From owner-freebsd-current  Tue Dec 12 01:42:59 1995
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id BAA25293
          for current-outgoing; Tue, 12 Dec 1995 01:42:59 -0800 (PST)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id BAA25267
          for <current@freebsd.org>; Tue, 12 Dec 1995 01:42:17 -0800 (PST)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id UAA25385; Tue, 12 Dec 1995 20:40:17 +1100
Date: Tue, 12 Dec 1995 20:40:17 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199512120940.UAA25385@godzilla.zeta.org.au>
To: current@freebsd.org, terry@lambert.org
Subject: Re: What is this thing in locore.s?
Sender: owner-current@freebsd.org
Precedence: bulk

>The following code live in i386/i386/locore.s:
>=============================================================================
>/* now initialize the page dir, upages, and p0stack PT */

>	movl	$(1+UPAGES+1),%ecx	/* number of PTEs */
>	movl	%esi,%eax		/* phys address of PTD */
>	andl	$PG_FRAME,%eax		/* convert to PFN, should be a NOP */
>	orl	$PG_V|PG_KW,%eax	/* valid, kernel read/write */
>	movl	%esi,%ebx		/* calculate pte offset to ptd */
>	shrl	$PGSHIFT-2,%ebx
>	addl	%esi,%ebx		/* address of page directory */
>	addl	$((1+UPAGES+1)*NBPG),%ebx /* offset to kernel page tables */
>	fillkpt
>=============================================================================

>What is the purpose of:
>=============================================================================
>	movl	%esi,%ebx		/* calculate pte offset to ptd */
>	shrl	$PGSHIFT-2,%ebx
>	addl	%esi,%ebx		/* address of page directory */
>	addl	$((1+UPAGES+1)*NBPG),%ebx /* offset to kernel page tables */
>=============================================================================

This initializes %ebx :-).

>The kernel boots and runs fine without it -- it doesn't seem to do
>anything really useful?!?

There are some ifdefs before it.  Apparently %ebx is left with the
correct value for the case that you use.  I think it is left with the
correct value for all cases, but this is only true now that the kernel
is loaded above 1MB.  Previously, in the BDE_DEBUGGER case, %ebx was
left pointing to the pte for physical address 0xa0000, while the page
dir was either somewhat lower (at KERNend - KERNBASE) or perhaps at 1MB.
This depends on the page dir going at (KERNend - KERNBASE).

Bruce