Error correction with aes-looback / cryptoloop?

[Mike Oliver]

Quoting Anthony de Boer <adb-tlug-AbAJl/g/NLXk1uMJSBkQmQ at public.gmane.org>:

> Mike Oliver wrote:
>> It seems to work fine, but I'm not so happy about the
>> thought that a one-bit HDD error could make me lose the
>> entire partition.
>
> Are you really sure it works that way?  Disks are random-access beasts,
> and certainly after booting you can go to any part of the disk without
> having to sequentially go through everything ahead of it.


Good point.  Update: I made a small test AES128 filesystem-in-file using
cryptoloop, and went in with a hex editor and changed one byte in the
image file.  It garbaged a few bytes in one file of the filesystem, not
even a whole file.  So it seems to work in ECB mode or something like
it, for good reason as you point out.  I guess I had been assuming CFB
mode, not because it really made sense, but because it was the worst-case
scenario for purposes of my concern here.

> There are critical bits like the superblock and the directory at the root
> of the partition, but alternate superblocks and fsck can affect at least
> partial recovery from many errors.

Yeah, that's a point.  How much space do those occupy?  And do they change
as I add/remove files?  It would be nice to be able to back up a small piece
to protect against errors there.

> And you *ARE* going to keep backups, right?

Well, sure, from time to time.  But I don't back up every day or anything;
too much of a PITA.

--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists