From owner-freebsd-hackers  Fri Feb  7 11:02:13 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id LAA12711
          for hackers-outgoing; Fri, 7 Feb 1997 11:02:13 -0800 (PST)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id LAA12693
          for <hackers@freebsd.org>; Fri, 7 Feb 1997 11:02:10 -0800 (PST)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.3/8.6.9) id GAA13110; Sat, 8 Feb 1997 06:00:19 +1100
Date: Sat, 8 Feb 1997 06:00:19 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199702071900.GAA13110@godzilla.zeta.org.au>
To: bmcgover@cisco.com, dg@root.com, support@cyclades.com
Subject: Re: Flow control on Cyclades boards???
Cc: hackers@freebsd.org, waz@cyclades.com
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

>Ok. I now have the boards up and running between machines, but I'm having a 
>problem with silo overflows almost constantly on all the currently running 
>ports. I'm using pppd (options listed below) to set up a number of ppp
>sessions between two pentium pro 200s. 
>
>I originally tried to run 32 of them at 115200. It failed miserably. I then
>dropped back to 16 at 115200. Then to 16 at 57600. Still got silo
>errors constantly.

The PCI Cyclades driver doesn't support "fast" interrupt handlers yet
(because the PCI interrupt dispatcher doesn't), so at 115200 bps,
each fully active port will get N fifo full's worth of silo overflows
(12 * N characters) whenever the kernel disables tty interrupts for
N milliseconds.  The kernel sometimes disables tty interrupts for
several milliseconds.  This can't be fixed using flow control or any
other configuration option.

Bruce