RE: z-Re: Fixes to loop.c et al. for DES,IDEA,stego now done

Ray Van Tassle-CRV004 (Ray_Van_Tassle-CRV004@email.mot.com)
Fri, 9 Aug 1996 8:05:38 -0500

>
> -----BEGIN PGP SIGNED MESSAGE-----
>
> In article <Pine.LNX.3.91.890918054255.204A-100000@HellSpawn>,
> Damien Lucifer <root@hellspawn.Berkeley.EDU> wrote:
> >
> >
> >On Tue, 6 Aug 1996, Ian Goldberg wrote:
> >> This directory contains patches to the Linux kernel to enable
encryption and
> >> steganography of filesystems. Encryption allows you to have a scrambled
> >> partition or file that, with the proper pass phrase, you can mount,
just
> >So the question on my mind, is can the loop device(s) be
> >multi-threaded? I decided to be clever one evening and moved all my home
> >directories to cfs directories.. unfortunatly when two instances of the
> >same user try to hit their CFS home directory at once, the whole machine
> >goes flubflub and and needs to be rebooted. Single threading didnt seem
> >like such a terrible thing when I installed cfs, but lately its become
> >rather restrictive and ugly to deal with. Please tell me theres a better
> >way. :)
>
> Well, unlike CFS, loop.c is part of the Linux kernel, which is
single-threaded,
> so I guess loop.c is also single-threaded. However, I took care that
deadlocks
> be avoided (mounting a loop device as another loop device (for example,
> hiding an encrypted filesystem as stego) caused me to think a bit, but I'm
> fairly confident that it works now).
>
> - Ian

It's both better and worse than you thought. There _is_ a deadlock with the
loop device (and maybe MD in raid mode). But it's in the kernel, not in the
driver.
I finally tracked it down and fixed it, but am having trouble getting it to
Linus (operator error!)

I also have a performance improvement to loop.c, and I'll sent it along as
soon as I get the deadlock fix into the kernel.

Regards,
Ray