[dm-crypt] Questions about LUKS / LVM

[dm-crypt] Questions about LUKS / LVM

[Robbie Smith]

Hello everyone

I'm contemplating setting up my new laptop to make use of LUKS + LVM
when I install Arch Linux on it, but I've got a few questions.

How much of a load on the system would LUKS + LVM be? Is it likely to
result in a noticeable drop in performance? Does entering the key(s) at
boot decrypt the whole volume, or just provide a means for the kernel
module to decrypt and encrypt on-the-fly?

And… how does it work? The documentation makes mention of multiple
key-slots; but I'm a little baffled as to how different keys can be used
to decrypt the same volume. It is based on symmetric cryptography, isn't
it?

I'd just like to get as much information as possible before I set this
up, as I'm still not entirely sure if full-disk encryption is worth the
effort for a personal laptop.

Cheers if you can advise

Robbie

Re: [dm-crypt] Questions about LUKS / LVM

[Karl O. Pinc]

On 09/19/2011 10:45:52 AM, Robbie Smith wrote:

> How much of a load on the system would LUKS + LVM be?
> Is it likely to
> result in a noticeable drop in performance?

It all depends, but generally no because cpu is _so_ much faster
than disk these days.

> Does entering the key(s)
> at
> boot decrypt the whole volume, or just provide a means for the kernel
> module to decrypt and encrypt on-the-fly?

The latter.

> 
> And… how does it work? The documentation makes mention of multiple
> key-slots; but I'm a little baffled as to how different keys can be
> used
> to decrypt the same volume. It is based on symmetric cryptography,
> isn't
> it?

Yes, but the master key is encrypted by each key, separately, and 
that's what your multiple passwords decrypt.

See the tks-1 paper (iirc) referenced on the wiki for more info.




Karl <kop@meme.com>
Free Software:  "You don't pay back, you pay forward."
                 -- Robert A. Heinlein

Re: [dm-crypt] Questions about LUKS / LVM

[Quentin Lefebvre]

On 19/09/2011 18:00, Karl O. Pinc wrote :
> On 09/19/2011 10:45:52 AM, Robbie Smith wrote:
> 
>> How much of a load on the system would LUKS + LVM be? Is it
>> likely to result in a noticeable drop in performance?
> 
> It all depends, but generally no because cpu is _so_ much
> faster than disk these days.

Indeed, you won't notice it on a modern computer, especially for every
day use. It may be painful for data transfer from one encrypted disk to
another encrypted disk, but you can expect very good rates anyway. It
depends on the cipher you use, on your machine, ...

>> Does entering the key(s) at boot decrypt the whole volume,
>> or just provide a means for the kernel module
>> to decrypt and encrypt on-the-fly?
> 
> The latter.
> 
>> And… how does it work? The documentation makes mention of multiple
>> key-slots; but I'm a little baffled as to how different keys can be
>> used to decrypt the same volume. It is based on symmetric
>> cryptography isn't it?
> 
> Yes, but the master key is encrypted by each key, separately, and 
> that's what your multiple passwords decrypt.
> 
> See the tks-1 paper (iirc) referenced on the wiki for more info.

This is important as it means your disk encryption is basically as
strong as your "weaker key slot". e.g. a very bad password for one key
slot make all other passwords/keys inefficient.

Thanks, Karl, for the link.

Best,
Quentin

Re: [dm-crypt] Questions about LUKS / LVM

[Yves-Alexis Perez]

[-- Attachment #1: Type: text/plain, Size: 530 bytes --]

On lun., 2011-09-19 at 18:11 +0200, Quentin Lefebvre wrote:
> > It all depends, but generally no because cpu is _so_ much
> > faster than disk these days.
> 
> Indeed, you won't notice it on a modern computer, especially for every
> day use. It may be painful for data transfer from one encrypted disk
> to
> another encrypted disk, but you can expect very good rates anyway. It
> depends on the cipher you use, on your machine, ...
> 
Using a modern intel cpu with aesni helps a lot too.

Regards,
-- 
Yves-Alexis

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]