Re: [Bug-ddrescue] Bug-ddrescue Digest, Vol 87, Issue 15

[Felix Ehlermann]

Dear Garegin,

I was very surprised to hear about this apparently known issue of badblocks.
I've been using the tool over the past decade to erase disks which were 
to be disposed or to be sent in for warranty and I don't remember any 
case where it would not have finished the next day.
Google was not very helpful in providing me with any substantiated 
reports about performance issues either.
Unfortunately I also did not keep track of the required time or the size 
of the drives I had wiped in the past, so I actually started a badblocks 
run on a 1tb-disk in my "ddrescue-box" last night.
It's an older Samsung drive (S-ATA, 5.4krpm, HD103**) which I normally 
use to write my ddrescue-created images to in order to do further 
analysis or repairs.
According to iotop badblocks wrote to the disk starting at ~110 MByte/s, 
eventually slowing down to ~60 MByte/s at the "end" of the drive.
After ~3 hours badblocks started over and was read from the drive 
(starting at 110 MByte/s again).
The performance of the drive reported by "hdparm -t /dev/sda" is 
basically the same.

address@hidden:~# date && badblocks -w -t random /dev/sda && date
Tue Apr 30 22:01:54 CEST 2013
Wed May  1 04:23:02 CEST 2013

Obviously the entire run took ~ 6h 20 minutes.

Assuming that you invoke badblocks with "badblocks -w /dev/sdx" it will 
iterate through the behavior described above four times (using the 
default patterns 0xaa, 0x55, 0xff, 0x00 according to manpage).
This would still finish after ~ 1 day on my 1tb drive.

This is indeed slow compared to a simple dd if=/dev/zero of=/dev/sda, 
but I don't see why this would be a performance issue Theodore should fix.
If you're in a hurry you can just cancel the run after the first 
write-pass anyway and you will be done after ~3 hours.

So when purging data from a damaged drive I assume that the fill-mode of 
ddrescue will provide superior performance if it is invoked to wipe the 
"good" areas of a damaged drive only.
However personally I prefer to wipe the entire drive with multiple 
patterns if sensitive data is stored on it. For me this seems easier to 
do with badblocks in most cases.


Kind Regards
Felix

On 30.04.2013 05:59, Garegin Asatryan wrote:
> badblocks has a problem with speed. this has been widely reported, but 
> its developer- Theodore Tso, has constantly ignored pleas to fix it. 
> as a result it literally takes two days to scan a 1tb hdd vs. 
> ddrescue's couple of hours.