Most disk encryption guides recommend to overwrite the disk with random data before encryption, this is discussed here.
However the time this takes with the "standard" source of random data can be rather slow:
# dd if=/dev/urandom of=/dev/sdb1 bs=10M 29+1 records in 29+0 records out 304087040 bytes (304 MB) copied, 15.5476 s, 19.6 MB/s
On an encrypted device (specifically, with LUKS), are there any drawbacks when one just fills the encrypted partition with zeros?
# dd if=/dev/zero of=/dev/mapper/data bs=50M 1099+0 records in 1099+0 records out 57619251200 bytes (58 GB) copied, 306.091 s, 188 MB/s
The only problem I could see would be that if known-plaintext attacks surface, this method might give an attacker plenty of sample data to play with.