Multi-block operations with VIA Padlock

[Michal Ludvig]

Herbert Xu wrote:
> On Sun, Apr 17, 2005 at 11:58:16PM +1200, Michal Ludvig wrote:
>>Yes, see http://www.logix.cz/michal/devel/padlock/bench.xp - the 
>>multiblock operation is definitely worth having!
> 
> Excellent.  Good to see that you've already thought this through
> months ago :)

Well that's what the PadLock is about ;-) Unfortunately only a stripped
down version of the patch was accepted to the kernel so far.

>>The patch for 2.6.10 is here: 
>>http://www.logix.cz/michal/devel/padlock/crypto-api-2.6.10.diff
>>Yes I know 2.6.10 is too old but you'll get an idea of what I did. If 
>>you'd like the approach I'm ready to forward port it to 2.6.12(-rc?).
> 
> I didn't find the padlock-specific code there, did I miss it?

http://www.logix.cz/michal/devel/padlock/crypto-dev-padlock-2.6.10.diff
BTW it is a whole driver, not a diff against the single-block version.
More patches are at http://www.logix.cz/michal/devel/padlock (e.g. the
"preference patch" which I still think is usefull).

> As to the API change: Does it make sense to continue to use the
> crypto scatterlist walking code? Would it be better for the padlock
> driver to do its own walking?

I don't know. Would the benefit be worth the possible troubles? I just
cleaned up and extended what was already there but I don't say it's the
best option...

Michal Ludvig
-- 
* Personal homepage: http://www.logix.cz/michal