Re: [1/1 take 2] HIFN: preliminary HIFN 795x driver for new async cryptoapi.

[Evgeniy Polyakov <johnpol@2ka.mipt.ru>]

On Tue, Jun 05, 2007 at 05:07:31PM +0400, Evgeniy Polyakov (johnpol@2ka.mipt.ru) wrote:
> This is tricky, hifn requires to have a key per crypto setup, so this
> one must be in a crypto context, but context is not used right now, so
> yes, it supports only one user. I will update it to use context, since
> it is smaller that hifn_device.

Hmm, I was wrong, it is not possible (from first view).

Here is a sequence of events to setup a crypto session:
1. user calls setkey, which results in tfm. It is possible to get a
context from tfm. Context can be initialized via cra_init() callback in
registered algorithm. But there is no way to provide a private pointer
into that callback. So, there is no way to provide a pointer to
hifn_device and thus tfm context can not find a hardware device to setup
this session.

Herbert, could you please elaborate how to provide a private pointer
into cra_init without searching via static lists of all devices?

-- 
	Evgeniy Polyakov