Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190712091239.716978-1-arnd@arndb.de> <20190712120249.GA27820@rkaganb.sw.ru>
 <CAK8P3a3+QSRQkitXiDFLYvyYvOS+Q4sXb=xA_XPeX2O2zQ5kgw@mail.gmail.com> <b7da5e91-f23c-9f5d-2c61-07e7fc7af9b1@redhat.com>
In-Reply-To: <b7da5e91-f23c-9f5d-2c61-07e7fc7af9b1@redhat.com>
From:   Arnd Bergmann <arnd@arndb.de>
Date:   Fri, 12 Jul 2019 15:32:19 +0200
Message-ID: <CAK8P3a1yiEBt0MgXJPKdnEMiTL6yUWsiT7Upi-kJAtqKJ5VDqw@mail.gmail.com>
Subject: Re: [PATCH 1/2] x86: kvm: avoid -Wsometimes-uninitized warning
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     Roman Kagan <rkagan@virtuozzo.com>,
        =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= <rkrcmar@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        "x86@kernel.org" <x86@kernel.org>,
        "H. Peter Anvin" <hpa@zytor.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Liran Alon <liran.alon@oracle.com>,
        "kvm@vger.kernel.org" <kvm@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "clang-built-linux@googlegroups.com" 
        <clang-built-linux@googlegroups.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Jul 12, 2019 at 3:14 PM Paolo Bonzini <pbonzini@redhat.com> wrote:
>
> On 12/07/19 15:02, Arnd Bergmann wrote:
> > I think what happens here is that clang does not treat the return
> > code of track the return code of is_64_bit_mode() as a constant
> > expression, and therefore assumes that the if() condition
> > may or may not be true, for the purpose of determining whether
> > the variable is used without an inialization. This would hold even
> > if it later eliminates the code leading up to the if() in an optimization
> > stage. IS_ENABLED(CONFIG_X86_64) however is a constant
> > expression, so with the patch, it understands this.
> >
> > In contrast, gcc seems to perform all the inlining first, and
> > then see if some variable is used uninitialized in the final code.
> > This gives additional information to the compiler, but makes the
> > outcome less predictable since it depends on optimization flags
> > and architecture specific behavior.
> >
> > Both approaches have their own sets of false positive and false
> > negative warnings.
>
> True, on the other hand constant returns are not really rocket science. :)
>
> Maybe change is_long_mode to a macro if !CONFIG_X86_64?  That would be
> better if clang likes it.

I had to also get rid of the temporary variable to make it work.
Sending v2 now.

      Arnd