Re: [PATCH] compiler.h: Fix barrier_data() on clang

From: Nick Desaulniers <ndesaulniers@google.com>
To: Andreas Schwab <schwab@linux-m68k.org>,
	Palmer Dabbelt <palmerdabbelt@google.com>
Cc: Arvind Sankar <nivedita@alum.mit.edu>,
	Nathan Chancellor <natechancellor@gmail.com>,
	clang-built-linux <clang-built-linux@googlegroups.com>,
	LKML <linux-kernel@vger.kernel.org>,
	linux-riscv@lists.infradead.org
Subject: Re: [PATCH] compiler.h: Fix barrier_data() on clang
Date: Mon, 16 Nov 2020 11:31:13 -0800	[thread overview]
Message-ID: <CAKwvOdn8aOr0=22idJ_KetgJ_rhk5-7i29VvxofoLrRo138f_g@mail.gmail.com> (raw)
In-Reply-To: <87sg999ot0.fsf@igel.home>

On Mon, Nov 16, 2020 at 9:47 AM Andreas Schwab <schwab@linux-m68k.org> wrote:
>
> On Okt 14 2020, Arvind Sankar wrote:
>
> > Commit
> >   815f0ddb346c ("include/linux/compiler*.h: make compiler-*.h mutually exclusive")
> >
> > neglected to copy barrier_data() from compiler-gcc.h into
> > compiler-clang.h. The definition in compiler-gcc.h was really to work
> > around clang's more aggressive optimization, so this broke
> > barrier_data() on clang, and consequently memzero_explicit() as well.
> >
> > For example, this results in at least the memzero_explicit() call in
> > lib/crypto/sha256.c:sha256_transform() being optimized away by clang.
> >
> > Fix this by moving the definition of barrier_data() into compiler.h.
> >
> > Also move the gcc/clang definition of barrier() into compiler.h,
> > __memory_barrier() is icc-specific (and barrier() is already defined
> > using it in compiler-intel.h) and doesn't belong in compiler.h.
> >
> > Signed-off-by: Arvind Sankar <nivedita@alum.mit.edu>
> > Fixes: 815f0ddb346c ("include/linux/compiler*.h: make compiler-*.h mutually exclusive")
>
> This breaks build on riscv:
>
>   CC [M]  drivers/net/ethernet/emulex/benet/be_main.o
> In file included from ./include/vdso/processor.h:10,
>                  from ./arch/riscv/include/asm/processor.h:11,
>                  from ./include/linux/prefetch.h:15,
>                  from drivers/net/ethernet/emulex/benet/be_main.c:14:
> ./arch/riscv/include/asm/vdso/processor.h: In function 'cpu_relax':
> ./arch/riscv/include/asm/vdso/processor.h:14:2: error: implicit declaration of function 'barrier' [-Werror=implicit-function-declaration]
>    14 |  barrier();
>       |  ^~~~~~~
> cc1: some warnings being treated as errors
> make[5]: *** [scripts/Makefile.build:283: drivers/net/ethernet/emulex/benet/be_main.o] Error 1
> make[4]: *** [scripts/Makefile.build:500: drivers/net/ethernet/emulex/benet] Error 2
> make[3]: *** [scripts/Makefile.build:500: drivers/net/ethernet/emulex] Error 2
> make[2]: *** [scripts/Makefile.build:500: drivers/net/ethernet] Error 2
> make[1]: *** [scripts/Makefile.build:500: drivers/net] Error 2
> make: *** [Makefile:1799: drivers] Error 2
>
> Andreas.

A lot of VDSO's reset KBUILD_CFLAGS or use a new variable for their
compiler flags.  As such, they're missing `-include` command line flag
that injects include/linux/compiler_types.h, which `#includes`
numerous other headers if `__KERNEL__` is defined (`-D__KERNEL__`).
So the RISCV VDSO Makefile might need to `-include
$(srctree)/include/linux/compiler_types.h -D__KERNEL__`, or `#include
<linux/compiler.h>"` directly in
arch/riscv/include/asm/vdso/processor.h.
-- 
Thanks,
~Nick Desaulniers