linux-kernel - Re: [PATCH 1/2] compiler.h: Avoid using inline asm operand modifiers

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: <your-ad-here.call-01636386038-ext-6578@work.hours>
Date:   Mon, 8 Nov 2021 16:40:38 +0100
From:   Vasily Gorbik <gor@...ux.ibm.com>
To:     Josh Poimboeuf <jpoimboe@...hat.com>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...nel.org>, Borislav Petkov <bp@...e.de>,
        Miroslav Benes <mbenes@...e.cz>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/2] compiler.h: Avoid using inline asm operand modifiers

On Fri, Nov 05, 2021 at 09:54:18AM -0700, Josh Poimboeuf wrote:
> On Wed, May 19, 2021 at 03:03:08PM +0200, Vasily Gorbik wrote:
> > The expansion of annotate_reachable/annotate_unreachable on s390 will
> > result in a compiler error if the __COUNTER__ value is high enough.
> > For example with "i" (154) the "%c0" operand of annotate_reachable
> > will be expanded to -102:
> > 
> >         -102:
> >         .pushsection .discard.reachable
> >         .long -102b - .
> >         .popsection
> > 
> > This is a quirk of the gcc backend for s390, it interprets the %c0
> > as a signed byte value. Avoid using operand modifiers in this case
> > by simply converting __COUNTER__ to string, with the same result,
> > but in an arch assembler independent way.
> > 
> > Signed-off-by: Vasily Gorbik <gor@...ux.ibm.com>
> 
> This patch causes these macros to break with Clang and
> CONFIG_TRACE_BRANCH_PROFILING.
> 
> I get a lot of warnings like
> 
>   arch/x86/kernel/traps.o: warning: objtool: handle_xfd_event()+0x134: unreachable instruction
> 
> Without an asm input, 'volatile' is ignored for some reason and Clang
> feels free to move the reachable() annotation away from its intended
> location.
> 
> I wonder if we could go back to the original approach of providing
> __COUNTER__ as an input to the asm, but then mask it to < 128.
> 
> Does this work on s390?
> 
> diff --git a/include/linux/compiler.h b/include/linux/compiler.h
> index 3d5af56337bd..42935500a712 100644
> --- a/include/linux/compiler.h
> +++ b/include/linux/compiler.h
> @@ -115,24 +115,18 @@ void ftrace_likely_update(struct ftrace_likely_data *f, int val,
>   * The __COUNTER__ based labels are a hack to make each instance of the macros
>   * unique, to convince GCC not to merge duplicate inline asm statements.
>   */
> -#define __stringify_label(n) #n
> -
> -#define __annotate_reachable(c) ({					\
> -	asm volatile(__stringify_label(c) ":\n\t"			\
> +#define annotate_reachable() ({						\
> +	asm volatile("%c0:\n\t"						\
>  		     ".pushsection .discard.reachable\n\t"		\
> -		     ".long " __stringify_label(c) "b - .\n\t"		\
> -		     ".popsection\n\t");				\
> +		     ".long %c0b - .\n\t"				\
> +		     ".popsection\n\t" : : "i" (__COUNTER__ & 0x7f));	\
>  })

hm, could we just add asm input back and not use it? and keep
__stringify_label(c) as is? would that work as well?