From mboxrd@z Thu Jan  1 00:00:00 1970
From: Derek Fawcus <dfawcus@cisco.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] pathetic
Message-ID: <20040225154145.G7383@edinburgh.cisco.com>
References: <Pine.LNX.4.44.0402250822260.9147-100000@maxroach.lanl.gov>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <Pine.LNX.4.44.0402250822260.9147-100000@maxroach.lanl.gov>; from rminnich@lanl.gov on Wed, Feb 25, 2004 at 08:26:01AM -0700
Date: Wed, 25 Feb 2004 15:41:46 +0000
Topicbox-Message-UUID: f541e130-eacc-11e9-9e20-41e7f4b1d025

On Wed, Feb 25, 2004 at 08:26:01AM -0700, ron minnich wrote:
> Actually I'm puzzled anyway as the segment descriptors on x86 have code
> and data bits. I'd be willing to be the real issue is that XP uses the
> moral equivalent of self-modifying-code, and that now that XP is being
> cleaned up they can actually use those bits. But I'm guessing.

As I understand it,  this NX bit is at the page level,  and so can alter
a individual 4K page within the linear memory space.  If they used the
segment facility to do this,  the memory space would be non linear and
on the x86 this'd take a massive performance hit (48 bit addresses).

Basically the 386 lost the ability to have non executable address space
when using the paging h/w and trying to present a simple 32 bit address
space.

DF