Re: [PATCH] lib/bitmap: waive const_eval test as it breaks the build

[Nick Desaulniers]

On Mon, Jul 17, 2023 at 1:31 PM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
>
> Hi Yury,
>
> On Mon, Jul 17, 2023 at 12:58:13PM -0700, Yury Norov wrote:
> > When building with clang, and when KASAN and GCOV_PROFILE_ALL are both
> > enabled, the test fails to build [1]:
> >
> > >> lib/test_bitmap.c:920:2: error: call to '__compiletime_assert_239' declared with 'error' attribute: BUILD_BUG_ON failed: !__builtin_constant_p(res)
> >            BUILD_BUG_ON(!__builtin_constant_p(res));
> >            ^
> >    include/linux/build_bug.h:50:2: note: expanded from macro 'BUILD_BUG_ON'
> >            BUILD_BUG_ON_MSG(condition, "BUILD_BUG_ON failed: " #condition)
> >            ^
> >    include/linux/build_bug.h:39:37: note: expanded from macro 'BUILD_BUG_ON_MSG'
> >    #define BUILD_BUG_ON_MSG(cond, msg) compiletime_assert(!(cond), msg)
> >                                        ^
> >    include/linux/compiler_types.h:352:2: note: expanded from macro 'compiletime_assert'
> >            _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
> >            ^
> >    include/linux/compiler_types.h:340:2: note: expanded from macro '_compiletime_assert'
> >            __compiletime_assert(condition, msg, prefix, suffix)
> >            ^
> >    include/linux/compiler_types.h:333:4: note: expanded from macro '__compiletime_assert'
> >                            prefix ## suffix();                             \
> >                            ^
> >    <scratch space>:185:1: note: expanded from here
> >    __compiletime_assert_239
> >
> > Originally it was attributed to s390, which now looks seemingly wrong. The
> > issue is also not related to bitmap code itself, but it breaks build for
> > a given configuration. So, disabling the test unless the compiler will
> > get fixed.
> >
> > [1] https://github.com/ClangBuiltLinux/linux/issues/1874
> >
> > Fixes: dc34d5036692 ("lib: test_bitmap: add compile-time optimization/evaluations assertions")
> > Signed-off-by: Yury Norov <yury.norov@xxxxxxxxx>
> > ---
> >  lib/test_bitmap.c | 17 +++++------------
> >  1 file changed, 5 insertions(+), 12 deletions(-)
> >
> > diff --git a/lib/test_bitmap.c b/lib/test_bitmap.c
> > index 187f5b2db4cf..a791fdb7a8c9 100644
> > --- a/lib/test_bitmap.c
> > +++ b/lib/test_bitmap.c
> > @@ -1163,6 +1163,9 @@ static void __init test_bitmap_print_buf(void)
> >
> >  static void __init test_bitmap_const_eval(void)
> >  {
> > +#if defined(CONFIG_CC_IS_CLANG) && defined(CONFIG_KASAN) && defined(CONFIG_GCOV_PROFILE_ALL)
> > +#warning "FIXME: Clang breaks compile time evaluations when KASAN and GCOV are enabled"
>
> Making this a '#warning' will basically just replace the current error
> with a different one in the face of CONFIG_WERROR, which seems pointless
> to me:
>
>   lib/test_bitmap.c:1167:2: error: "FIXME: Clang breaks compile time evaluations when KASAN and GCOV are enabled" [-Werror,-W#warnings]
>    1167 | #warning "FIXME: Clang breaks compile time evaluations when KASAN and GCOV are enabled"
>         |  ^
>   1 error generated.
>
> Could we just opt out of GCOV for test_bitmap.c if KASAN is enabled with
> clang? That does not seem too bad of a workaround. I highly doubt there
> are many people who are interested in debugging test_bitmap.c with KASAN
> while profiling it with GCOV when building with clang, since they would
> have hit this already and reported it already; as far as I can tell,
> only the Intel robot has reported this with a randconfig.

Yury, Nathan,
Thanks for the patches and discussion.  Yes, I think a combo of:

1. Nathan's lib/Makefile change plus
2. Yury's removal of the current preprocessor guards in lib/test_bitmap.c

make most sense.


>
> diff --git a/lib/Makefile b/lib/Makefile
> index 42d307ade225..c6f60ae42769 100644
> --- a/lib/Makefile
> +++ b/lib/Makefile
> @@ -83,6 +83,10 @@ obj-$(CONFIG_TEST_DYNAMIC_DEBUG) += test_dynamic_debug.o
>  obj-$(CONFIG_TEST_PRINTF) += test_printf.o
>  obj-$(CONFIG_TEST_SCANF) += test_scanf.o
>  obj-$(CONFIG_TEST_BITMAP) += test_bitmap.o
> +# https://github.com/ClangBuiltLinux/linux/issues/1874
> +ifeq ($(CONFIG_CC_IS_CLANG)$(CONFIG_KASAN),yy)
> +GCOV_PROFILE_test_bitmap.o := n
> +endif
>  obj-$(CONFIG_TEST_UUID) += test_uuid.o
>  obj-$(CONFIG_TEST_XARRAY) += test_xarray.o
>  obj-$(CONFIG_TEST_MAPLE_TREE) += test_maple_tree.o
>
> Cheers,
> Nathan
>
> > +#else
> >       DECLARE_BITMAP(bitmap, BITS_PER_LONG);
> >       unsigned long initvar = BIT(2);
> >       unsigned long bitopvar = 0;
> > @@ -1177,20 +1180,9 @@ static void __init test_bitmap_const_eval(void)
> >        * in runtime.
> >        */
> >
> > -     /*
> > -      * Equals to `unsigned long bitmap[1] = { GENMASK(6, 5), }`.
> > -      * Clang on s390 optimizes bitops at compile-time as intended, but at
> > -      * the same time stops treating @bitmap and @bitopvar as compile-time
> > -      * constants after regular test_bit() is executed, thus triggering the
> > -      * build bugs below. So, call const_test_bit() there directly until
> > -      * the compiler is fixed.
> > -      */
> > +     /* Equals to `unsigned long bitmap[1] = { GENMASK(6, 5), }` */
> >       bitmap_clear(bitmap, 0, BITS_PER_LONG);
> > -#if defined(__s390__) && defined(__clang__)
> > -     if (!const_test_bit(7, bitmap))
> > -#else
> >       if (!test_bit(7, bitmap))
> > -#endif
> >               bitmap_set(bitmap, 5, 2);
> >
> >       /* Equals to `unsigned long bitopvar = BIT(20)` */
> > @@ -1220,6 +1212,7 @@ static void __init test_bitmap_const_eval(void)
> >       /* ~BIT(25) */
> >       BUILD_BUG_ON(!__builtin_constant_p(~var));
> >       BUILD_BUG_ON(~var != ~BIT(25));
> > +#endif
> >  }
> >
> >  static void __init selftest(void)
> > --
> > 2.39.2
> >
>


-- 
Thanks,
~Nick Desaulniers