Rescue broken HDD

[schneida]

Hello everybody,

I was asked by a friend to help him out with a broken HDD. Its a 2.5" Sata drive of a laptop wich was dropped and now you can clearly hear that something is wrong although you can still acess the drive. I started rescueing the whole disk with ddrescue (GNU) using the following command:

ddrescue --no-split /dev/sdb /media/home/image01 logfile

The problem is that ddrescue is damn slow (about 5 KB a second, at this rate it would take me 44 days to complete the rescue process) and I guess that some kind of kernel abstraction prevents ddrescue from skippping bad sectors. I have no proof for that and it's just a guessing but I imagine that ddrescue accesses the drive through the kernel which interacts with the drive and as the drive returns bad data, the kernel tries to re-read that data until it finally gives up and returns command to ddrescue which then continue with the next block.

Am I right with my guess and is there any way of disabling such mechanisms? I read that there is a "--direct" option for ddrescue but the manual states that it's probably taking more time then..?

In case you need to know something about my kernel. I'm using archlinux kernel 2.6.36.

Dmesg outputs the following:

ata6.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
ata6.00: irq_stat 0x40000001
ata6.00: failed command: READ DMA
ata6.00: cmd c8/00:00:e0:16:05/00:00:00:00:00/e0 tag 0 dma 131072 in
         res 51/40:29:b7:17:05/00:00:00:00:00/e0 Emask 0x9 (media error)
ata6.00: status: { DRDY ERR }
ata6.00: error: { UNC }
ata6.00: configured for UDMA/100
ata6: EH complete
ata6.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
ata6.00: irq_stat 0x40000001
ata6.00: failed command: READ DMA
ata6.00: cmd c8/00:00:e0:16:05/00:00:00:00:00/e0 tag 0 dma 131072 in
         res 51/01:29:b7:17:05/00:00:00:00:00/e0 Emask 0x1 (device error)
ata6.00: status: { DRDY ERR }
ata6.00: configured for UDMA/100
ata6: EH complete
ata6.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
ata6.00: irq_stat 0x40000001
ata6.00: failed command: READ DMA
ata6.00: cmd c8/00:00:e0:16:05/00:00:00:00:00/e0 tag 0 dma 131072 in
         res 51/40:29:b7:17:05/00:00:00:00:00/e0 Emask 0x9 (media error)

Any help is appreciated!!

[henrikau]

as a general rule of thumb - whenever you do this, try to make an *exact* copy of the disk before you do anything else!

dd if=/dev/sdb of=/media/home/image01

If this succeeds, you should have all the data in a binary file which you then can mount as a disk via the loopdev like so

mount -o loop media/home/image01 /some/target/folder/

[schneida]

Thanks for your reply. I did try to copy files with cp and then with dd first but as both just returend I/O Errors I started to look for an alternative.

I originally planned to use dd_rescue with dd_rhelper but on the author's homepage I read that ddrescue should be preferred as it is probably faster and not as quirky as the shell script is.

I'm now with 670 MB (of 192 GB) and got about 260 errors. I was faster (actually the average rate at the moment is at 45 kb/s) but now I' down at 102 B/s again....

EDIT: Ok, I think I've got all bad sectors behind me now as the data rate went up to 40 MB/s and I've got about 40 GB rescued data.