[Qemu-devel] qemu dd sizes

[Qemu-devel] qemu dd sizes

[jmf]

When using dd to create a disk image to boot through qemu, does the block size really matter, or can I omit that parameter?  If the size of the original disk is 20 gb but there is only 2 gb of data on the disk, how can I make the image size smaller? 





_______________________________________________
Join Excite! - http://www.excite.com
The most personalized portal on the Web!

Re: [Qemu-devel] qemu dd sizes

[Kyle Hayes]

On Friday 24 September 2004 19:15, jmf wrote:
> When using dd to create a disk image to boot through qemu, does the
> block size really matter, or can I omit that parameter?  If the size of
> the original disk is 20 gb but there is only 2 gb of data on the disk,
> how can I make the image size smaller?

It is a little unclear what you are asking here.   dd uses the block size 
and a count in order to determine how much to copy.  If you are making a 
file with holes in it, you can use a block size, offset and then set 
count=0.  If you omit the count and block size, it will use the default 
block size (512 bytes I think) and will simply copy until either the 
source runs out of data or the destination runs out of space.

If there is 2GB, it depends what the original filesystem was.  Ext2/3 
(under Linux) puts data all over the disk in a sparse manner.  So, just 
copying the first 2GB would not get all the data.  If it is a Windows 
partition and you are using VFAT, then running the defragmenter program 
will push all the data to one end of the disk (beginning?).  With NTFS, 
I'm not sure what happens.

You could try to resize the 20GB partition down to 2.5GB and then copy that 
with dd.  What kind of filesystem is it?

Best,
Kyle

Re: [Qemu-devel] qemu dd sizes

[jmf]

It's UFS, FreeBSD. I think UFS does a good job keeping all the data in a non-sparse manner. 



I could resize the UFS partition down to a smaller size and then use dd, however, I want a faster operation that will not alter my original installation. I was hoping there is a way to manipulate the images that dd produces, for more convenient use with qemu.





_______________________________________________
Join Excite! - http://www.excite.com
The most personalized portal on the Web!

Re: [Qemu-devel] qemu dd sizes

[Paul Brook]

On Saturday 25 September 2004 14:08, jmf wrote:
> It's UFS, FreeBSD. I think UFS does a good job keeping all the data in a
> non-sparse manner.
>
>
>
> I could resize the UFS partition down to a smaller size and then use dd,
> however, I want a faster operation that will not alter my original
> installation. I was hoping there is a way to manipulate the images that dd
> produces, for more convenient use with qemu.

On linux you can mount images using the loopback device. You may need to tell 
it to skip the partition table (the first track, usually 32256 bytes).

I usually find it easier and more reliable to format/partition the image 
inside the guest, then mount it via loopback device on the host and use 
normal tools (cp, tar, etc) to copy the files across.

Paul

Re: [Qemu-devel] qemu dd sizes

[Bartosz Fabianowski]

> On linux you can mount images using the loopback device.

You can do the same on FreeBSD, although you use a different command.

So one way to go about this would be as follows. Boot up the native 
FreeBSD. Create an empty image file of appropriate size - say 4GB. Mount 
this image and copy the installed FreeBSD to the image. While doing this 
using native tools, remember to skip copying the image file onto itself, 
as that would result in a recursion you don't want. Finally, move the 
image you just created to your host system. This might not be the 
quickest way of doing it, but it seems simple enough.

- Bartosz

Re: [Qemu-devel] qemu dd sizes

[Mark A. Williamson]

> If the size of the
> original disk is 20 gb but there is only 2 gb of data on the disk, how can
> I make the image size smaller?

If you're trying to save space, I'd advise you to check the output of ls -ls.  
The -s flag also prints out the actual space taken on disk, as well as the 
nominal length of the file usually output by -l.

Most filesystems will represent zero-filled areas of a file with a small 
constant amount of on-disk space.  If you're only using 2GB then the actual 
on disk size might only be just over 2GB.  As you add more data, the on disk 
size will grow to fit.

It's worth checking this before you bother resizing!

Cheers,
Mark

>
>
>
>
>
> _______________________________________________
> Join Excite! - http://www.excite.com
> The most personalized portal on the Web!
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel

Re: [Qemu-devel] qemu dd sizes

[Bartosz Fabianowski]

[This was sent to me off-list]

> This was what I was looking for, however, I'm a little confused about
> how to mount the file through the loopback device. Mount normally
> requires a filesystem type as a parameter but if I create the "empy
> image" with dd if=/dev/zero, there is no filesystem in the image. I
> can't use newfs on the file until I mount it with the loopback
> device..... How do I make the first mount ?

Quite simple really. First, create an empty disk image:

# dd bs=1 seek=4194303 count=1 if=/dev/zero of=image
1+0 records in
1+0 records out
1 bytes transferred in 0.000104 secs (9620 bytes/sec)

Then, create a loopback device:

# mdconfig -a -t vnode -f image
md0

Note the previous command told you what device name it created. Now, run 
makefs on that device:

# newfs /dev/md0
/dev/md0: 4.0MB (8192 sectors) block size 16384, fragment size 2048
         using 4 cylinder groups of 1.02MB, 65 blks, 192 inodes.
super-block backups (for fsck -b #) at:
  160, 2240, 4320, 6400

Finally, you can mount the disk:

# mount /dev/md0 /mnt

As you can see, the file system is there, fully functional:

# df -H /mnt
Filesystem    Size    Used   Avail Capacity  Mounted on
/dev/md0      3.8M    4.1k    3.5M     0%    /mnt

To clean up things, you'd first unmount the disk:

# umount /mnt

Then, you'd free the loopback device:

# mdconfig -d -u md0

That's all you need to do. Note however, that mdconfig is a FreeBSD 5.x 
/ 6.x tool. If you're still on 4.x, you will need to use vnconfig 
instead (it's not as intuitive, but quite easy to use).

- Bartosz