linux-kernel - RE: [PATCH] scsi: aic7xxx: Fix unintended sign extension issue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <483f04b2ad5a4858b56c2be52cea625e@AcuMS.aculab.com>
Date:   Thu, 25 Oct 2018 15:54:38 +0000
From:   David Laight <David.Laight@...LAB.COM>
To:     'James Bottomley' <jejb@...ux.vnet.ibm.com>,
        Colin King <colin.king@...onical.com>,
        Hannes Reinecke <hare@...e.com>,
        "Martin K . Petersen" <martin.petersen@...cle.com>,
        "linux-scsi@...r.kernel.org" <linux-scsi@...r.kernel.org>
CC:     "kernel-janitors@...r.kernel.org" <kernel-janitors@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [PATCH] scsi: aic7xxx: Fix unintended sign extension issue

From: James Bottomley
> Sent: 25 October 2018 16:33
> 
> On Thu, 2018-10-25 at 16:13 +0100, Colin King wrote:
> > From: Colin Ian King <colin.king@...onical.com>
> >
> > In the expression "ahc_inb(ahc, port+3) << 24", the initial value is
> > a u8, but is promoted to a signed int, then sign-extended to
> > uint64_t.
> 
> Why is this, that's highly non intuitive?  The compiler is supposed to
> promote to the biggest type, which is uint64_t and then do the
> calculation

Do not doubt the wisdom on the ANSI C committee that decided to do
'value preserving' integer promotions instead of the 'sign preserving'
ones of K&R C.

So 'unsigned char' is promoted to 'int' almost everywhere it is used
(unless they are both the same size - which is allowed).
This means that ahc_inb() << 24 is actually undefined (signed integer
overflow can do anything it likes).

By far the best fix is to change the return type of ahc_inb() to
be 'unsigned int'.
On systems without byte sized registers (about everything except x86)
this will almost certainly generate better code.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)