Re: [PATCH next 08/14] bits: simplify GENMASK_TYPE()

[David Laight <david.laight.linux@gmail.com>]

On Sat, 7 Feb 2026 21:36:57 -0500
Yury Norov <ynorov@nvidia.com> wrote:

> On Wed, Jan 21, 2026 at 02:57:25PM +0000, david.laight.linux@gmail.com wrote:
> > From: David Laight <david.laight.linux@gmail.com>
> > 
> > Since the type is always unsigned (T)-1 is always the correct value
> > so there is no need to use type_max().
> > 
> > Signed-off-by: David Laight <david.laight.linux@gmail.com>  
> 
> Here you implicitly add a new restriction that all types passed in
> GENMASK_TYPE() must be unsigned. Can you enforce it with
> is_signed_type(), or at lease with a corresponding comment?

GENMASK_TYPE() is an internal define used in the expansions of,
for example, GENMASK_U32() - it is always passed an unsigned type.
I don't really think it should be used elsewhere at all.

In any case I don't think it makes any difference.
The value or type_max(t) has to match BITS_PER_TYPE(t) or else
the result is wrong, so it has never worked for signed types.

I did try adding a signedness check to FIELD_GET(), there is one
place in the x86-64 build that uses a signed type and requires the
right shift sign extend.
That code (in the fault handler) probably ought to be changed.

	David

> 
> > ---
> >  include/linux/bits.h | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> > 
> > diff --git a/include/linux/bits.h b/include/linux/bits.h
> > index a40cc861b3a7..697318f2a47d 100644
> > --- a/include/linux/bits.h
> > +++ b/include/linux/bits.h
> > @@ -45,8 +45,7 @@
> >   */
> >  #define GENMASK_TYPE(t, h, l)					\
> >  	((t)(GENMASK_INPUT_CHECK(h, l) +			\
> > -	     (type_max(t) << (l) &				\
> > -	      type_max(t) >> (BITS_PER_TYPE(t) - 1 - (h)))))
> > +	     ((t)-1 << (l) & (t)-1 >> (BITS_PER_TYPE(t) - 1 - (h)))))
> >  
> >  #define GENMASK(h, l)		GENMASK_TYPE(unsigned long, h, l)
> >  #define GENMASK_ULL(h, l)	GENMASK_TYPE(unsigned long long, h, l)
> > -- 
> > 2.39.5