Re: [PATCH v11 1/4] bitops: Introduce the for_each_set_clump macro

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Fri, Oct 16, 2020 at 04:23:05AM +0530, Syed Nayyar Waris wrote:
> On Tue, Oct 6, 2020 at 4:56 PM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > On Tue, Oct 06, 2020 at 02:52:16PM +0530, Syed Nayyar Waris wrote:

...

> > > +             return (map[index] >> offset) & GENMASK(nbits - 1, 0);
> >
> > Have you considered to use rather BIT{_ULL}(nbits) - 1?
> > It maybe better for code generation.
> 
> Yes I have considered using BIT{_ULL} in earlier versions of patchset.
> It has a problem:
> 
> This macro when used in both bitmap_get_value and
> bitmap_set_value functions, it will give unexpected results when nbits or clump
> size is BITS_PER_LONG (32 or 64 depending on arch).
> 
> Actually when nbits (clump size) is 64 (BITS_PER_LONG is 64, for example),
> (BIT(nbits) - 1)
> gives a value of zero and when this zero is ANDed with any value, it
> makes it full zero. This is unexpected, and incorrect calculation occurs.
> 
> What actually happens is in the macro expansion of BIT(64), that is 1
> << 64, the '1' overflows from leftmost bit position (most significant
> bit) and re-enters at the rightmost bit position (least significant
> bit), therefore 1 << 64 becomes '0x1', and when another '1' is
> subtracted from this, the final result becomes 0.
> 
> This is undefined behavior in the C standard (section 6.5.7 in the N1124)

I see, indeed, for 64/32 it is like this.

...

> Yes I have incorporated your suggestion to use the '<<' operator. Thank You.

One side note, consider the use round_up() vs. roundup(). I don't remember
which one is optimized to divisor being power of 2.

-- 
With Best Regards,
Andy Shevchenko