copying /dev/da0 with dd(1) to file: output differs

[Warren Block]

On Fri, 19 Nov 2010, Matthias Apitz wrote:

> El d?a Friday, November 19, 2010 a las 07:16:53PM +0100, Hans Petter Selasky escribi?:
>
>>> I was thinking in a tool just reading each file block by block,
>>> comparing the blocks and noting the 1st diff with block offset number.
>>> (some 10 lines of C code :-))
>>>
>>> 	matthias
>>
>> Maybe you need to write a small C-program to do that.
>>
>> You can use bcmp() to compare two buffers.
>
> Will do that tomorrow.
>
> Just an idea: The USB key in question was new and I only created the
> file system on it the usual way (fdisk, bsdlabel, newfs). Then I
> restored the dump on it (which took 26 hours for 3.1 GByte dump file).
> The USB key boots fine, btw.

26 *hours*?  USB 1.1?

> Could it be that unwritten/unformatted blocks are read as random data
> from that USB key?

Well, they could be anything.  But the original filesystem has 
formerly-used blocks with old data from deleted files.  dump doesn't 
copy those, but dd does.

> Should I overwrite the full USB key from /dev/zero?

Possibly there would still be differences.  Filesystem metadata like 
date last mounted, for example.  If you want a block-by-block duplicate, 
the brute-force method is to just dd the whole drive.  Use bs=64k or 
bs=1m to help reduce overhead.