2022-01-07 23:27:52

by Kees Cook

[permalink] [raw]
Subject: [PATCH] arm64: atomics: Dereference matching size

When building with -Warray-bounds (which is desired to be enabled
globally), the following warning is generated:

In file included from ./arch/arm64/include/asm/lse.h:16,
from ./arch/arm64/include/asm/cmpxchg.h:14,
from ./arch/arm64/include/asm/atomic.h:16,
from ./include/linux/atomic.h:7,
from ./include/asm-generic/bitops/atomic.h:5,
from ./arch/arm64/include/asm/bitops.h:25,
from ./include/linux/bitops.h:33,
from ./include/linux/kernel.h:22,
from kernel/printk/printk.c:22:
./arch/arm64/include/asm/atomic_lse.h:247:9: warning: array subscript 'long unsigned int[0]' is partly outside array bounds of 'atomic_t[1]' [-Warray-bounds]
247 | asm volatile( \
| ^~~
./arch/arm64/include/asm/atomic_lse.h:266:1: note: in expansion of macro '__CMPXCHG_CASE'
266 | __CMPXCHG_CASE(w, , acq_, 32, a, "memory")
| ^~~~~~~~~~~~~~
kernel/printk/printk.c:3606:17: note: while referencing 'printk_cpulock_owner'
3606 | static atomic_t printk_cpulock_owner = ATOMIC_INIT(-1);
| ^~~~~~~~~~~~~~~~~~~~

This is due to the compiler seeing an unsigned long * cast against
something (atomic_t) that is int sized. Replace the cast with the
matching size cast. This results in no change in binary output.

Cc: Will Deacon <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Boqun Feng <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: [email protected]
Signed-off-by: Kees Cook <[email protected]>
---
arch/arm64/include/asm/atomic_lse.h | 2 +-
arch/arm64/include/asm/cmpxchg.h | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/atomic_lse.h b/arch/arm64/include/asm/atomic_lse.h
index d955ade5df7c..5d460f6b7675 100644
--- a/arch/arm64/include/asm/atomic_lse.h
+++ b/arch/arm64/include/asm/atomic_lse.h
@@ -249,7 +249,7 @@ __lse__cmpxchg_case_##name##sz(volatile void *ptr, \
" mov %" #w "[tmp], %" #w "[old]\n" \
" cas" #mb #sfx "\t%" #w "[tmp], %" #w "[new], %[v]\n" \
" mov %" #w "[ret], %" #w "[tmp]" \
- : [ret] "+r" (x0), [v] "+Q" (*(unsigned long *)ptr), \
+ : [ret] "+r" (x0), [v] "+Q" (*(u##sz *)ptr), \
[tmp] "=&r" (tmp) \
: [old] "r" (x1), [new] "r" (x2) \
: cl); \
diff --git a/arch/arm64/include/asm/cmpxchg.h b/arch/arm64/include/asm/cmpxchg.h
index f9bef42c1411..497acf134d99 100644
--- a/arch/arm64/include/asm/cmpxchg.h
+++ b/arch/arm64/include/asm/cmpxchg.h
@@ -243,7 +243,7 @@ static inline void __cmpwait_case_##sz(volatile void *ptr, \
" cbnz %" #w "[tmp], 1f\n" \
" wfe\n" \
"1:" \
- : [tmp] "=&r" (tmp), [v] "+Q" (*(unsigned long *)ptr) \
+ : [tmp] "=&r" (tmp), [v] "+Q" (*(u##sz *)ptr) \
: [val] "r" (val)); \
}

--
2.30.2



2022-01-08 09:32:53

by Ard Biesheuvel

[permalink] [raw]
Subject: Re: [PATCH] arm64: atomics: Dereference matching size

On Sat, 8 Jan 2022 at 00:28, Kees Cook <[email protected]> wrote:
>
> When building with -Warray-bounds (which is desired to be enabled
> globally), the following warning is generated:
>
> In file included from ./arch/arm64/include/asm/lse.h:16,
> from ./arch/arm64/include/asm/cmpxchg.h:14,
> from ./arch/arm64/include/asm/atomic.h:16,
> from ./include/linux/atomic.h:7,
> from ./include/asm-generic/bitops/atomic.h:5,
> from ./arch/arm64/include/asm/bitops.h:25,
> from ./include/linux/bitops.h:33,
> from ./include/linux/kernel.h:22,
> from kernel/printk/printk.c:22:
> ./arch/arm64/include/asm/atomic_lse.h:247:9: warning: array subscript 'long unsigned int[0]' is partly outside array bounds of 'atomic_t[1]' [-Warray-bounds]
> 247 | asm volatile( \
> | ^~~
> ./arch/arm64/include/asm/atomic_lse.h:266:1: note: in expansion of macro '__CMPXCHG_CASE'
> 266 | __CMPXCHG_CASE(w, , acq_, 32, a, "memory")
> | ^~~~~~~~~~~~~~
> kernel/printk/printk.c:3606:17: note: while referencing 'printk_cpulock_owner'
> 3606 | static atomic_t printk_cpulock_owner = ATOMIC_INIT(-1);
> | ^~~~~~~~~~~~~~~~~~~~
>
> This is due to the compiler seeing an unsigned long * cast against
> something (atomic_t) that is int sized. Replace the cast with the
> matching size cast. This results in no change in binary output.
>
> Cc: Will Deacon <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> Cc: Boqun Feng <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: [email protected]
> Signed-off-by: Kees Cook <[email protected]>

Acked-by: Ard Biesheuvel <[email protected]>

> ---
> arch/arm64/include/asm/atomic_lse.h | 2 +-
> arch/arm64/include/asm/cmpxchg.h | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/include/asm/atomic_lse.h b/arch/arm64/include/asm/atomic_lse.h
> index d955ade5df7c..5d460f6b7675 100644
> --- a/arch/arm64/include/asm/atomic_lse.h
> +++ b/arch/arm64/include/asm/atomic_lse.h
> @@ -249,7 +249,7 @@ __lse__cmpxchg_case_##name##sz(volatile void *ptr, \
> " mov %" #w "[tmp], %" #w "[old]\n" \
> " cas" #mb #sfx "\t%" #w "[tmp], %" #w "[new], %[v]\n" \
> " mov %" #w "[ret], %" #w "[tmp]" \
> - : [ret] "+r" (x0), [v] "+Q" (*(unsigned long *)ptr), \
> + : [ret] "+r" (x0), [v] "+Q" (*(u##sz *)ptr), \
> [tmp] "=&r" (tmp) \
> : [old] "r" (x1), [new] "r" (x2) \
> : cl); \
> diff --git a/arch/arm64/include/asm/cmpxchg.h b/arch/arm64/include/asm/cmpxchg.h
> index f9bef42c1411..497acf134d99 100644
> --- a/arch/arm64/include/asm/cmpxchg.h
> +++ b/arch/arm64/include/asm/cmpxchg.h
> @@ -243,7 +243,7 @@ static inline void __cmpwait_case_##sz(volatile void *ptr, \
> " cbnz %" #w "[tmp], 1f\n" \
> " wfe\n" \
> "1:" \
> - : [tmp] "=&r" (tmp), [v] "+Q" (*(unsigned long *)ptr) \
> + : [tmp] "=&r" (tmp), [v] "+Q" (*(u##sz *)ptr) \
> : [val] "r" (val)); \
> }
>
> --
> 2.30.2
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> [email protected]
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

2022-01-10 10:28:49

by Mark Rutland

[permalink] [raw]
Subject: Re: [PATCH] arm64: atomics: Dereference matching size

Hi Kees,

On Fri, Jan 07, 2022 at 03:27:46PM -0800, Kees Cook wrote:
> When building with -Warray-bounds (which is desired to be enabled
> globally), the following warning is generated:
>
> In file included from ./arch/arm64/include/asm/lse.h:16,
> from ./arch/arm64/include/asm/cmpxchg.h:14,
> from ./arch/arm64/include/asm/atomic.h:16,
> from ./include/linux/atomic.h:7,
> from ./include/asm-generic/bitops/atomic.h:5,
> from ./arch/arm64/include/asm/bitops.h:25,
> from ./include/linux/bitops.h:33,
> from ./include/linux/kernel.h:22,
> from kernel/printk/printk.c:22:
> ./arch/arm64/include/asm/atomic_lse.h:247:9: warning: array subscript 'long unsigned int[0]' is partly outside array bounds of 'atomic_t[1]' [-Warray-bounds]
> 247 | asm volatile( \
> | ^~~
> ./arch/arm64/include/asm/atomic_lse.h:266:1: note: in expansion of macro '__CMPXCHG_CASE'
> 266 | __CMPXCHG_CASE(w, , acq_, 32, a, "memory")
> | ^~~~~~~~~~~~~~
> kernel/printk/printk.c:3606:17: note: while referencing 'printk_cpulock_owner'
> 3606 | static atomic_t printk_cpulock_owner = ATOMIC_INIT(-1);
> | ^~~~~~~~~~~~~~~~~~~~
>
> This is due to the compiler seeing an unsigned long * cast against
> something (atomic_t) that is int sized. Replace the cast with the
> matching size cast. This results in no change in binary output.

Just to check, I assume both GCC and Clang are happy with this applied?

I recall that (historically at least) clang would warn about size mismatches
for inline assembly and would sometimes require more care. I don't see anythign
for which that would matter, but I just want to check.

> Cc: Will Deacon <[email protected]>
> Cc: Peter Zijlstra <[email protected]>
> Cc: Boqun Feng <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: [email protected]
> Signed-off-by: Kees Cook <[email protected]>
> ---
> arch/arm64/include/asm/atomic_lse.h | 2 +-
> arch/arm64/include/asm/cmpxchg.h | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/include/asm/atomic_lse.h b/arch/arm64/include/asm/atomic_lse.h
> index d955ade5df7c..5d460f6b7675 100644
> --- a/arch/arm64/include/asm/atomic_lse.h
> +++ b/arch/arm64/include/asm/atomic_lse.h
> @@ -249,7 +249,7 @@ __lse__cmpxchg_case_##name##sz(volatile void *ptr, \
> " mov %" #w "[tmp], %" #w "[old]\n" \
> " cas" #mb #sfx "\t%" #w "[tmp], %" #w "[new], %[v]\n" \
> " mov %" #w "[ret], %" #w "[tmp]" \
> - : [ret] "+r" (x0), [v] "+Q" (*(unsigned long *)ptr), \
> + : [ret] "+r" (x0), [v] "+Q" (*(u##sz *)ptr), \
> [tmp] "=&r" (tmp) \
> : [old] "r" (x1), [new] "r" (x2) \
> : cl); \

It might be worth nothing that __ll_sc__cmpxchg_case_##name##sz already uses
the same constraint:

[v] "+Q" (*(u##sz *)ptr

... since that explains why we only need to update the LSE form and not the
LL/SC form, and indicates that this is unlikely to be problematic.

Either way:

Acked-by: Mark Rutland <[email protected]>

Thanks,
Mark.

> diff --git a/arch/arm64/include/asm/cmpxchg.h b/arch/arm64/include/asm/cmpxchg.h
> index f9bef42c1411..497acf134d99 100644
> --- a/arch/arm64/include/asm/cmpxchg.h
> +++ b/arch/arm64/include/asm/cmpxchg.h
> @@ -243,7 +243,7 @@ static inline void __cmpwait_case_##sz(volatile void *ptr, \
> " cbnz %" #w "[tmp], 1f\n" \
> " wfe\n" \
> "1:" \
> - : [tmp] "=&r" (tmp), [v] "+Q" (*(unsigned long *)ptr) \
> + : [tmp] "=&r" (tmp), [v] "+Q" (*(u##sz *)ptr) \
> : [val] "r" (val)); \
> }
>
> --
> 2.30.2
>

2022-01-12 20:19:16

by Kees Cook

[permalink] [raw]
Subject: Re: [PATCH] arm64: atomics: Dereference matching size

On Mon, Jan 10, 2022 at 10:27:59AM +0000, Mark Rutland wrote:
> Hi Kees,
>
> On Fri, Jan 07, 2022 at 03:27:46PM -0800, Kees Cook wrote:
> > When building with -Warray-bounds (which is desired to be enabled
> > globally), the following warning is generated:
> >
> > In file included from ./arch/arm64/include/asm/lse.h:16,
> > from ./arch/arm64/include/asm/cmpxchg.h:14,
> > from ./arch/arm64/include/asm/atomic.h:16,
> > from ./include/linux/atomic.h:7,
> > from ./include/asm-generic/bitops/atomic.h:5,
> > from ./arch/arm64/include/asm/bitops.h:25,
> > from ./include/linux/bitops.h:33,
> > from ./include/linux/kernel.h:22,
> > from kernel/printk/printk.c:22:
> > ./arch/arm64/include/asm/atomic_lse.h:247:9: warning: array subscript 'long unsigned int[0]' is partly outside array bounds of 'atomic_t[1]' [-Warray-bounds]
> > 247 | asm volatile( \
> > | ^~~
> > ./arch/arm64/include/asm/atomic_lse.h:266:1: note: in expansion of macro '__CMPXCHG_CASE'
> > 266 | __CMPXCHG_CASE(w, , acq_, 32, a, "memory")
> > | ^~~~~~~~~~~~~~
> > kernel/printk/printk.c:3606:17: note: while referencing 'printk_cpulock_owner'
> > 3606 | static atomic_t printk_cpulock_owner = ATOMIC_INIT(-1);
> > | ^~~~~~~~~~~~~~~~~~~~
> >
> > This is due to the compiler seeing an unsigned long * cast against
> > something (atomic_t) that is int sized. Replace the cast with the
> > matching size cast. This results in no change in binary output.
>
> Just to check, I assume both GCC and Clang are happy with this applied?
>
> I recall that (historically at least) clang would warn about size mismatches
> for inline assembly and would sometimes require more care. I don't see anythign
> for which that would matter, but I just want to check.

Yup, Clang is happy with this AFAICT.

>
> > Cc: Will Deacon <[email protected]>
> > Cc: Peter Zijlstra <[email protected]>
> > Cc: Boqun Feng <[email protected]>
> > Cc: Catalin Marinas <[email protected]>
> > Cc: [email protected]
> > Signed-off-by: Kees Cook <[email protected]>
> > ---
> > arch/arm64/include/asm/atomic_lse.h | 2 +-
> > arch/arm64/include/asm/cmpxchg.h | 2 +-
> > 2 files changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/arm64/include/asm/atomic_lse.h b/arch/arm64/include/asm/atomic_lse.h
> > index d955ade5df7c..5d460f6b7675 100644
> > --- a/arch/arm64/include/asm/atomic_lse.h
> > +++ b/arch/arm64/include/asm/atomic_lse.h
> > @@ -249,7 +249,7 @@ __lse__cmpxchg_case_##name##sz(volatile void *ptr, \
> > " mov %" #w "[tmp], %" #w "[old]\n" \
> > " cas" #mb #sfx "\t%" #w "[tmp], %" #w "[new], %[v]\n" \
> > " mov %" #w "[ret], %" #w "[tmp]" \
> > - : [ret] "+r" (x0), [v] "+Q" (*(unsigned long *)ptr), \
> > + : [ret] "+r" (x0), [v] "+Q" (*(u##sz *)ptr), \
> > [tmp] "=&r" (tmp) \
> > : [old] "r" (x1), [new] "r" (x2) \
> > : cl); \
>
> It might be worth nothing that __ll_sc__cmpxchg_case_##name##sz already uses
> the same constraint:
>
> [v] "+Q" (*(u##sz *)ptr
>
> ... since that explains why we only need to update the LSE form and not the
> LL/SC form, and indicates that this is unlikely to be problematic.

Good idea, I'll note that in a v2.

>
> Either way:
>
> Acked-by: Mark Rutland <[email protected]>

Thanks!

-Kees

>
> Thanks,
> Mark.
>
> > diff --git a/arch/arm64/include/asm/cmpxchg.h b/arch/arm64/include/asm/cmpxchg.h
> > index f9bef42c1411..497acf134d99 100644
> > --- a/arch/arm64/include/asm/cmpxchg.h
> > +++ b/arch/arm64/include/asm/cmpxchg.h
> > @@ -243,7 +243,7 @@ static inline void __cmpwait_case_##sz(volatile void *ptr, \
> > " cbnz %" #w "[tmp], 1f\n" \
> > " wfe\n" \
> > "1:" \
> > - : [tmp] "=&r" (tmp), [v] "+Q" (*(unsigned long *)ptr) \
> > + : [tmp] "=&r" (tmp), [v] "+Q" (*(u##sz *)ptr) \
> > : [val] "r" (val)); \
> > }
> >
> > --
> > 2.30.2
> >

--
Kees Cook