For a full-disk encryption setup, with a physical disk X and the induced logical volume X_crypt, there are several ways to initially fill the disk:
- (A): Fill X with zeros;
- (B): Fill X with random data;
- (C): Fill X_crypt with zeros;
- (D): Fill X_crypt with random data.
All theses methods have advantages and disadvantages:
- (i): (B), (C) and (D) allow you to hide which parts of the disk are used, while (A) does not.
- (ii): (A) and (C) allow you to prove that you have fully decrypted the disk, which (B) and (D) do not.
- (iii): (C) would be slightly easier to attack if a known plaintext attack was discovered for your cypher.
It looks to me like (ii) is necessary if you ever plan on traveling to countries that value security a lot more than they value privacy, which only leaves (A) and (C). Then, since I don't really see any danger to (i), and with (iii) in mind, I would be tempted to choose (A). However, most tutorial seem to recommend either (C) or (D).
Is there something I'm missing?