On Wed, Aug 28, 2019 at 10:53 AM Nathan Chancellor natechancellor@gmail.com wrote:
On Wed, Aug 28, 2019 at 11:43:53PM +1000, Michael Ellerman wrote:
Nathan Chancellor natechancellor@gmail.com writes:
Commit aea447141c7e ("powerpc: Disable -Wbuiltin-requires-header when setjmp is used") disabled -Wbuiltin-requires-header because of a warning about the setjmp and longjmp declarations.
r367387 in clang added another diagnostic around this, complaining that there is no jmp_buf declaration.
In file included from ../arch/powerpc/xmon/xmon.c:47: ../arch/powerpc/include/asm/setjmp.h:10:13: error: declaration of built-in function 'setjmp' requires the declaration of the 'jmp_buf' type, commonly provided in the header <setjmp.h>. [-Werror,-Wincomplete-setjmp-declaration] extern long setjmp(long *); ^ ../arch/powerpc/include/asm/setjmp.h:11:13: error: declaration of built-in function 'longjmp' requires the declaration of the 'jmp_buf' type, commonly provided in the header <setjmp.h>. [-Werror,-Wincomplete-setjmp-declaration] extern void longjmp(long *, long); ^ 2 errors generated.
Take the same approach as the above commit by disabling the warning for the same reason, we provide our own longjmp/setjmp function.
Cc: stable@vger.kernel.org # 4.19+ Link: https://github.com/ClangBuiltLinux/linux/issues/625 Link: https://github.com/llvm/llvm-project/commit/3be25e79477db2d31ac46493d97eca8c... Signed-off-by: Nathan Chancellor natechancellor@gmail.com
It may be worth using -fno-builtin-setjmp and -fno-builtin-longjmp instead as it makes it clear to clang that we are not using the builtin longjmp and setjmp functions, which I think is why these warnings are appearing (at least according to the commit that introduced this waring).
Sample patch: https://github.com/ClangBuiltLinux/linux/issues/625#issuecomment-519251372
Couldn't we just add those flags to CFLAGS for the whole kernel? Rather than making them per-file.
Yes, I don't think this would be unreasonable. Are you referring to the cc-disable-warning flags or the -fno-builtin flags? I personally think the -fno-builtin flags convey to clang what the kernel is intending to do better than disabling the warnings outright.
The `-f` family of flags have dire implications for codegen, I'd really prefer we think long and hard before adding/removing them to suppress warnings. I don't think it's a solution for this particular problem.
I mean there's no kernel code that wants to use clang's builtin setjmp/longjmp implementation at all right?
cheers
I did a quick search of the tree and it looks like powerpc and x86/um are the only architectures that do anything with setjmp/longjmp. x86/um avoids this by using a define flag to change setjmp to kernel_setjmp:
arch/um/Makefile: -Dlongjmp=kernel_longjmp -Dsetjmp=kernel_setjmp \
Seems like adding those flags should be safe.
Cheers, Nathan