From owner-freebsd-hackers  Sun Feb  4 14:18:06 1996
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id OAA10964
          for hackers-outgoing; Sun, 4 Feb 1996 14:18:06 -0800 (PST)
Received: from labinfo.iet.unipi.it (labinfo.iet.unipi.it [131.114.9.5])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id OAA10955
          for <freebsd-hackers@freebsd.org>; Sun, 4 Feb 1996 14:18:00 -0800 (PST)
Received: from localhost (luigi@localhost) by labinfo.iet.unipi.it (8.6.5/8.6.5) id XAA19281; Sun, 4 Feb 1996 23:18:18 +0100
From: Luigi Rizzo <luigi@labinfo.iet.unipi.it>
Message-Id: <199602042218.XAA19281@labinfo.iet.unipi.it>
Subject: Re: FAT filesystem performance
To: j@uriah.heep.sax.de (J Wunsch)
Date: Sun, 4 Feb 1996 23:18:18 +0100 (MET)
Cc: freebsd-hackers@freebsd.org
In-Reply-To: <199602041822.TAA01613@uriah.heep.sax.de> from "J Wunsch" at Feb 4, 96 07:21:54 pm
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Sender: owner-hackers@freebsd.org
Precedence: bulk

> As Luigi Rizzo wrote:
> 
> > I am wondering: how hard would it be to add mmap() to, say, wd.c ?
> > Would it have other useful applications ?
> 
> This would only shift the problem from the msdosfs code into the
> various device drivers.

Well, let me first say that I don't know exactly how mmapping files
works internally. But what I expect is to have the kernel do its
best to keep blocks in core and write them back (if necessary) when
the file is unmapped.

Now assume for a moment the existence of a Very Simple File System
(VSFS), which always contains a single file spanning the whole
partition. mmapping the unique file of this VSFS should do exactly
what I want, and the VSFS is device independant (it's a file
system!).

This VSFS is logically above the device driver, but it might just
be a standard piece of the system which all device drivers invoke
to implement mmap().

Does the above sound reasonable ?

I have no idea on how complex would it be to implement this, but
the feeling is that it is more of a design problem rather than a
coding problem (for those who know how to write such code, of course
:))

	Luigi
====================================================================
Luigi Rizzo                     Dip. di Ingegneria dell'Informazione
email: luigi@iet.unipi.it       Universita' di Pisa
tel: +39-50-568533              via Diotisalvi 2, 56126 PISA (Italy)
fax: +39-50-568522              http://www.iet.unipi.it/~luigi/
====================================================================