On Fri, Oct 31, 2003 at 08:12:19PM +0200, Jukka Marin wrote:
> 
> True, but it seems that writing to CF is unreliable.  Every dd produces
> a different copy - reading the CF card using dd works reliably.  The
> error in write seems to occur at offset (disk block boundary + 1) bytes
> (like 0x06c00001 or 0x00150001).  After this point, the rest of the CF
> contents are wrong (as if part of the data written was skipped or
> something).

Some CF cards do not write correctly in "controller mode" (the way they
operate when connected to a PCMCIA bus); it's possible (though I haven't
seen it) that some don't write correctly in "disk mode", that is, when
directly connected to an IDE controller.  I've also encountered a situation
in which a particular CF card required 12V to write, but the PCMCIA 
controller in question did not supply 12V to the socket by default and the
operating system in use didn't enable 12V either.  The result was exactly
what you describe... so you might look into this.