Messages in this thread | | | From | "Khimenko Victor" <> | Date | Wed, 15 Dec 1999 07:13:03 +0300 (MSK) | Subject | Re: crypted swapspace? |
| |
In <Pine.LNX.3.96.991214144130.13635B-100000@idoru.mine.nu> Lucca (sdlucca@mindspring.com) wrote: >>But "swap space" is different: It is completely acceptable if the >>kernel grabs a random key from the entropy pool on the first swap.
> That's a good idea. Linux already has a beautiful API for encrypted block > devices (to be used for filesystems ultimately) that is available at > http://www.kerneli.org/ .
>>From what I understand, those can't be used to make a secure swap > partition because they occaisionally need to allocate more memory.
What for ???
> A swap-out would cause a very nasty catch-22.
> I know very little about "strong" cryptographic algorithms, so I'm not > sure if any exist that would have reasonably stable memory usage.
Ok. Most "strong" symmetric (we need symmetric crypto for swap!) cryptgraphic algorithms have the same memory usage: some structure (not very big) is allocated on initialization but NO allocations are needed in crypto process itself. IMO much bigger problem is speed. Blowfish will give you 4MiB/sec with 100% processor usage on Pentium200MMX and blowfish is pretty fast (but of course not fastest available) algorithm. If you need real security it can be tolerable though...
> Also... grabbing a key from the entropy available at "first swap" isn't > necessarily a good idea. What if there isn't enough entropy?
Block and wait.
> Go unsecure? Worse, block until entropy replenishes... (as is normal for > /dev/random) yuck.
Why it's so problematic ?
> Many distributions have a random seed they store when they shut down, and > restore when they start up. This keeps a sufficient level of entropy, but > then you are depending on userland to do something before swapping is > necessary. This isn't necessarily bad, after all userland needs to > /sbin/swapon before anything happens anyway. The Mounting of the swap > seems to be logical place to pull a key out of entropy.
- 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/
| |