Re: [PATCH] ASoC: soc-dapm: Fix comparison of pointers

[Rasmus Villemoes <linux@rasmusvillemoes.dk>]

On Fri, Dec 12 2014, Mark Brown <broonie@kernel.org> wrote:

> On Tue, Dec 09, 2014 at 10:53:41PM +0100, Rasmus Villemoes wrote:
>
>>  	if (a->reg != b->reg)
>>  		return a->reg - b->reg;
>> -	if (a->dapm != b->dapm)
>> -		return (unsigned long)a->dapm - (unsigned long)b->dapm;
>> +	if (a->dapm < b->dapm)
>> +		return -1;
>> +	if (a->dapm > b->dapm)
>> +		return 1;
>
> If we're worrying about standards conformance type stuff here this is
> also buggy since it's out of spec to compare pointers that are not part
> of the same array like this.  Casting to uintptr_t is a better fix here.

I'm not worrying about standards conformance. Casting pointers to and
from [unsigned] long is done everywhere in the kernel, and that's
fine. But subtracting ints or longs and using the sign of the result for
comparison is simply wrong. See acbbe6fbb2 (kcmp: fix standard
comparison bug) for the long story. In this case the narrowing (when
BITS_PER_LONG==64) of the result from long to int just makes it even
more wrong.

Rasmus