Hi,

Thanks everybody for their comments!

On Thu, 11 Jan 2007, Jachym Holecek wrote:
> rnd is initialized before device autoconfiguration, opencrypto is
> initialized when the first user calls crypto_get_driverid(). So, no
> problem here I think.

I tried this before, and it didn't work. But this turns out to be a 
(trivial) bug in crypto.c, causing reinitialization of opencrypto after 
each crypto_get_driverid() call. I have filed a PR (kern/35412) with a 
fix. It would be great if someone could 'ack' the fix, since this also 
affects support for hifn and ubsec (though they will override cryptosoft).

-- Daniel