want to link to this message? Use this URL: <>