On Sun, Aug 14, 2016 at 4:02 PM, Dave Taht <dave.taht@gmail.com> wrote:

> On Mon, Aug 15, 2016 at 12:49 AM, Aaron Wood <woody77@gmail.com> wrote:
> > On Sun, Aug 14, 2016 at 3:21 PM, Dave Taht <dave.taht@gmail.com> wrote:
> >>
> >>
> >> Ideas as to why crypto chops off so much throughput in the first place
> >> (from 120mbit to 80), and as to what else to poke into with crypto on,
> >> would be nice.
> >
> >
> > Crypto as in WPA2?  If 1 client is fine, but multiple clients is not, my
> > guess is that changing the AES key in the crypto accelerator is
> > slow/difficult, and so what you're seeing is the lost time as it switches
> > security contexts.
>
> 1 client suffers degradation with more than one flow in flight, being
> fq'd, and crypted with wpa2.
>
> I'll have plots in the morning.
>

So it's not, that, then.  Unless it's doing something dumb with the keys
when it switches flows (since that's now an interrupt back up to the
driver?)

-Aaron