Re: Using compression before encryption in device-mapper

[Guillaume Lacôte]

Thank you for your answers.
Le Mardi 13 Avril 2004 19:45, Jörn Engel a écrit :

>> 0) Has this problem already been adressed, and if yes, where ?
> Yes, on the filesystems level.  Jffs2 is usable, although not
> well-suited for disks and similar, ext2compr appears to be unusable.
> On the device level, I haven't heard of anything yet.
Thank you, I didn't know about Jffs2; however I believe it is not an 
implemendation at the device level as I would like.

> > 1) Using dm:
> I'd go for a dm implementation.
>
> > 2) Block I/O boundaries:
> > 3) Compressed sectors have varying sizes ...
> > 4) Block allocation on writes: 
>
> If you really want to deal with this, you end up with a device that
> can grow and shrink depending on the data.  Unless you have a strange
> fetish for pain, you shouldn't even think about it.
Since space efficiency is _not_ my aim I plan to forcibly allocate 3 physical 
blocks for every 2 "compressed" blocks (as it should (?) always fit with a 
dynamic Huffman encoding).

>
> > 5) As a workaround to 2,3,4 I plan to systematically allocate 2 sectors
> > per real sector (space efficiency is _not_ my aim, growing entropy per
> > bit is) and to use a trivial dynamic huffman compression algorithm. Is
> > this solution (which means having half less space than physically
> > available) acceptable ?
>
> Makes sense.  One of the zlib developers actually calculated the
> maximum expansion when zlib-compressing data, so you could even get
> away with more than 50% net size, but that makes the code more
> complicated.  Your call.
Oops ! I thought it was possible to guarantee with the Huffman encoding (which 
is more basic than Lempev-Zif) that the compressed data use no more than 1 
bit for every byte (i.e. 12,5% more space).

>
> Performance should not be a big issue, as encryption is a performance
> killer anyway.
I am not sure that this is good news ;)
>
> Whether it is acceptable depends on the user.  Make it optional and
> let the user decide.
>
> > 6) Shall this whole idea of compression be ruled out of dm and only be
> > implemented at the file-system level (e.g. as a plugin for ReiserFS4) ?
>
> Again, depends on the user.  But from experience, there are plenty of
> users who want something like this.
Unfortunately I failed to find substantial code/documentation on encryption 
plugin for Reiser4, for example. Do you know about some ?

>
> Jörn
Thank you, Guillaume.

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