linux-kernel - Re: [RFC] Add missing #include <linux/bug.h>

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Mon, 25 Apr 2016 11:36:41 -0700
From:	Andrew Morton <akpm@...ux-foundation.org>
To:	Eric Engestrom <eric.engestrom@...tec.com>
Cc:	<linux-kernel@...r.kernel.org>, Joe Perches <joe@...ches.com>,
	Hannes Frederic Sowa <hannes@...essinduktion.org>,
	Rasmus Villemoes <linux@...musvillemoes.dk>,
	"David S. Miller" <davem@...emloft.net>,
	Daniel Borkmann <daniel@...earbox.net>,
	Andrey Ryabinin <aryabinin@...tuozzo.com>
Subject: Re: [RFC] Add missing #include <linux/bug.h>

On Mon, 25 Apr 2016 19:10:55 +0100 Eric Engestrom <eric.engestrom@...tec.com> wrote:

> ARRAY_SIZE uses BUILD_BUG_ON_ZERO, which is undefined is you don't
> include linux/bug.h first, which just happened to me.
> 
> Is there any reason this include isn't here? A quick grep found 595
> other files using a define from bug.h without ever including it.

yeah.  A lot of the kernel compiles by accident ;) We just fix stuff
when it blows up - it ain't pretty but this process works acceptably
well.

> --- a/include/linux/compiler-gcc.h
> +++ b/include/linux/compiler-gcc.h
> @@ -2,6 +2,8 @@
>  #error "Please don't include <linux/compiler-gcc.h> directly, include <linux/compiler.h> instead."
>  #endif
>  
> +#include <linux/bug.h>
> +
>  /*
>   * Common definitions for all gcc versions go here.
>   */

hm, it Seems Wrong to include bug.h into compiler.h - the latter is
such a low-level infrastructury thing.  As evidenced by the fact that
bug.h explicitly includes compiler.h!

The bottom line is that our headers are just too large and complex and
try to do too much stuff.  Going finer-grained is always a good fix,
but I don't think you want to be adding a new array_size.h!

I guess we could move __must_be_array() into kernel.h and use #ifdef
__GNUC__ (or whatever).  But that adds to kernel.h (another) dependency
on bug.h, which kernel.h doesn't (and probably shouldn't and possibly
can't) include.  Sucks.