2024-04-24 19:18:42

by Kees Cook

Subject: [PATCH 2/4] arm64: atomics: lse: Silence intentional wrapping addition

Annotate atomic_add_return() and atomic_sub_return() to avoid signed
overflow instrumentation. They are expected to wrap around.

Signed-off-by: Kees Cook <[email protected]>
---
Cc: Will Deacon <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Boqun Feng <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: [email protected]
---
arch/arm64/include/asm/atomic_lse.h | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/arch/arm64/include/asm/atomic_lse.h b/arch/arm64/include/asm/atomic_lse.h
index 87f568a94e55..a33576b20b52 100644
--- a/arch/arm64/include/asm/atomic_lse.h
+++ b/arch/arm64/include/asm/atomic_lse.h
@@ -10,6 +10,8 @@
#ifndef __ASM_ATOMIC_LSE_H
#define __ASM_ATOMIC_LSE_H

+#include <linux/overflow.h>
+
#define ATOMIC_OP(op, asm_op) \
static __always_inline void \
__lse_atomic_##op(int i, atomic_t *v) \
@@ -82,13 +84,13 @@ ATOMIC_FETCH_OP_SUB( )
static __always_inline int \
__lse_atomic_add_return##name(int i, atomic_t *v) \
{ \
- return __lse_atomic_fetch_add##name(i, v) + i; \
+ return wrapping_add(int, __lse_atomic_fetch_add##name(i, v), i);\
} \
\
static __always_inline int \
__lse_atomic_sub_return##name(int i, atomic_t *v) \
{ \
- return __lse_atomic_fetch_sub(i, v) - i; \
+ return wrapping_sub(int, __lse_atomic_fetch_sub(i, v), i); \
}

ATOMIC_OP_ADD_SUB_RETURN(_relaxed)
@@ -189,13 +191,13 @@ ATOMIC64_FETCH_OP_SUB( )
static __always_inline long \
__lse_atomic64_add_return##name(s64 i, atomic64_t *v) \
{ \
- return __lse_atomic64_fetch_add##name(i, v) + i; \
+ return wrapping_add(s64, __lse_atomic64_fetch_add##name(i, v), i); \
} \
\
static __always_inline long \
__lse_atomic64_sub_return##name(s64 i, atomic64_t *v) \
{ \
- return __lse_atomic64_fetch_sub##name(i, v) - i; \
+ return wrapping_sub(s64, __lse_atomic64_fetch_sub##name(i, v), i); \
}

ATOMIC64_OP_ADD_SUB_RETURN(_relaxed)
--
2.34.1

2024-05-02 11:21:47

by Will Deacon

Subject: Re: [PATCH 2/4] arm64: atomics: lse: Silence intentional wrapping addition

On Wed, Apr 24, 2024 at 12:17:35PM -0700, Kees Cook wrote:
> Annotate atomic_add_return() and atomic_sub_return() to avoid signed
> overflow instrumentation. They are expected to wrap around.
>
> Signed-off-by: Kees Cook <[email protected]>
> ---
> Cc: Will Deacon <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> Cc: Boqun Feng <[email protected]>
> Cc: Mark Rutland <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: [email protected]
> ---
> arch/arm64/include/asm/atomic_lse.h | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)

How come the ll/sc routines (in atomic_ll_sc.h) don't need the same
treatment? If that's just an oversight, then maybe it's better to
instrument the higher-level wrappers in asm/atomic.h?

Will

2024-05-02 15:01:07

by Kees Cook

Subject: Re: [PATCH 2/4] arm64: atomics: lse: Silence intentional wrapping addition

On Thu, May 02, 2024 at 12:21:28PM +0100, Will Deacon wrote:
> On Wed, Apr 24, 2024 at 12:17:35PM -0700, Kees Cook wrote:
> > Annotate atomic_add_return() and atomic_sub_return() to avoid signed
> > overflow instrumentation. They are expected to wrap around.
> >
> > Signed-off-by: Kees Cook <[email protected]>
> > ---
> > Cc: Will Deacon <[email protected]>
> > Cc: Peter Zijlstra <[email protected]>
> > Cc: Boqun Feng <[email protected]>
> > Cc: Mark Rutland <[email protected]>
> > Cc: Catalin Marinas <[email protected]>
> > Cc: [email protected]
> > ---
> > arch/arm64/include/asm/atomic_lse.h | 10 ++++++----
> > 1 file changed, 6 insertions(+), 4 deletions(-)
>
> How come the ll/sc routines (in atomic_ll_sc.h) don't need the same
> treatment? If that's just an oversight, then maybe it's better to
> instrument the higher-level wrappers in asm/atomic.h?

Those are all written in asm, so there's no open-coded C arithmetic that
the sanitizers will notice. All is well there! :)

--
Kees Cook