linux-kernel - Re: [PATCH v2 4/6] compiler-gcc.h: add asm

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: <20190905144725.GQ9749@gate.crashing.org>
Date:   Thu, 5 Sep 2019 09:47:25 -0500
From:   Segher Boessenkool <segher@...nel.crashing.org>
To:     Rasmus Villemoes <linux@...musvillemoes.dk>
Cc:     Nick Desaulniers <ndesaulniers@...gle.com>,
        "maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)" <x86@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Miguel Ojeda <miguel.ojeda.sandonis@...il.com>,
        "gcc-patches@....gnu.org" <gcc-patches@....gnu.org>
Subject: Re: [PATCH v2 4/6] compiler-gcc.h: add asm_inline definition

On Thu, Sep 05, 2019 at 04:23:11PM +0200, Rasmus Villemoes wrote:
> On 05/09/2019 15.45, Segher Boessenkool wrote:
> > On Thu, Sep 05, 2019 at 01:07:11PM +0200, Rasmus Villemoes wrote:
> >> Perhaps something like below, though that
> >> won't affect the already released gcc 9.1 and 9.2, of course.
> > 
> > That is one reason to not want such a predefined macro.  Another reason
> > is that you usually need to compile some test programs *anyway*, to see if
> > some bug is present for example, or to see if the exact implementation of
> > the feature is beneficial (or harmful) to your program in some way.
> 
> OK, I think I'll just use a version check for now, and then switch to a
> Kconfig test if and when clang grows support.
> 
> >> gcc maintainers, WDYT? Can we add a feature test macro for asm inline()?
> > 
> > Why would GCC want to have macros for all features it has? 
> 
> Well, gcc has implemented __has_attribute() which is similar - one could
> detect support by compiling a trivial test program.

It is not a macro, it doesn't spill over the place, and it is for
detecting things that are already fixed strings, much easier to do :-)

> Or the same could be
> said for many of the predefined macros that are conditionally defined,
> e.g. __HAVE_SPECULATION_SAFE_VALUE.

That one happened because of the Great Security Scare of 2017/2018, it's
not a good precedent.  And, how it is set is target-specific, it can
depend on CPU model selected, target code generation options, or whatnot.

> But I was just throwing the question into the air, I won't pursue this
> further.

Maybe GCC should have a has_feature thing, it might fit in well there.
As preprocessor macros, not so much, IMO.


Segher