Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   YingChi Long <me@inclyc.cn>
To:     me@inclyc.cn
Cc:     bp@alien8.de, chang.seok.bae@intel.com,
        dave.hansen@linux.intel.com, hpa@zytor.com,
        linux-kernel@vger.kernel.org, mingo@redhat.com,
        ndesaulniers@google.com, pbonzini@redhat.com, tglx@linutronix.de,
        x86@kernel.org, david.laight@aculab.com
Subject: [PATCH v3] x86/fpu: use _Alignof to avoid UB in TYPE_ALIGN
Date:   Thu,  6 Oct 2022 22:14:42 +0800
Message-Id: <20221006141442.2475978-1-me@inclyc.cn>
In-Reply-To: <20220925153151.2467884-1-me@inclyc.cn>
References: <20220925153151.2467884-1-me@inclyc.cn>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

WG14 N2350 made very clear that it is an UB having type definitions with
in "offsetof". This patch change the implementation of macro
"TYPE_ALIGN" to builtin "_Alignof" to avoid undefined behavior.

I've grepped all source files to find any type definitions within
"offsetof".

    offsetof\(struct .*\{ .*,

This implementation of macro "TYPE_ALIGN" seemes to be the only case of
type definitions within offsetof in the kernel codebase.

I've made a clang patch that rejects any definitions within
__builtin_offsetof (usually #defined with "offsetof"), and tested
compiling with this patch, there are no error if this patch applied.

ISO C11 _Alignof is subtly different from the GNU C extension
__alignof__. __alignof__ is the preferred alignment and _Alignof the
minimal alignment. For 'long long' on x86 these are 8 and 4
respectively.

The macro TYPE_ALIGN we're replacing has behavior that matches
_Alignof rather than __alignof__.

Signed-off-by: YingChi Long <me@inclyc.cn>
Link: https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2350.htm
Link: https://godbolt.org/z/sPs1GEhbT
Link: https://gcc.gnu.org/onlinedocs/gcc/Alignment.html
Link: https://reviews.llvm.org/D133574
---
v3:
- commit message changes suggested by Nick and David

v2: https://lore.kernel.org/all/20220927153338.4177854-1-me@inclyc.cn/
---
 arch/x86/kernel/fpu/init.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kernel/fpu/init.c b/arch/x86/kernel/fpu/init.c
index 621f4b6cac4a..de96c11e1fe9 100644
--- a/arch/x86/kernel/fpu/init.c
+++ b/arch/x86/kernel/fpu/init.c
@@ -133,9 +133,6 @@ static void __init fpu__init_system_generic(void)
 	fpu__init_system_mxcsr();
 }

-/* Get alignment of the TYPE. */
-#define TYPE_ALIGN(TYPE) offsetof(struct { char x; TYPE test; }, test)
-
 /*
  * Enforce that 'MEMBER' is the last field of 'TYPE'.
  *
@@ -143,8 +140,8 @@ static void __init fpu__init_system_generic(void)
  * because that's how C aligns structs.
  */
 #define CHECK_MEMBER_AT_END_OF(TYPE, MEMBER) \
-	BUILD_BUG_ON(sizeof(TYPE) != ALIGN(offsetofend(TYPE, MEMBER), \
-					   TYPE_ALIGN(TYPE)))
+	BUILD_BUG_ON(sizeof(TYPE) !=         \
+		     ALIGN(offsetofend(TYPE, MEMBER), _Alignof(TYPE)))

 /*
  * We append the 'struct fpu' to the task_struct:
--
2.35.1