Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:2 as permitted sender) client-ip=2620:137:e000::3:2;
Content-Type: text/plain;
        charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3731.700.6\))
Subject: Re: Missing clobber on alternative use on Linux UM 32-bit
From:   Nadav Amit <nadav.amit@gmail.com>
In-Reply-To: <a55f6941-ddf6-2355-271d-5ed0db5a2a62@kot-begemot.co.uk>
Date:   Sat, 4 Nov 2023 11:40:48 +0200
Cc:     Richard Weinberger <richard@nod.at>, linux-um@lists.infradead.org,
        Thomas Gleixner <tglx@linutronix.de>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Peter Zijlstra <peterz@infradead.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <1616BF59-5C18-46C7-86AB-92604B683652@gmail.com>
References: <24BD0906-C6FE-499C-9A4A-00C56E6EE84A@gmail.com>
 <a55f6941-ddf6-2355-271d-5ed0db5a2a62@kot-begemot.co.uk>
To:     Anton Ivanov <anton.ivanov@kot-begemot.co.uk>
Precedence: bulk


> On Nov 4, 2023, at 11:34 AM, Anton Ivanov =
<anton.ivanov@kot-begemot.co.uk> wrote:
>=20
> On 04/11/2023 09:25, Nadav Amit wrote:
>>=20
>> I was reading (again) the x86 C macro of =E2=80=9Calternative()=E2=80=9D=
 and I was a bit
>> surprised it does clobber the flags (=E2=80=9Ccc=E2=80=9D) as a =
precaution.
>>=20
>>  #define alternative(oldinstr, newinstr, ft_flags) \
>> 	asm_inline volatile (ALTERNATIVE(oldinstr, newinstr, ft_flags) : =
: : "memory")
>>=20
>> Actually there seems to be only one instance of problematic cases - =
in um/32-bit:
>>=20
>>  #define mb() alternative("lock; addl $0,0(%%esp)", "mfence", =
X86_FEATURE_XMM2)
>>  #define rmb() alternative("lock; addl $0,0(%%esp)", "lfence", =
X86_FEATURE_XMM2)
>>  #define wmb() alternative("lock; addl $0,0(%%esp)", "sfence", =
X86_FEATURE_XMM)
>>=20
>> Presumably, if XMM or XMM2 are not supported, there would be =
instances where addl
>> would be able to change eflags arithmetic flags without the compiler =
being aware
>> of it.
>>=20
>> As it only affects 32-bit Linux UM - I don=E2=80=99t easily have an =
environment to test
>> the fix. An alternative (word-pun unintended) is to add =E2=80=9Ccc=E2=80=
=9D as a precaution
>> to the alternative macro.
>>=20
> Application alternatives in um is presently a NOP. It always uses the =
"blunt and heavy instrument" - the most conservative option.
>=20
> It is on the TODO list.

Thanks for the quick response. But I don=E2=80=99t see how it prevents =
the problem
(it actually makes it worse - affecting XMM/XMM2 CPUs as well) since you
keep the =E2=80=9Clock; addl $0,0(%%esp)=E2=80=9D in the running code, =
affecting eflags
without telling the compiler that you do so through a =E2=80=9Ccc=E2=80=9D=
 clobber.