Re: [PATCH 4/4] rt2x00: Fix power of 2 check

[Ivo van Doorn <ivdoorn@gmail.com>]

On Thursday 08 March 2007 22:14, Ivo van Doorn wrote:
> Cleanup code, make gcc happy, make sparse happy. What a happy patch ;)
> This cleans up the code used to check the validity of the register fields,
> this will also reduce the required memory usage while working with sparse.
> This patch was inspired on the codesuggestions from Linus.

Please cancel this 4th patch, the previous 3 were correct but this one is terribly wrong.

> +#define is_power_of_two(x)	( !((x) & ((x)-1)) )
> +#define low_bit_mask(x)		( ((x)-1) & ~(x) )
> +#define is_valid_mask(x)	is_power_of_two(1 + (x) + low_bit_mask(x))
> +
> +#define FIELD16(__mask)					\
> +({							\
> +	BUILD_BUG_ON(!__mask || !is_valid_mask(__mask));\
> +	(struct rt2x00_field16) {			\
> +		(__scanbit((__mask), 16) - 1), (__mask) \
> +	};						\
> +})
> +
> +#define FIELD32(__mask)					\
> +({							\
> +	BUILD_BUG_ON(!__mask || !is_valid_mask(__mask));\
> +	(struct rt2x00_field32) {			\
> +		(__scanbit((__mask), 32) - 1), (__mask) \
> +	};						\
> +})

The problem is here, I have used __scanbit to determine the first bit
that is set. __scanbit was used since this doesn't require a pointer to the value
while the other functions do require the pointer (and the macro receives the direct
value hence the reason for the __scanbit choice).
Unfortunately I did not check if __scanbit was available on all arches, and this one
seems to only excist for x86_64.. :S
I'll try to fix this asap and come with a new patch.

Ivo