Re: [PATCH] linux/bits.h: Fix compilation error with GENMASK

[Yury Norov <yury.norov@gmail.com>]

On Thu, May 20, 2021 at 08:46:40PM +0200, Rikard Falkeborn wrote:
> On Wed, May 12, 2021 at 02:53:27PM +0200, Ard Biesheuvel wrote:
> > (+ Arnd)
> > 
> > On Tue, 11 May 2021 at 22:37, Rikard Falkeborn
> > <rikard.falkeborn@gmail.com> wrote:
> > >
> > > GENMASK() has an input check which uses __builtin_choose_expr() to enable
> > > a compile time sanity check of its inputs if they are known at compile
> > > time. However, it turns out that __builtin_constant_p() does not always
> > > return a compile time constant [0]. It was thought this problem was fixed
> > > with gcc 4.9 [1], but apparently this is not the case [2].
> > >
> > > Switch to use __is_constexpr() instead which always returns a compile
> > > time constant, regardless of its inputs.
> > >
> > > [0]: https://lore.kernel.org/lkml/42b4342b-aefc-a16a-0d43-9f9c0d63ba7a@rasmusvillemoes.dk
> > > [1]: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=19449
> > > [2]: https://lore.kernel.org/lkml/1ac7bbc2-45d9-26ed-0b33-bf382b8d858b@I-love.SAKURA.ne.jp
> > >
> > > Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> > > Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com>
> > > ---
> > > Feedback on placing __is_constexpr() in const.h is welcome, at least the
> > > name is appropriate...
> > >
> > >  include/linux/bits.h        |  2 +-
> > >  include/linux/const.h       |  8 ++++++++
> > >  include/linux/minmax.h      | 10 ++--------
> > >  tools/include/linux/bits.h  |  2 +-
> > >  tools/include/linux/const.h |  8 ++++++++
> > >  5 files changed, 20 insertions(+), 10 deletions(-)
> > >
> > > diff --git a/include/linux/bits.h b/include/linux/bits.h
> > > index 7f475d59a097..87d112650dfb 100644
> > > --- a/include/linux/bits.h
> > > +++ b/include/linux/bits.h
> > > @@ -22,7 +22,7 @@
> > >  #include <linux/build_bug.h>
> > >  #define GENMASK_INPUT_CHECK(h, l) \
> > >         (BUILD_BUG_ON_ZERO(__builtin_choose_expr( \
> > > -               __builtin_constant_p((l) > (h)), (l) > (h), 0)))
> > > +               __is_constexpr((l) > (h)), (l) > (h), 0)))
> > >  #else
> > >  /*
> > >   * BUILD_BUG_ON_ZERO is not available in h files included from asm files,
> > > diff --git a/include/linux/const.h b/include/linux/const.h
> > > index 81b8aae5a855..435ddd72d2c4 100644
> > > --- a/include/linux/const.h
> > > +++ b/include/linux/const.h
> > > @@ -3,4 +3,12 @@
> > >
> > >  #include <vdso/const.h>
> > >
> > > +/*
> > > + * This returns a constant expression while determining if an argument is
> > > + * a constant expression, most importantly without evaluating the argument.
> > > + * Glory to Martin Uecker <Martin.Uecker@med.uni-goettingen.de>
> > > + */
> > > +#define __is_constexpr(x) \
> > > +       (sizeof(int) == sizeof(*(8 ? ((void *)((long)(x) * 0l)) : (int *)8)))
> > > +
> > >  #endif /* _LINUX_CONST_H */
> > > diff --git a/include/linux/minmax.h b/include/linux/minmax.h
> > > index c0f57b0c64d9..5433c08fcc68 100644
> > > --- a/include/linux/minmax.h
> > > +++ b/include/linux/minmax.h
> > > @@ -2,6 +2,8 @@
> > >  #ifndef _LINUX_MINMAX_H
> > >  #define _LINUX_MINMAX_H
> > >
> > > +#include <linux/const.h>
> > > +
> > >  /*
> > >   * min()/max()/clamp() macros must accomplish three things:
> > >   *
> > > @@ -17,14 +19,6 @@
> > >  #define __typecheck(x, y) \
> > >         (!!(sizeof((typeof(x) *)1 == (typeof(y) *)1)))
> > >
> > > -/*
> > > - * This returns a constant expression while determining if an argument is
> > > - * a constant expression, most importantly without evaluating the argument.
> > > - * Glory to Martin Uecker <Martin.Uecker@med.uni-goettingen.de>
> > > - */
> > > -#define __is_constexpr(x) \
> > > -       (sizeof(int) == sizeof(*(8 ? ((void *)((long)(x) * 0l)) : (int *)8)))
> > > -
> > >  #define __no_side_effects(x, y) \
> > >                 (__is_constexpr(x) && __is_constexpr(y))
> > >
> > > diff --git a/tools/include/linux/bits.h b/tools/include/linux/bits.h
> > > index 7f475d59a097..87d112650dfb 100644
> > > --- a/tools/include/linux/bits.h
> > > +++ b/tools/include/linux/bits.h
> > > @@ -22,7 +22,7 @@
> > >  #include <linux/build_bug.h>
> > >  #define GENMASK_INPUT_CHECK(h, l) \
> > >         (BUILD_BUG_ON_ZERO(__builtin_choose_expr( \
> > > -               __builtin_constant_p((l) > (h)), (l) > (h), 0)))
> > > +               __is_constexpr((l) > (h)), (l) > (h), 0)))
> > >  #else
> > >  /*
> > >   * BUILD_BUG_ON_ZERO is not available in h files included from asm files,
> > > diff --git a/tools/include/linux/const.h b/tools/include/linux/const.h
> > > index 81b8aae5a855..435ddd72d2c4 100644
> > > --- a/tools/include/linux/const.h
> > > +++ b/tools/include/linux/const.h
> > > @@ -3,4 +3,12 @@
> > >
> > >  #include <vdso/const.h>
> > >
> > > +/*
> > > + * This returns a constant expression while determining if an argument is
> > > + * a constant expression, most importantly without evaluating the argument.
> > > + * Glory to Martin Uecker <Martin.Uecker@med.uni-goettingen.de>
> > > + */
> > > +#define __is_constexpr(x) \
> > > +       (sizeof(int) == sizeof(*(8 ? ((void *)((long)(x) * 0l)) : (int *)8)))
> > > +
> > >  #endif /* _LINUX_CONST_H */
> > > --
> > > 2.31.1
> > >
> 
> Friendly ping.

I added this patch in my Bitmap tree, but since it's actually a build
bug fix, I think it's worth to move it faster with Arnd's or Andrew's
trees?