From owner-freebsd-isp  Thu May 29 11:17:48 1997
Return-Path: <owner-isp>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id LAA29509
          for isp-outgoing; Thu, 29 May 1997 11:17:48 -0700 (PDT)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id LAA29502;
          Thu, 29 May 1997 11:17:44 -0700 (PDT)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.5/8.6.9) id EAA24203; Fri, 30 May 1997 04:14:28 +1000
Date: Fri, 30 May 1997 04:14:28 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199705291814.EAA24203@godzilla.zeta.org.au>
To: hackers@FreeBSD.ORG, paulo@nlink.com.br
Subject: Re: Cyclades PCI Strange Behavior
Cc: isp@FreeBSD.ORG
Sender: owner-isp@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>a Pentium 133 with Intel i430vx Motherboard
>32MB RAM.
>1.2GB IDE HardDrive
>a Trident PCI Video Card
>A cyclades PCI Cy32 card mapped to high memory
>32 dial-in modems

>We are getting a lot of silo overflows that get more frequently when we 
>start using the computer ( very frequent when we compile the kernel ).
>If we start an upload to the computer, we have very much silo overflows 
>than in a download.

Silo overflows are probably normal for the PCI version of the driver.
It uses a normal low-priority interrupt handler which can be delayed
for several msec by other interrupt handlers.  At 115200 bps, there is
about 1 silo overflow for ever msec of delay.  Delays of 3-5 msec are
normal for updating the keyboard LEDs.  Delays of 0.1-1 msec are normal
for IDE drives and PIO ethernet cards.  Delays accumulate.

>After about 1 day with the system running it JUST RESETS without any 
>message in the screen if we SWITCH ONE OF THE MODEMS OFF AND ON AGAIN ( 
>Just after we turn if on ).

This is probably unrelated.  There is a software bug that causes the
timeout table to fill up under certain overloads, but this normally
causes panics with a message.

Bruce