On Tue, Mar 05, 2002 at 08:19:39AM -0800, IT3 Stuart Blake Tener, USNR-R wrote:
> I must admit, I do not clearly understand why something such as
> the CryptoAPI must live in kernel space? Can a user space implementation
> of it not exist?
That is the current problem. Many user space applications have their
own implimentations. There are different libraries you can compile
with in user space.
You could possibly do an efficient user-space file crypto file system in
a microkernel architecture. But that would not help you.
File systems are just not portable. They have to implimented under
each OS on which they are used because efficiency is required, different
OS's have different abstractions, etc.
It *might* be possible to write a portable file system emulator in
user space. But it would be slow, slow, slow as it would be essentially
a loopback file operated on in user space. I don't even want to imagine
how many context swaps would be involved in looking at a file.
> Presuming for the sake of understanding that it does need to
> live in kernel space, if it were written as a module (in much the same
> way loop-aes is), could it not be made to support a plethora more
> kernels?
It is part of the kernel because it isn't only a file system. It is an
encryption/decryption Application Programmer Interface that can be
utilized by other modules or by OS calls from application programs.
It just happens that it *also* supports a *linux* loopback file system.
> Can the CryptoAPI be made to work in user space?
I stated this above, but I'll also add that I prefer the security
of one single well audited common toolset compiled into the kernel.
I don't even use modular kernels. I've been told by some
"former blackhat" friends that fiddling the kernel via a
module was one of their favorite backdoor tricks. He could do it without
a reboot that someone would probably notice, and it then hid everything
he did from normal user space tools.
> The problem (for me) is the fact that it is a kernel patch, and
> I cannot just plug it into any kernel I need to use it with, this is not
> a small issue, it is quite critical in my decision.
Herbert is dealing with this. Personally I've never
found it a problem (except for the period in which the
core 2.4 loopback in linux was broken). Typically the patches
work for many versions. I've never been left uncovered except
for that one broken at birth 2.4 kernel (2.4.14 or something?)
--
------------------------------------------------------
Nuke bin Laden: Dale Amon, CEO/MD
improve the global Islandone Society
gene pool. www.islandone.org
------------------------------------------------------
-
Linux-crypto: cryptography in and on the Linux system
Archive: http://mail.nl.linux.org/linux-crypto/
