Re: [linux-next:master 7012/7430] include/linux/compiler_types.h:328:38: error: call to '__compiletime_assert_183' declared with attribute error: unexpected size in kmalloc_index()

From: Hyeonggon Yoo <42.hyeyoo@gmail.com>
To: Nathan Chancellor <nathan@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>,
	kernel test robot <lkp@intel.com>,
	kbuild-all@lists.01.org,
	 Linux Memory Management List <linux-mm@kvack.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	 Nick Desaulniers <ndesaulniers@google.com>,
	 clang-built-linux <clang-built-linux@googlegroups.com>
Subject: Re: [linux-next:master 7012/7430] include/linux/compiler_types.h:328:38: error: call to '__compiletime_assert_183' declared with attribute error: unexpected size in kmalloc_index()
Date: Sat, 12 Jun 2021 09:31:51 +0900	[thread overview]
Message-ID: <CAB=+i9RJgjHP9RExixQx66hkWmEuK64nAaD+wH8SeDjxyMoFsw@mail.gmail.com> (raw)
In-Reply-To: <99de1f59-1e38-6410-86ff-0ea1f016c49f@kernel.org>

[-- Attachment #1: Type: text/plain, Size: 1797 bytes --]

On Sat, Jun 12, 2021, 1:56 AM Nathan Chancellor <nathan@kernel.org> wrote:

> On 6/11/2021 1:58 AM, Hyeonggon Yoo wrote:> After playing with clang
> more a bit, I got to know that
> > compiletime_assert makes weird link error (undefined reference to
> > compiletime_assert_XXX), Not a compile error.
> >
> >
> > I think it's time to CC ClangBuiltLinux maintainers, who work on
> > clang/llvm build support.
> >
> > [+CC Nathan and Nick]
> >
> > I assumeed that compiletime_assert (in linux/compiler.h) will make
> > compiler error, but it makes no compile error, just makes weird link
> error.
> >
> > I'm not sure it it works well with clang, or somewhat buggy status?
>
> I am guessing this alone is why we were keyed into the thread so I am
> just going to respond to this.
>

Thank you for quick reply!

Unfortunately, this is a known issue with clang:
>
> https://github.com/ClangBuiltLinux/linux/issues/1173
>
> https://bugs.llvm.org/show_bug.cgi?id=16428

> As you noticed, building the full kernel will result in a link error but
> it would certainly be nicer if it were a compiler error. Something for
> us to improve indeed but I am not sure when we will be able to allocate
> resources for that.

I wanted to be sure if we can use compiletime_assert for clang. Then there
is room for improvement, but it seems okay.

Until then, you can build a full kernel to get the
> failing translation unit then use nm or readelf when building the single
> translation unit to see if there are any "__compiletime_assert" symbols.
>

Okay, then we should find symbol like that until improved.

It may confuse developer, but it seems okay for our code to support clang
as it meets minimal condition - build failure.

And I hope it become improved in future!

Thanks,
Hyeonggon

Cheers,
> Nathan
>

[-- Attachment #2: Type: text/html, Size: 3515 bytes --]