From owner-freebsd-hackers  Wed Jun  7 08:27:39 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id IAA22837
          for hackers-outgoing; Wed, 7 Jun 1995 08:27:39 -0700
Received: from asstdc.scgt.oz.au (root@asstdc.scgt.oz.au [202.14.234.65])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id IAA22798
          for <freebsd-hackers@freefall.cdrom.com>; Wed, 7 Jun 1995 08:27:21 -0700
Received: (from imb@localhost)
	by asstdc.scgt.oz.au (8.6.12/BSD4.4)
	id BAA13001; Thu, 8 Jun 1995 01:26:46 +1000
From: michael butler <imb@scgt.oz.au>
Message-Id: <199506071526.BAA13001@asstdc.scgt.oz.au>
Subject: Re: silo overflows
To: jleppek@suw2k.ess.harris.com (James Leppek)
Date: Thu, 8 Jun 1995 01:26:45 +1000 (EST)
Cc: freebsd-hackers@freefall.cdrom.com
In-Reply-To: <9506071209.AA20749@borg.ess.harris.com> from "James Leppek" at Jun 7, 95 08:09:08 am
X-Mailer: ELM [version 2.4 PL24beta]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 897       
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

James Leppek writes:

> Thanks for the info.
> The only bus hogging DMA I could think of is the 1542CF but it is
> at the default settings which I thought were safe.

In the interest of disk performance, FreeBSD uses bus-on/bus-off times which
are quite aggressive (11/4uS ?) as compared to, say, Interactive at about 9
and 6uS. 

As I understand it, heavy disk work, therefore, only yields 4uS of CPU time
(to service the serial port(s)) before grabbing another 11uS for disk
transfers .. probably not including bus-request and acknowledge latencies.

> Is there a "cleaner" way to adjust the fifo trigger levels than making
> another constant like FIFO_TRIGGER_10? perhaps a sysctl or stty option?

You only have four options .. 1, 4, 8 or 14 .. Bruce's suggestion will
afford that flexibility in a later version without necessitating
kernel recompilation to "tune" a workable result,

	michael