On Wed, 30 Jul 1997, Bill Studenmund wrote:

> ?? The in-tree driver doesn't support inbound flow control yet. Just
> outbound (us hushing when the other side asks).

Wait wait wait.  I'm going to be confused soon.  Reading EIA specs is
_not_ clearing up the matter for me, either, though it's plain that
whatever interpretation I'm reading, there is no inbound flow control
going on.  Thanks for pointing out that much.

I used to know this stuff.  Really. %-\

> I suspect that the act of reporting an overrun can delay things enough
> to generate another. :-)

Certainly probable.

Now as for what's holding off zs interrupts, I have no clue.  I have been
contemplating lowering splsched() to IPL3, though, which should help
matters a little.

--scott