MIME-Version: 1.0
References: <20230125012801.362496-1-ebiggers@kernel.org> <14506678-918f-81e1-2c26-2b347ff50701@intel.com>
 <CAG48ez1NaWarARJj5SBdKKTYFO2MbX7xO75Rk0Q2iK8LX4BwFA@mail.gmail.com> <394c92e2-a9aa-37e1-7a34-d7569ac844fd@intel.com>
In-Reply-To: <394c92e2-a9aa-37e1-7a34-d7569ac844fd@intel.com>
From:   Jann Horn <jannh@google.com>
Date:   Thu, 26 Jan 2023 18:52:25 +0100
Message-ID: <CAG48ez0ZK3pMqkto4DTZPNyddYcv8jPHQDNhYoFEPvSRLf80fQ@mail.gmail.com>
Subject: Re: [PATCH] x86: enable Data Operand Independent Timing Mode
To:     Dave Hansen <dave.hansen@intel.com>
Cc:     Eric Biggers <ebiggers@kernel.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        "H . Peter Anvin" <hpa@zytor.com>, x86@kernel.org,
        linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-hardening@vger.kernel.org,
        Peter Zijlstra <peterz@infradead.org>,
        Roxana Bradescu <roxabee@chromium.org>,
        Adam Langley <agl@google.com>,
        Ard Biesheuvel <ardb@kernel.org>,
        "Jason A . Donenfeld" <Jason@zx2c4.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Jan 26, 2023 at 5:40 PM Dave Hansen <dave.hansen@intel.com> wrote:
> On 1/26/23 05:52, Jann Horn wrote:
> > On Wed, Jan 25, 2023 at 4:30 PM Dave Hansen <dave.hansen@intel.com> wrote:
> >> Translating from Intel-speak: Intel thinks that DOITM purely a way to
> >> make the CPU run slower if you haven't already written code specifically
> >> to mitigate timing side channels.  All pain, no gain.
> >>
> >> The kernel as a whole is not written that way.
> >
> > The kernel as a whole also doesn't really use the FPU registers for
> > anything other than checksumming and cryptography and stuff like that
> > (it's disabled in the compiler flags because the FPU registers
> > normally contain userspace state that must not be clobbered). The
> > instructions listed on that Intel help page are all weird PM* and VP*
> > arithmetic instructions that can't be generated from C code in the
> > kernel (except for weird subsystems in which every function is only
> > callable in kernel-FPU-enabled mode and the compiler is set to enable
> > FPU instruction generation, by which I mean amdgpu).
>
> Maybe I'm totally missing something, but I thought the scope here was
> the "non-data operand independent timing behavior for the listed
> instructions" referenced here:
>
> > https://www.intel.com/content/www/us/en/developer/articles/technical/software-security-guidance/best-practices/data-operand-independent-timing-isa-guidance.html
>
> where the "listed instructions" is this list:
>
> > https://www.intel.com/content/www/us/en/developer/articles/technical/software-security-guidance/resources/data-operand-independent-timing-instructions.html
>
> For example, that includes XOR with the 0x31 and 0x81 opcodes which
> there are plenty of in the kernel.

That list says at the top: "The table below lists instructions that
have data-independent timing."

And the "MCDT (MXCSR-Sensitivity)" column that marks instructions that
do not actually have data-independent timing if you set the MSR to the
unsafe state is only marked for PMADDUBSW, PMADDWD, PMULDQ, PMULHRSW,
PMULHUW, PMULHW, PMULLD, PMULLW, PMULUDQ, VPLZCNTD, VPLZCNTQ,
VPMADD52HUQ, VPMADD52LUQ, VPMADDUBSW, VPMADDWD, VPMULDQ, VPMULHRSW,
VPMULHUW, VPMULHW, VPMULLD, VPMULLQ, VPMULLW, VPMULUDQ. All the others
are guaranteed to always have data-independent timing, if I understand
the table correctly.

> That's a bit wider scope than the crazy instructions like VPLZCNTD.  The
> crazy instructions list that I _think_ you were grepping for is the
> "Instructions That May Exhibit MCDT Behavior".  That's also a fun one,
> but it is more narrow than the DOITM list.