Re: [PATCH 05/21] powerpc: Avoid comparison of unsigned long >= 0 in pfn_valid

From: Segher Boessenkool
Date: Mon Feb 26 2018 - 03:46:41 EST


On Mon, Feb 26, 2018 at 07:32:03AM +0100, Christophe LEROY wrote:
> Le 25/02/2018 à 18:22, Mathieu Malaterre a écrit :
> >-#define pfn_valid(pfn) ((pfn) >= ARCH_PFN_OFFSET && (pfn) <
> >max_mapnr)
> >+#define pfn_valid(pfn) \
> >+ (((pfn) - ARCH_PFN_OFFSET) < (max_mapnr - ARCH_PFN_OFFSET))
>
> What will happen when ARCH_PFN_OFFSET is not nul and pfn is lower than
> ARCH_PFN_OFFSET ?

It will work fine.

Say you are asking for a <= x < b so (in actual integers, no overflow)
that is 0 <= x-a < b-a and you also assume x-a overflows, so that we
are actually comparing x-a+M < b-a with M = 2**32 or such (the maximum
value in the unsigned integer type plus one). This comparison is
obviously always false.

(It also works if b < a btw).


Segher