From owner-cvs-all  Fri Jun 22 22:42:48 2001
Delivered-To: cvs-all@freebsd.org
Received: from mailman.zeta.org.au (mailman.zeta.org.au [203.26.10.16])
	by hub.freebsd.org (Postfix) with ESMTP
	id A68DE37B401; Fri, 22 Jun 2001 22:42:38 -0700 (PDT)
	(envelope-from bde@zeta.org.au)
Received: from bde.zeta.org.au (bde.zeta.org.au [203.2.228.102])
	by mailman.zeta.org.au (8.9.3/8.8.7) with ESMTP id PAA07057;
	Sat, 23 Jun 2001 15:42:31 +1000
Date: Sat, 23 Jun 2001 15:40:39 +1000 (EST)
From: Bruce Evans <bde@zeta.org.au>
X-Sender: bde@besplex.bde.org
To: Matthew Jacob <mjacob@feral.com>
Cc: John Baldwin <jhb@FreeBSD.org>, cvs-all@FreeBSD.org,
	cvs-committers@FreeBSD.org, Peter Wemm <peter@FreeBSD.org>
Subject: Re: D'oh!
In-Reply-To: <20010622152459.X21527-100000@wonky.feral.com>
Message-ID: <Pine.BSF.4.21.0106231508410.45551-100000@besplex.bde.org>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-cvs-all@FreeBSD.ORG
Precedence: bulk
List-ID: <cvs-all.FreeBSD.ORG>
List-Archive: <http://docs.freebsd.org/mail/> (Web Archive)
List-Help: <mailto:majordomo?subject=help> (List Instructions)
List-Subscribe: <mailto:majordomo?subject=subscribe%20cvs-all>
List-Unsubscribe: <mailto:majordomo?subject=unsubscribe%20cvs-all>
X-Loop: FreeBSD.ORG

On Fri, 22 Jun 2001, Matthew Jacob wrote:

> Aghh.... internally to the kernel physmem is in units of pages for alpha or
> the oldie moldy 'clicks' for i386. What *was* I thinking?  It's really just
> the sysctl representation, no?

No.  The kernel variable is in pages, but the sysctl scales it to bytes
using broken scaling code:

On i386's, the sshould-be-deprecated MI interface ctob() is used for
scaling.  ctob() is:

	/* clicks to bytes */
	#define ctob(x)	((x)<<PAGE_SHIFT)

when (x) has type int (like the physmem variable), this overflows to
a negative int when the result should be >= 2GB.  When the result
should be 4GB, it overflows to 0.  When the result should be > 4GB,
it overflows worse.  Next, this int is passed to sysctl_int() which
copies it out using:

	error = SYSCTL_IN(req, arg1, sizeof(int));

so userland sees an int.  Peter's change just hides the previous overflows
by making sysctl(8) print this int using the bogus format %u.

On alphas, the scaling code is actually correct.  ctob() is broken on
alphas in the same was as on i386's, but the sysctl doesn't use it; it
uses alpha_ptob(), which works up to size (2^64 - 1 page).  The result
of alpha_ptob() is unsigned long.  Next, this unsigned long is corrupted
to an int by passing it to sysctl_int().  The remaining bugs are the same
as on i386's.

There are many other bugs and bogons in the conversion macros in
<machine/param.h>.  Most seriously, dbtob() is broken on alphas.

Bruce


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message