From owner-freebsd-current  Fri Sep 13 12:27:29 1996
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id MAA10819
          for current-outgoing; Fri, 13 Sep 1996 12:27:29 -0700 (PDT)
Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.211])
          by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id MAA10807;
          Fri, 13 Sep 1996 12:27:26 -0700 (PDT)
Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id MAA09428; Fri, 13 Sep 1996 12:25:30 -0700
From: Terry Lambert <terry@lambert.org>
Message-Id: <199609131925.MAA09428@phaeton.artisoft.com>
Subject: Re: VM problem (OK!!)
To: dyson@FreeBSD.org
Date: Fri, 13 Sep 1996 12:25:29 -0700 (MST)
Cc: jdli@FreeBSD.csie.NCTU.edu.tw, freebsd-current@FreeBSD.org
In-Reply-To: <199609130307.WAA00231@dyson.iquest.net> from "John S. Dyson" at Sep 12, 96 10:07:55 pm
X-Mailer: ELM [version 2.4 PL24]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-current@FreeBSD.org
X-Loop: FreeBSD.org
Precedence: bulk

> > >  Addition of page coloring support.  Various levels of coloring are
> > >  afforded.  The default level works with minimal overhead, but one
> > >  can also enable full, efficient use of a 512K cache.  (Parameters
> > >  can be generated to support arbitrary cache sizes also.)
> > 
> > 	BTW, I don't understand this, does it mean that I can tune my kernel
> > 	to use cache better ? (I got 512k pb-cache, 32MB DRAM, ASUS T2P4 M/B)
>
> Yes!!!  I haven't documented the options yet, their names are probably
> not final.

Hold up a second.

When you said you had page coloring working, I immediately thought in
Microsoft terms of using a segment type identifier on page creation
for gross management of code.

Specifically, I thought I could do this with an ELF kernel:


Color	Meaning

0x00	Immovable data
0x01	Immovable code
0x02	Discardable data
0x03	Discardable code
0x04	Pageable data
0x05	Pageable code
0x06	Driver locked data
0x07	Driver locked code
0x08	Driver pageable data
0x09	Driver pageable code

[ ... ]

Then I would make the console driver by putting the interrupt handler
code in 0x07, the data referenced by the interrupt handler code (ie:
volatile data) in 0x06, one time initialization code/data in 0x03 and
0x02, and the rest of the driver in 0x09 and 0x08.

Then I could replace the console driver with a card-specific driver
once I had booted and throw away all pages that contained the default
console driver.

Similarly, I could replace a VM86()-based BIOS-using disk driver with
a controller-specific driver, letting me boot on all 386 or better
hardware DOS can boot on, run at diminished performance if there was no
protected mode driver, and load a protected mode driver for better
performance if it were available (and throw away the BIOS-using driver
once I had done that).


It seems that you are using page coloring in terms of L2 cache here;
am I missing something?


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.