Booting NOT-Windows

Stephen Ryan stephen at sryanfamily.info
Thu Aug 21 09:44:31 EDT 2008


On Thu, 2008-08-21 at 08:16 -0400, Neil Joseph Schelly wrote:
> On Wednesday 20 August 2008 20:13, Bill McGonigle wrote:
> > So, one thing I like to do is to create a disk image of the damaged
> > disk before trying anything else.  That way you can go back if
> > 'recovery' attempts do more damage than good.  This is largely a
> > question of what your data is worth and what your risk aversion is
> > like.  NewEgg recently had $500GB disk on sale for $50, so we're not
> > talking very expensive here.
> 
> I've troubleshooting a lot of Windows 2000 and XP laptops lately for tasks 
> that require moving filesystems from one hard disk to another.  I have found 
> that my trusty dd methods really don't work that well, at least not wth a 
> suspect drive.
> 
> To move a partition from one disk to another, I was using dd (and sometimes 
> just cat) with bzip2 and netcat to rip an image from one disk and dump it to 
> another disk in another laptop in realtime.  I found that while most of the 
> filesystem arrived at its destination intact, many parts of it did not.  I 
> could boot up the old drive with a Knoppix disk and get access to those 
> files, retransmit them as files over sftp/scp or something, and then they'd 
> work, but in the image, they were not kept intact.
> 
> Has anyone else run into this or know what causes this?  I guess I never 
> bothered with the "bs=8M conv=noerror,sync" options you specified.  Do you 
> think those would have made a difference?

Yes it does.  Without those options, dd is nearly useless; with them,
you have a chance of getting useful things off the disk.  

If there is an error while reading the source disk, "conv=noerror,sync"
ensures that dd tries to read the rest of the disk, and that it fills in
the spot with the read error with zeroes, rather than skipping it
entirely, which shifts the rest of the disk image over, thus destroying
every file stored after the point of the error. The "bs=8M" gives dd a
block size to work with; it should match the sector size of the source
disk so that if an error does occur during reading, only that sector is
lost.

> -N
> _______________________________________________
> gnhlug-discuss mailing list
> gnhlug-discuss at mail.gnhlug.org
> http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/



More information about the gnhlug-discuss mailing list