Re: crypto and fsck guys.

Erez Zadok (ezk@cs.columbia.edu)
Mon, 18 Oct 1999 12:38:28 -0400 (EDT)

In article <E11dB0T-0008Jk-00@the-village.bc.nu> you write:
> > I have just lost _ABSOLUELY_ everything
> > of importance to myself. The reason? put
> > quite plain and simply, two groups need to
> > get their act together:
>
> Actually to be brutal the reason is because you didnt make a backup.
>
> > I run fsck to check the ext2 fs on my
> > encrypted partition before it gets mounted,
> > I don't expect fsck to provide the added
> > "feature" of destroying that filesystem when
> > it was in fact completely stable, usuable
> > and NOT in need of repair. I'm going
> > to have to see my doctor now as I am the
> > most depressed that I have ever been.
>
> fsck actually doesn't even know about encrypted data. The encrypt/decrypt
> happens below the level either it or the file system code sees. If you
> tried to fsck the crypted data not the uncrypted loopback I imagine
> e2fsck would think your fs was quite odd however.
>
> Alan

My cryptfs is a stackable file system which mounts over any existing
directory. No keys are stored on disk, and the underlying file system
(ext2, nfs, whatever you want) does NOT change at all. Therefore, you use
normal fsck and other utilities, w/o fear of corruption. You do your backup
on the lower-level (ciphertext) directories, which is (1) faster b/c you
don't need to decrypt and (2) safer b/c the backup operator doesn't need to
decrypt the data.

The cost of this stacking is 5-7% over the lower-level (stacked-on) file
system. This cost is well worth it IMHO, esp. given that crypto algorithms
cost much more than that.

You can get software and several papers on the subject from

http://www.cs.columbia.edu/~ezk/research/
http://www.cs.columbia.edu/~ezk/research/software/

-- 
Erez Zadok.
Columbia University Department of Computer Science.

---
EMail: ezk@cs.columbia.edu           Web: http://www.cs.columbia.edu/~ezk


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/