[LAU] OT: Bash help to check new USB keys.

[David Christensen]

On 07/08/2015 09:54 AM, Kaza Kore wrote:
> I have recently bought some large USB keys to use as backup storage
> to replace my old moving disk external hard drive. Due to the worry
> of the reports of fakes reporting to be large drives, even when
> plugged into the system, I wanted to test these before using them for
> real. Seems best way to do this is fill the drive up and check data
> for integrity. So first step will be to create a 8GB file which I
> will copy multiple times.

Have you heard of the ISAAC cryptographic pseudo-random number generator 
(CPRNG)?

     http://burtleburtle.net/bob/rand/isaacafa.html

Given seed value(s), it will generate a repeatable sequence of "random" 
integer values.  Cycles should be more than long enough to fill your 
device before the sequence repeats (you could save the first block of 
values for detection and generate a warning if this occurs).


You could generate a sequence of values, write them to the device one 
block at a time (4096 bytes, far more efficient and easier on your flash 
memory than writing one integer at a time), wait for the device to fill, 
reset ISAAC, read the blocks back, compare to ISAAC values, and keep a 
running count of good and bad blocks, etc..  This will confirm if your 
device really has memory behind the addressable blocks, as every block 
will contain a unique pattern that the USB device firmware won't be able 
to fake.


When done, you'll probably want to fill the device with zeros (e.g. 
erase it).


Rather than userland tools and a Bourne shell script, I'd suggest doing 
it in C or some other systems programming/ scripting language.


On 07/08/2015 11:14 AM, Kaza Kore wrote:
> It's not a full testing of a flash device I'm worried about,
> especially as they are new, I just want to know they really are the
> size they report as being...

You could use the above idea to check every 2**N blocks plus the last 
block, or some other sampling of the addressable space.  This could be 
much faster than checking every single block (especially with large 
and/or USB 2.0 and slower devices).


David