Date: Mon, 29 Jul 2013 06:54:11 -0700
From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: caf@guarana.org
Cc: linux-kernel@vger.kernel.org, john.stultz@linaro.org, davem@davemloft.net,
        arnd@arndb.de, mingo@kernel.org, torvalds@linux-foundation.org
Subject: Re: [PATCH jiffies] Avoid undefined behavior from signed overflow
Message-ID: <20130729135411.GM26694@linux.vnet.ibm.com>
Reply-To: paulmck@linux.vnet.ibm.com
References: <20130727225828.GA11864@linux.vnet.ibm.com>
 <20130729153035.18805zox54aoyubf@chicago.guarana.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20130729153035.18805zox54aoyubf@chicago.guarana.org>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2323
Lines: 46

On Mon, Jul 29, 2013 at 03:30:35PM +1000, caf@guarana.org wrote:
> Quoting "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>:
> 
> >According to the C standard 3.4.3p3, overflow of a signed integer results
> >in undefined behavior.  This commit therefore changes the definitions
> >of time_after() and time_after_eq() to avoid this undefined behavior.
> >The trick is that the subtraction is done using unsigned arithmetic,
> >which according to 6.2.5p9 cannot overflow because it is defined as
> >modulo arithmetic.  This has the added (though admittedly quite small)
> >benefit of shortening two lines of code by four characters each.
> >
> >Note that the C standard considers the cast from signed to
> >unsigned to be implementation-defined, see 6.3.1.3p3.  However, on a
> >two-complement system, an implementation that defines anything other
> >than a reinterpretation of the bits is free come to me, and I will be
> >happy to act as a witness for its being committed to an insane asylum.
> >(Although I have nothing against saturating arithmetic or signals in
> >some cases, these things really should not be the default.)
> 
> Don't worry, the case from signed to unsigned is actually well-defined -
> the relevant part is 6.3.1.3p2 (in C99):
> 
> >Otherwise, if the new type is unsigned, the value is converted by
> >repeatedly adding or subtracting one more than the maximum value that
> >can be represented in the new type until the value is in the range of
> >the new type.

Yep, but we are going in the other direction, from unsigned to signed.

> ...which ends up just being reinterpretation of the bits on a two's
> complement system, as you'd hope (after sign-extension to the width of
> the target unsigned type, that is).  This actually means if you were
> mad enough to implement C on a sign-magnitude system, you'd be forced to
> do a non-trivial conversion in this case.

Fortunately, I never used signed-magnitude systems.  And even when I used
ones-complement systems back in my misguided youth, I didn't write C
programs for them.  ;-)

							Thanx, Paul

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/