From owner-freebsd-current  Wed Oct  9 12:49:20 1996
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id MAA26250
          for current-outgoing; Wed, 9 Oct 1996 12:49:20 -0700 (PDT)
Received: from rocky.mt.sri.com (rocky.mt.sri.com [206.127.76.100])
          by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id MAA26244
          for <current@freebsd.org>; Wed, 9 Oct 1996 12:49:13 -0700 (PDT)
Received: (from nate@localhost) by rocky.mt.sri.com (8.7.5/8.7.3) id NAA11505; Wed, 9 Oct 1996 13:49:10 -0600 (MDT)
Date: Wed, 9 Oct 1996 13:49:10 -0600 (MDT)
Message-Id: <199610091949.NAA11505@rocky.mt.sri.com>
From: Nate Williams <nate@mt.sri.com>
To: current@freebsd.org
Subject: SIO problems
Sender: owner-current@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

OK, I *really* think something is funky with the serial drivers in
2.1.5.  I'm getting really good results most of the time, but the # of
overflows I'm seeing concerns me.

HW:
486/66 - 16MB memory
Aha1542B - running at 'slow' transfer rates (3.3MB/sec or 5MB/sec,
           whichever is the slowest, I can't remember).
multi-I/O card
 - Slot 0 has a 16450 and is connected to my mouse
 - Slot 1 has a TRUE-BLUE 16550AFN and is connected to a USR Sporster
   28.8K running SLIP at 115K

/etc/rc.serial is running modem() on slot 1, but is modified to 115200
baud since that's what I'm running the modem at.  I'm running slattach
at 115200 baud.

The machine seems to have problems when uses as a Remote-CVS server,
which hits both the network connection and the disks hard, but I've also
gotten overflows when I was away on business in the middle of the day,
when *NO* significant traffic was on the line.

Background:
  I've been running this box for years at high-speeds with no problems,
but as soon as I upgraded to 2.1.5 I started seeing problems.  Bruce
fixed a bug that supposedly allowed overflows to go un-noticed in the
past, but are now recognized, so that's what I assumed.

However, I started seeing lots of serial overflows.  I even ran a
program Bruce provided to check the latency of my system under heavy
load, but I was well under the 'problem' mark.

Snippets from a boot:
sio1 at 0x2f8-0x2ff irq 3 on isa
sio1: type 16550A

I uncommented the 'change FIFO level' code since it happens rarely to
see if helps, with the following results:

sio1: 1 more silo overflow (total 1)
sio1: reduced fifo trigger level to 8
sio1: 1 more silo overflow (total 2)
sio1: reduced fifo trigger level to 4
sio1: 1 more silo overflow (total 3)
sio1: reduced fifo trigger level to 1
sio1: 1 more silo overflow (total 4)
sio1: reduced fifo trigger level to 4

Obviously the dynamic code doesn't work quite as good as it should since
it changed the level from 1 -> 4, but in any case I'm getting overflows
and I really shouldn't be.  The box has been up 9 days with this.  I
have a 486/66 box on the other end running 3 full-time connections that
gets beat up even more than my box and it has no problem, although it's
an IDE system that never touches it's disk.

I'm really confused as to why I'm seeing these errors, so if anyone can
shed light on it I'd appreciate it.


Nate