Christoph Anton Mitterer wrote:
I did about 5 complete tests today and dozens of mkfs.ext3's but I wasn't able to reproduce any of the two errors... very very strange. (used the same sequence of commands, with and without using the USB-stick)... I'll do some other tests tomorrow because these problems were real and I cannot believe, that they're simply gone... And IMHO hardware problems are still very unlikely, or am I wrong?
LUKs operates at the block level - it seems unlikely to have a fault condition that affects single bits in bytes.
On the other hand hardware errors in memory devices can do just that. Other's on the list have given their experiences. I once had verification errors when burning CDs. Bit errors in the un-recorded part of the CD (normally set to 0xFF) were the symptom but the fault was eventually tracked to sticky bits in the L2 cache of the processor. Curiously other faults occurred so infrequently and inconsistently that they pretty much went unnoticed or unexplained.
In your case it might be that because the encryption produces more varieties of bit pattern than usually found in unencrypted data that the opportunity for a particular memory error to be seen is greater.
Anyway,.. is there anybody who made deeper tests of dmcrypt? I mean real massive tests perhaps with different filesystems and so on?
IMHO I don't think you should be blaming dmcrypt in this case. Dick --------------------------------------------------------------------- dm-crypt mailing list - http://www.saout.de/misc/dm-crypt/ To unsubscribe, e-mail: dm-crypt-unsubscribe@xxxxxxxx For additional commands, e-mail: dm-crypt-help@xxxxxxxx
