These are the files for Sun serial net:

tty_sl.c        - the driver, installed as /sys/sys/tty_sl.c (see below)
tty_conf.c.diff - context diff of changes to /sys/sys/tty_conf.c, 10 lines added
ioctl.h.diff    - context diff of changes to /sys/h/ioctl.h, two lines added
files.diff      - context diff of changes to /sys/conf/files, one line added
slattach.c      - program for /etc (or where ever)
sldetach.c      - program for /etc (Ibid.)
sl_stats        - adb script for looking at sl_stats struct, if you want to

Installing the driver as /sys/sys/tty_sl.c follows ECF's practice, the
alternative is /sys/sunif/if_sl.c .  Personally I don't care much.

The driver has run a heavily-used 19200 baud connection for a number
of weeks through an MTI 16 port multiplexor on a Sun 3/180 and we have
yet to see an error of any kind.

Bugs:

The mti multiplexor board, despite claims of a huge input silo, is run
interrupt per input character by the Sun driver.  This means that incoming
network traffic beats on the Sun pretty hard.  An incoming rcp from a
Microvax to /dev/null over a 19200 baud line causes the Sun to take 1650
interrupts per second and consumes 20% of the CPU.  Caveat emptor (sp?).

The driver is an offspring of Thomson's VAX serial net driver.  While
this means that it is both cleaner than the original and easier on your
CPU, it also means that it buffers large numbers of outbound characters
in clists attached to the terminal output queue where the network can't
see them.  What this implies is that if you are rlogin'd over the line
and someone else is doing an rcp, your terminal response will be even
more rotten than the old original vax driver.  This applies to Thomson's
VAX version of the driver as well so you shouldn't blame it on this one.

The mti board complains of a silo overflow when you do the slattach,
sometimes.  I know not why.  This isn't important but it annoys me.  If
someone fixes this please tell me how.
