> The sun3 serial port driver uses a two-level interrupt scheme
> (and PDMA) which makes "silo overflow" events rare.  There were
> problems with H/W flow control until somewhat recently, but
> other than that I've not not heard many complaints.
> 
> Note that the sparc does not use the same driver, and the
> mac68k does not have H/W interrupt assignments that allow
> reliable handling of serial port PDMA (or so I've heard).

I think mac68k's fine for the PDMA. I think the problem's else- where.
I'm not sure where, though not having input flow control is probably
part of the problem. The new driver was able to get the other side to
hush before things got bad. I think most of the fifo overruns happen
soon after the first one. Like duh, but I think the process of
recording a message causes enough delays to make another one more
likely.

Though we seem to have broken something just recently. A number of
folks are complaining that they just started getting fifo overruns,
which is weird as the PDMA routine is on the highest interrupt level
of all.

Take care,

Bill