linux-kernel - Re: [PATCH 2/2] include/linux/typecheck.h: Zero initialize dummy variables

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

Message-ID: <20250502023647.GC1744689@ax162>
Date: Thu, 1 May 2025 19:36:47 -0700
From: Nathan Chancellor <nathan@...nel.org>
To: Al Viro <viro@...iv.linux.org.uk>
Cc: Linus Torvalds <torvalds@...ux-foundation.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Masahiro Yamada <masahiroy@...nel.org>,
	Nicolas Schier <nicolas.schier@...ux.dev>,
	Nick Desaulniers <nick.desaulniers+lkml@...il.com>,
	Bill Wendling <morbo@...gle.com>,
	Justin Stitt <justinstitt@...gle.com>, linux-kbuild@...r.kernel.org,
	linux-kernel@...r.kernel.org, llvm@...ts.linux.dev,
	patches@...ts.linux.dev, stable@...r.kernel.org,
	Linux Kernel Functional Testing <lkft@...aro.org>,
	Marcus Seyfarth <m.seyfarth@...il.com>
Subject: Re: [PATCH 2/2] include/linux/typecheck.h: Zero initialize dummy
 variables

On Fri, May 02, 2025 at 03:05:34AM +0100, Al Viro wrote:
> On Thu, May 01, 2025 at 06:24:49PM -0700, Nathan Chancellor wrote:
> 
> > > How long has that been valid? Because this is certainly new to the
> > > kernel, and sparse does complain about this initializer.
> > 
> > As you noted, brace initialization for scalars appears to always be
> > valid (at least in my testing) but as Al points out, empty braces for
> > scalars is only supported in GCC 13+ and Clang 17+ (I think [1] was the
> > clang commit), so that is not going to fly...
> 
> From some digging around it looks like
> 	* {} for compounds had been an extension for quite a while
> 	* C++11 got it into standard, with semantics defined as "same
> value you get for static-duration variables of that type without an
> explicit initializer".  For scalar types as well, with the same
> semantics.
> 	* On C side that happened (again, with scalar types allowed)
> in 2022; N2912 is the first draft with that change already merged,
> N2913 is the corresponding editor's report, saying that change in question
> (N2900) got merged in January/February virtual meeting.
> 	IOW, C23 has it, no previous versions do.  For C17 this syntax
> is an error, and AFAICS you need at least -std=c2x or -std=gnu2x to have
> it acceptable.

Neat, thanks for digging around.

> We can make sparse accept it (either unconditionally or with sufficient
> -std in arguments), but that won't do a damn thing for cc(1).  Does
> clang (any version) really accept it with -std=gnu11?

Yes, it appears that both GCC and clang accept it even with -std=gnu89:
https://godbolt.org/z/GYKrKhTdf

The clang commit mentions that this is exposed to older C modes like the
GNU extension was.

I guess another option to locally silence the warning would be to insert
something like

  #if defined(__clang__) && __clang_major__ >= 21
  #define typecheck_init = {}
  #else
  #define typecheck_init
  #endif

  #define typecheck(type,x) \
  ({    type __dummy typecheck_init; \
        typeof(x) __dummy2 typecheck_init; \
        (void)(&__dummy == &__dummy2); \
        1; \
  })

but maybe that is just too ugly or worthless.

Cheers,
Nathan