Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Thu, 07 May 2020 00:02:32 -0700
User-Agent: K-9 Mail for Android
In-Reply-To: <CAMzpN2idWF2_4wtPebM2B2HVyksknr9hAqK8HJi_vjQ06bgu2g@mail.gmail.com>
References: <20200505174423.199985-1-ndesaulniers@google.com> <CAMzpN2idWF2_4wtPebM2B2HVyksknr9hAqK8HJi_vjQ06bgu2g@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable
Subject: Re: [PATCH] x86: bitops: fix build regression
To:     Brian Gerst <brgerst@gmail.com>,
        Nick Desaulniers <ndesaulniers@google.com>
CC:     Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Sedat Dilek <sedat.dilek@gmail.com>,
        stable <stable@vger.kernel.org>,
        Jesse Brandeburg <jesse.brandeburg@intel.com>,
        "kernelci . org bot" <bot@kernelci.org>,
        Andy Shevchenko <andriy.shevchenko@intel.com>,
        Ilie Halip <ilie.halip@gmail.com>,
        the arch/x86 maintainers <x86@kernel.org>,
        Marco Elver <elver@google.com>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        "Peter Zijlstra (Intel)" <peterz@infradead.org>,
        Daniel Axtens <dja@axtens.net>,
        Masahiro Yamada <yamada.masahiro@socionext.com>,
        Luc Van Oostenryck <luc.vanoostenryck@gmail.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        clang-built-linux@googlegroups.com
From:   hpa@zytor.com
Message-ID: <6A99766A-59FB-42DF-9350-80EA671A42B0@zytor.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On May 6, 2020 11:18:09 PM PDT, Brian Gerst <brgerst@gmail=2Ecom> wrote:
>On Tue, May 5, 2020 at 1:47 PM Nick Desaulniers
><ndesaulniers@google=2Ecom> wrote:
>>
>> From: Sedat Dilek <sedat=2Edilek@gmail=2Ecom>
>>
>> It turns out that if your config tickles __builtin_constant_p via
>> differences in choices to inline or not, this now produces invalid
>> assembly:
>>
>> $ cat foo=2Ec
>> long a(long b, long c) {
>>   asm("orb\t%1, %0" : "+q"(c): "r"(b));
>>   return c;
>> }
>> $ gcc foo=2Ec
>> foo=2Ec: Assembler messages:
>> foo=2Ec:2: Error: `%rax' not allowed with `orb'
>>
>> The "q" constraint only has meanting on -m32 otherwise is treated as
>> "r"=2E
>>
>> This is easily reproducible via
>Clang+CONFIG_STAGING=3Dy+CONFIG_VT6656=3Dm,
>> or Clang+allyesconfig=2E
>>
>> Keep the masking operation to appease sparse (`make C=3D1`), add back
>the
>> cast in order to properly select the proper 8b register alias=2E
>>
>>  [Nick: reworded]
>>
>> Cc: stable@vger=2Ekernel=2Eorg
>> Cc: Jesse Brandeburg <jesse=2Ebrandeburg@intel=2Ecom>
>> Link: https://github=2Ecom/ClangBuiltLinux/linux/issues/961
>> Link:
>https://lore=2Ekernel=2Eorg/lkml/20200504193524=2EGA221287@google=2Ecom/
>> Fixes: 1651e700664b4 ("x86: Fix bitops=2Eh warning with a moved cast")
>> Reported-by: Sedat Dilek <sedat=2Edilek@gmail=2Ecom>
>> Reported-by: kernelci=2Eorg bot <bot@kernelci=2Eorg>
>> Suggested-by: Andy Shevchenko <andriy=2Eshevchenko@intel=2Ecom>
>> Suggested-by: Ilie Halip <ilie=2Ehalip@gmail=2Ecom>
>> Tested-by: Sedat Dilek <sedat=2Edilek@gmail=2Ecom>
>> Signed-off-by: Sedat Dilek <sedat=2Edilek@gmail=2Ecom>
>> Signed-off-by: Nick Desaulniers <ndesaulniers@google=2Ecom>
>> ---
>>  arch/x86/include/asm/bitops=2Eh | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/x86/include/asm/bitops=2Eh
>b/arch/x86/include/asm/bitops=2Eh
>> index b392571c1f1d=2E=2E139122e5b25b 100644
>> --- a/arch/x86/include/asm/bitops=2Eh
>> +++ b/arch/x86/include/asm/bitops=2Eh
>> @@ -54,7 +54,7 @@ arch_set_bit(long nr, volatile unsigned long *addr)
>>         if (__builtin_constant_p(nr)) {
>>                 asm volatile(LOCK_PREFIX "orb %1,%0"
>>                         : CONST_MASK_ADDR(nr, addr)
>> -                       : "iq" (CONST_MASK(nr) & 0xff)
>> +                       : "iq" ((u8)(CONST_MASK(nr) & 0xff))
>
>I think a better fix would be to make CONST_MASK() return a u8 value
>rather than have to cast on every use=2E
>
>Also I question the need for the "q" constraint=2E  It was added in
>commit 437a0a54 as a workaround for GCC 3=2E4=2E4=2E  Now that the minimu=
m
>GCC version is 4=2E6, is this still necessary?
>
>--
>Brian Gerst

Yes, "q" is needed on i386=2E
--=20
Sent from my Android device with K-9 Mail=2E Please excuse my brevity=2E