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, 15 Oct 2020 14:34:57 -0700
User-Agent: K-9 Mail for Android
In-Reply-To: <20201015161216.1563600-1-irogers@google.com>
References: <20201015161216.1563600-1-irogers@google.com>
MIME-Version: 1.0
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable
Subject: Re: [PATCH v2] x86/insn, tools/x86: Fix some potential undefined behavior.
To:     Ian Rogers <irogers@google.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        x86@kernel.org, "Peter Zijlstra (Intel)" <peterz@infradead.org>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        Josh Poimboeuf <jpoimboe@redhat.com>,
        linux-kernel@vger.kernel.org,
        Adrian Hunter <adrian.hunter@intel.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>
CC:     Numfor Mbiziwo-Tiapo <nums@google.com>
From:   hpa@zytor.com
Message-ID: <5E711F64-A9BF-48EC-83A3-3C644D80F848@zytor.com>
Precedence: bulk

On October 15, 2020 9:12:16 AM PDT, Ian Rogers <irogers@google=2Ecom> wrote=
:
>From: Numfor Mbiziwo-Tiapo <nums@google=2Ecom>
>
>Don't perform unaligned loads in __get_next and __peek_nbyte_next as
>these are forms of undefined behavior=2E
>
>These problems were identified using the undefined behavior sanitizer
>(ubsan) with the tools version of the code and perf test=2E Part of this
>patch was previously posted here:
>https://lore=2Ekernel=2Eorg/lkml/20190724184512=2E162887-4-nums@google=2E=
com/
>
>v2=2E removes the validate_next check and merges the 2 changes into one
>as
>requested by Masami Hiramatsu <mhiramat@kernel=2Eorg>
>
>Signed-off-by: Ian Rogers <irogers@google=2Ecom>
>Signed-off-by: Numfor Mbiziwo-Tiapo <nums@google=2Ecom>
>---
> arch/x86/lib/insn=2Ec       | 4 ++--
> tools/arch/x86/lib/insn=2Ec | 4 ++--
> 2 files changed, 4 insertions(+), 4 deletions(-)
>
>diff --git a/arch/x86/lib/insn=2Ec b/arch/x86/lib/insn=2Ec
>index 404279563891=2E=2Ebe88ab250146 100644
>--- a/arch/x86/lib/insn=2Ec
>+++ b/arch/x86/lib/insn=2Ec
>@@ -20,10 +20,10 @@
> 	((insn)->next_byte + sizeof(t) + n <=3D (insn)->end_kaddr)
>=20
> #define __get_next(t, insn)	\
>-	({ t r =3D *(t*)insn->next_byte; insn->next_byte +=3D sizeof(t); r; })
>+	({ t r; memcpy(&r, insn->next_byte, sizeof(t)); insn->next_byte +=3D
>sizeof(t); r; })
>=20
> #define __peek_nbyte_next(t, insn, n)	\
>-	({ t r =3D *(t*)((insn)->next_byte + n); r; })
>+	({ t r; memcpy(&r, (insn)->next_byte + n, sizeof(t)); r; })
>=20
> #define get_next(t, insn)	\
>	({ if (unlikely(!validate_next(t, insn, 0))) goto err_out;
>__get_next(t, insn); })
>diff --git a/tools/arch/x86/lib/insn=2Ec b/tools/arch/x86/lib/insn=2Ec
>index 0151dfc6da61=2E=2E92358c71a59e 100644
>--- a/tools/arch/x86/lib/insn=2Ec
>+++ b/tools/arch/x86/lib/insn=2Ec
>@@ -20,10 +20,10 @@
> 	((insn)->next_byte + sizeof(t) + n <=3D (insn)->end_kaddr)
>=20
> #define __get_next(t, insn)	\
>-	({ t r =3D *(t*)insn->next_byte; insn->next_byte +=3D sizeof(t); r; })
>+	({ t r; memcpy(&r, insn->next_byte, sizeof(t)); insn->next_byte +=3D
>sizeof(t); r; })
>=20
> #define __peek_nbyte_next(t, insn, n)	\
>-	({ t r =3D *(t*)((insn)->next_byte + n); r; })
>+	({ t r; memcpy(&r, (insn)->next_byte + n, sizeof(t)); r; })
>=20
> #define get_next(t, insn)	\
>	({ if (unlikely(!validate_next(t, insn, 0))) goto err_out;
>__get_next(t, insn); })

Wait, what?

You are taking about x86-specific code, and on x86 unaligned memory access=
es are supported, well-defined, and ubiquitous=2E=20

This is B=2ES=2E at best, and unless the compiler turns the memcpy() right=
 back into what you started with, deleterious for performance=2E

If you have a *very* good reason for this kind of churn, wrap it in the un=
aligned access macros, but using memcpy() is insane=2E All you are doing is=
 making the code worse=2E

--=20
Sent from my Android device with K-9 Mail=2E Please excuse my brevity=2E