linux-kernel - Re: [RFC PATCH 0/5] make use of gcc 9's "asm inline()"

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: <c3b71dff-c884-28b6-a4c7-ff2bf12720d4@rasmusvillemoes.dk>
Date:   Fri, 30 Aug 2019 09:45:07 +0200
From:   Rasmus Villemoes <linux@...musvillemoes.dk>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     the arch/x86 maintainers <x86@...nel.org>,
        Linux List Kernel Mailing <linux-kernel@...r.kernel.org>,
        Ingo Molnar <mingo@...hat.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        "H. Peter Anvin" <hpa@...or.com>, Nadav Amit <namit@...are.com>,
        Miguel Ojeda <miguel.ojeda.sandonis@...il.com>,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>
Subject: Re: [RFC PATCH 0/5] make use of gcc 9's "asm inline()"

On 29/08/2019 18.05, Linus Torvalds wrote:
> On Thu, Aug 29, 2019 at 1:32 AM Rasmus Villemoes
> <linux@...musvillemoes.dk> wrote:
>>
>> But since we #define the identifier inline to attach some attributes,
>> we have to use the alternate spelling __inline__ of that
>> keyword. Unfortunately, we also currently #define that one (to
>> inline), so we first have to get rid of all (mis)uses of
>> __inline__. Hence the huge diffstat.
> 
> Ugh. Not pretty, but I guess we're stuck with it.
> 
> However, it worries me a bit that you excluide the UAPI headers where
> we still use "__inline__", and now the semantics of that will change
> for the kernel (for some odd gcc versions).

Yeah, as I wrote I was aware of that, but didn't have any good ideas, so
I was fishing. Doing

#ifdef __KERNEL__
#define __uapi_inline inline
#else
#define __uapi_inline __inline__
#endif

was one of the bad ideas...

> I suspect we should just bite the bullet and you should do it to the
> uapi headers too. We already use "inline" in a lot of them, so it's
> not the case that we're using __inline__ because of some namespace
> issue, as far as I can tell.
> 
> One option might be to just use "__inline" for the asm_inline thing.
> We have way fewer of those. That would make the noise much less for
> this patch series.

Ah, interesting. I didn't know that was also a compiler provided alias.
It seems to be entirely undocumented -
https://gcc.gnu.org/onlinedocs/gcc/Alternate-Keywords.html just mentions
the __inline__ (and __asm__) spellings. But it's clear as day in the gcc
sources

  { "__inline",         RID_INLINE,     0 },
  { "__inline__",       RID_INLINE,     0 },

and has been that way since forever, AFAICT.

So yes, let's just start by getting rid of the __inline define and fix
the staging (+acpi,zstd) users, to allow asm_inline to progress. I'll
respin.

Rasmus