From mboxrd@z Thu Jan 1 00:00:00 1970 From: Derek Fawcus To: 9fans@cse.psu.edu Subject: Re: [9fans] pathetic Message-ID: <20040226011539.N7383@edinburgh.cisco.com> References: <20040225154145.G7383@edinburgh.cisco.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: ; from rminnich@lanl.gov on Wed, Feb 25, 2004 at 09:02:45AM -0700 Date: Thu, 26 Feb 2004 01:15:39 +0000 Topicbox-Message-UUID: f72edfe8-eacc-11e9-9e20-41e7f4b1d025 On Wed, Feb 25, 2004 at 09:02:45AM -0700, ron minnich wrote: > On Wed, 25 Feb 2004, Derek Fawcus wrote: > > well ... on e.g. linux, the segment descriptors are used. Paging on x86 is Well the descriptors have to be there, but they're ususally set for 1:1 mapping (phys == linear). It used to be that linux used the segments for proc protection, then just for the 3G limit, now I believe they are set 1:1 and everything done with pages. > in addition to segments. segment are earlier in address translation path > than pages, and they are still there when paging is turned on. It's quite > weird. err - later. Virt => Linear (via paging h/w). Then Linear => Phys (via segment h/w). > I don't see an NX bit in my pentium manual. That's what AMD are supposed to have added. That's what the whole hoohaw is about, they've added a bit somewhere (I'd guess a spare bit from the page descriptor) such that a page can be non execute. DF