Wednesday, May 14, 2008

HOWTO - Work with Disk Images

Warning! Extra Geeky Content Ahead!

What is a disk image?

The way I'm using it, a disk image is a bit-by-bit copy of the information on a data storage device. For a broader exploration of the topic, check out the "Disk image" article over at the Wikipedia.

I'll talk about hard disk images, and CD images (the infamous .iso file).

Why would I care about disk images?

Maybe you wouldn't. Mostly, I think we want to interact with our disks through the standard filesystem tools to work with files.

There ARE several situations where working with disk images might be useful. Among them might be:

  • Downloading and burning the latest Ubuntu Live CD.
  • Copying a CD, to distribute or to have a backup copy just in case.
  • Accessing the contents of these CD images without burning them onto CD.
  • Duplicating smaller disks (e.g. SD cards, CompactFlash cards)
  • Saving a perfect data copy of a disk as a first step in sensitive situations involving data (e.g., data recovery or forensics work).
  • Keeping an archival copy of a disk (e.g. I have a disk image of the 120MB hard disk from my old 386 computer for nostalgia purposes; one of these days I'm going to figure out how to emulate it...).

Note that while technically possible, imaging a large number of computers this way is a very long process compared to tools that are specifically designed for that sort of work (such as partimage and the SystemImager suite).

What tools do we need?

In UNIX (and Linux, and BSD, OSX, etc.), we will use the dd utility to access data from the storage device directly (as opposed to via standard file manipulation, which is (thankfully) abstracted from direct access through concepts like partitioning and filesystems) in order to be able to read and write disk images.

dd, or a version of it, is available for every platform, I believe.

I'll also discuss using the UNIX mount utility to mount the disk images as if they were real disks. I believe Mac OSX has a similar functionality built right in, and I know there are all kinds of programs for mounting virtual drives in Windows.

Creating a Disk Image

You'll need the disk you're copying from to not be mounted. You'll also need a place to save the image file that has sufficient free space. This may mean:

  • You're copying from a smaller disk to a larger disk—with more free space than the smaller disk has total capacity— neither of which is mounted on your computer.
  • You're booted into a Live CD environment in order to have un-mounted access to your main (operating system) disk in order to be able to copy it off onto an external storage device, be that over USB or over the network.

Whatever the case, once you have things ready, the syntax we use is:

# dd if=input of=output

Where input is the disk device node and output is the file you want to write the disk's data to.

SO, if I want to take /dev/sda and make a disk image of it called sda.dd in the current working directory, I run:

# dd if=/dev/sda of=sda.dd

And now I have a file named sda.dd which contains an exact bit-by-bit copy of my /dev/sda disk!

Writing a Disk Image to a Disk

So, you created a disk image of your drive, and then you did something stupid and ruined the contents of your drive? or your drive died and you got a new one? or (more optimistically) you're just duplicating the hard disk and that's why you have an image? No worries, we can simply write the image right back onto that disk!

We'll need to have access to the disk image file, and the destination disk will need to be available and unmounted.

The pattern for dd stays the same:

# dd if=input of=output

Except now input is the disk image file and and output is the disk device node you want to write the disk's data to.

SO, if I want to take /dev/sda and write a disk image called sda.dd onto it, I run:

# dd if=sda.dd of=/dev/sda


Incidentally, you can also do this with a partition rather than a full disk, by giving it a partition node instead of a disk node (e.g. /dev/sda1 is the first partition on /dev/sda) so dd will create an image of just the first partition rather than one of the full disk).

CD Images

Compact Disc images have been made really easy to work with.

In Ubuntu, you need only to right-click on a .iso file in order to be presented with the option to Write to Disc.... Also, using Brasero, you can run a "Disc copy" project and copy your disc to a "File image".

Alternatively, on the command-line, CD images work just like you might expect from the above. In order to create them you can just:

# dd if=/dev/scd0 of=discimage.iso

And in order to burn the disc image to a blank disc, you can just reverse the direction:

# dd if=discimage.iso of=/dev/scd0

Accessing Disk Image Partitions

So, we can write the disk images onto disks...

We can also access and manipulate them directly by treating the image file as a disk. I'll be discussing the use of the UNIX mount utility, which is responsible for mounting disks onto the UNIX file system, to do this.

The pattern for using mount (for this) is:

# mount -t fstype -o options device directory


  • fstype is the type of filesystem you're trying to mount; necessary if you're working with non-native filesystems, like NTFS.
  • options are extra options you may need to use, e.g. loop is the option we feed mount to let it know we're feeding it a disk image file rather than a real disk. If we want to mount a partition from inside a full disk image, we'll also need to use the offset option to let it know where the partition starts.
  • device is the device (in our case disk image) to be mounted.
  • directory is the destination directory in the UNIX file system where you want the image mounted.

So, in order to mount a CD iso, we just do:

# mount -o loop cdimage.iso /media/iso

And now running ls /media/iso will show us the root directory of the CD.

On Ubuntu 8.04 (and others), so long as you mount it in the /media directory, an icon will show up on the GNOME desktop to represent the "disc".

If we have the image of a disk partition, we can similarly mount it as above with:

# mount -o loop partition.img /media/partition

However, if we want to mount a partition from inside a full disk image, we'll first need to locate the spot in the image that the partition begins.

We need to use fdisk to get the needed information out of the disk image, with the -l option to list partition information and the -u option to show us the sizes in sectors.

For example, as I was working on recovering data off of a teacher's dying hard drives, I got this response:

root@om:/mnt# fdisk -lu output 
You must set cylinders.
You can do this from the extra functions menu.

Disk output: 0 MB, 0 bytes
255 heads, 63 sectors/track, 0 cylinders, total 0 sectors
Units = sectors of 1 * 512 = 512 bytes
Disk identifier: 0x9dc96e9e

 Device Boot      Start         End      Blocks   Id  System
output1              63       80324       40131   de  Dell Utility
output2   *       80325   464840774   232380225    7  HPFS/NTFS
Partition 2 has different physical/logical endings:
     phys=(1023, 254, 63) logical=(28934, 254, 63)

So, from the above we can tell that the second partition in the disk image "output" starts at sector 80325.

We also know that each sector is 512 bytes.

Multiply the two and we know that the second partition starts at byte number 41126400.

We can also tell the partition type is NTFS (It's a Windows XP partition...).

I created a directory called /mnt/C to be the mount point for the NTFS partition, and plugged all the information in the right order for mount:

root@om:/mnt# mount -o loop,offset=41126400 -t ntfs output C

And voilà!

the second partition of the disk image file "output" is now mounted on my system under the /mnt/C folder!

I proceed to grip all that teacher's years of work and yank it back from the jaws of oblivion. (For this, I just used graphical filesystem tools to copy out his Documents and Settings folder)


There are certain situations where knowing how to work with raw disk data can be useful.

These are the things I had to learn to figure out how to do it right. Now I've got a reference to look back on the next time I have to do it.

Also, hopefully others will find it useful or at least mildly entertaining.



Anonymous Anonymous said...

Nice article :)

You either missed or never encountered a need to create a raw disk image, say for an SD Card, USB stick or CompactFlash.

How do I create one or more msdos/vfat or other partitions on raw disk image using dd like:

dd if=/dev/zero of=disk.img bs=512 count=1984k

4:24:00 PM  
Blogger Simón A. Ruiz said...


You can partition a raw disk image that was created using dd with the same tool you use to partition a normal disk, fdisk.

The problem I've had is that fdisk usually gets its geometry values directly from the disk it's trying to work with, and these files aren't actual disks, so that information needs to be given to fdisk manually.

The above is a page I found that seems to address the issue pretty well, starting from step 3.

I imagine the image file you got from using dd is functionally the same as what they got from the first couple of steps in that post, except for size...

Hope this helps.


8:26:00 AM  

Post a Comment

Links to this post:

Create a Link

<< Home