shred vs writing zeros to wipe a drive

[Matt Graham]

On 2013-08-19 10:39, Carruth, Rusty wrote:
> Don't' know if I need to see the code, but I'm certainly curious to 
> know
> how you made it slightly more efficient :-)

If you do "dd if=/dev/zero of=/dev/sdb bs=32k" , then dd is constantly 
reading from /dev/zero .  Sure, reading from /dev/zero is fast, but

zeroes=calloc(1,32768*sizeof(char));
/* later */
fwrite(zeroes,32768,1,WRITE_PTR);

doesn't do *any* I/O to /dev/zero .

When doing the "write random data" pass, randbuffer contains 32K of 
data read from /dev/urandom .  After the data is written, randbuffer is 
only rewritten with new random data if rand() % 2 == 0.  This means that 
there's 50% less I/O to /dev/urandom at the price of having the random 
data repeated for more than 1 block in a mostly non-deterministic way.  
Which means the program spends more time writing and less time reading.

> I figure /dev/urandom is at least one order of magnitude better
> than /dev/zero (in some base or other) :-) But, yes, if you REALLY
> want good random numbers use /dev/random and be prepared for a wait.

If you need more than about 256 bytes of stuff out of /dev/random, 
you'll be waiting for a looooong time.  At least IME.  If you have 
serious needs for lots of high-quality non-deterministic entropy, I 
guess you build an RNG made out of the beta emitters in old smoke 
detectors.  (Fun for the whole family! :-) )

-- 
Crow202 Blog: http://crow202.org/wordpress
There is no Darkness in Eternity
But only Light too dim for us to see.