Re: [PATCH] log2.h: Macro-ize is_power_of_2() for use in BUILD_BUG_ON

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 06 Jan 2010 12:21:13 -0800
Roland Dreier <rdreier@cisco.com> wrote:

> When code relies on a constant being a power of 2:
> 
> 	#define FOO	512	/* must be a power of 2 */
> 
> it would be nice to be able to do:
> 
> 	BUILD_BUG_ON(!is_power_of_2(FOO));
> 
> However applying an inline function does not result in a compile-time
> constant that can be used with BUILD_BUG_ON(), so trying that gives
> results in:
> 
> 	error: bit-field '<anonymous>' width not an integer constant
> 
> We can fix this by changing is_power_of_2() to a macro; we leave the
> inline function for the non-constant case, to avoid evaluating the
> parameter more than once.  (gcc does not accept a multi-statement
> expression like "({ unsigned long __n = n; ... })" as a compile-time
> constant so that solution doesn't work)
> 
> Signed-off-by: Roland Dreier <rolandd@cisco.com>
> ---
> This is somewhat of an RFC -- I'm a bit undecided whether it's really
> worth making this change.  It's prompted by
> http://www.mail-archive.com/linux-rdma@vger.kernel.org/msg01941.html and
> http://www.mail-archive.com/linux-rdma@vger.kernel.org/msg01950.html
> which do ugly things to work around is_power_of_2 not being usable in
> BUILD_BUG_ON().
> 
> On the other hand maybe just
> 
> /* FOO must be a power of 2 */
> #define FOO_SHIFT 9
> #define FOO (1 << FOO_SHIFT)
> 
> is good enough.
> 
>  include/linux/log2.h |    9 ++++++++-
>  1 files changed, 8 insertions(+), 1 deletions(-)
> 
> diff --git a/include/linux/log2.h b/include/linux/log2.h
> index 25b8086..248f69c 100644
> --- a/include/linux/log2.h
> +++ b/include/linux/log2.h
> @@ -49,11 +49,18 @@ int __ilog2_u64(u64 n)
>   */
>  
>  static inline __attribute__((const))
> -bool is_power_of_2(unsigned long n)
> +bool __is_power_of_2(unsigned long n)
>  {
>  	return (n != 0 && ((n & (n - 1)) == 0));
>  }
>  
> +#define is_power_of_2(n)					\
> +(								\
> +	__builtin_constant_p(n) ?				\
> +		(((n) != 0) && (((n) & ((n) - 1)) == 0)) :	\
> +		__is_power_of_2(n)				\
> +)
> +

We've had recurring struggles with various versions of gcc screwing up
constructs of this form and trying to emit the non-constant code when
the arg was clearly a compile-time constant.  One episode which comes
to mind was when we made changes to kmalloc().

Of course, that might not bite us in this case - it would need a lot of
coverage testing to find out.


Perhaps we can avoid worrying about that via

#define BUILD_BUG_ON_NOT_POWER_OF_2(n)	\
	BUILD_BUG_ON((n != 0 && ((n & (n - 1)) == 0)))

?