Re: [Bug-ddrescue] Reading past errors on a non-scsi drive

[James Bardin]

Thanks,

I've tried many combinations in the past, but this time with a new intel 
controller, and a 2.6.19 kernel, I got through. I know I've tried this 
with older intel controllers, and older kernels, and had no luck. I 
guess whatever I needed was fixed in the past year or two.

DMA wasn't the problem, and it gets turned off when there is an ide 
reset (after a bunch of AddrNotFound errors).
It helped to turn dma back on after a bad area on the disk, and quickly 
get to the next set of errors.



One minor point:
When I start ddrescue with a log, it reports the error size equal to the 
recovered size. Still runs fine, and error size is reported correctly 
thereafter.

Thanks
-jim




Dave Burton wrote:
> James, the problem is probably not the defective disk drive,
> but, rather, the IDE controller you are using it with.  The
> disk drive won't wait forever.  It might be up to 10-15
> seconds, but not forever.  "Promise" brand controllers are
> especially bad -- they will often either freeze or just crash
> the computer if they hit a bad block on the disk.  Use a
> different computer and the problem will probably go away.
>
> Alternately, you can try disabling DMA; e.g.,
>
> hdparm -d0 /dev/hdc
>
> That might also avoid the problem, though it'll slow down the
> data recovery process, too.
>
> -Dave
>
>
>
> address@hidden (James Bardin) wrote:
>
>   
> > Hello,
> >
> > This may be a feature request, or just advice is it's already possible.
> >
> > I'm trying to read through a disk in which there are errors that cause
> > the disk to wait indefinitely. (I've had to recover many disks in this
> > state, so I'm surprised that I can't find a info on this)
> > When this happens on a scsi disc, the read times out, the controller
> > dumps the memory, and we move on.
> > On other types of disks, we hang. I've tried reading both from the block
> > device, and raw character devices.
> >
> > My current solution is to use a usb adapter, and pull the connection
> > when it stops reading, then start again from a different block. This is
> > fine when there a few bad areas, but I'm looking for an automated
> > solution for more serious recovery.
> >
> > I'm not sure if this is even possible without additional features in the
> > kernel. I've been following a related thread on lkml about a true forced
> > unmount that would drop pending io:
> > http://www.gossamer-threads.com/lists/linux/kernel/745783
> >
> > Or, if someone knows a way to forcefully remove, and reinitialize a usb
> > device via software, I could use that as a workaround for now.
> >
> > Thanks
> > -jim
> >     
>
>   

--

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
James Bardin - Systems Analyst / Administrator I
Boston University Department of Electrical and Computer Engineering
8 Saint Mary's St, Room 305, Boston, MA 02215
Ph:617-358-2785         http://www.bu.edu/ece/it
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */