diff mbox series

[2/2,AArch32] Fix 128-bit sequential consistency atomic operations.

Message ID YqC3E6H+wO35o+VR@arm.com
State Committed
Commit 5471f55f001af412e1125b04972ebaab9d4f7337
Headers
Series [1/2] AArch64 Fix 128-bit sequential consistency atomic operations. |

Commit Message

Tamar Christina June 8, 2022, 2:49 p.m. UTC 
  Hi All,

Similar to AArch64 the Arm implementation of 128-bit atomics is broken.

For 128-bit atomics we rely on pthread barriers to correct guard the address
in the pointer to get correct memory ordering.  However for 128-bit atomics the
address under the lock is different from the original pointer.

This means that one of the values under the atomic operation is not protected
properly and so we fail during when the user has requested sequential
consistency as there's no barrier to enforce this requirement.

As such users have resorted to adding an

#ifdef GCC
<emit barrier>
#endif

around the use of these atomics.

This corrects the issue by issuing a barrier only when __ATOMIC_SEQ_CST was
requested.  I have hand verified that the barriers are inserted
for atomic seq cst.


Bootstrapped Regtested on arm-none-linux-gnueabihf and no issues.

Ok for master? and for backporting to GCC 12, 11 and 10?

Thanks,
Tamar

libatomic/ChangeLog:

	PR target/102218
	* config/arm/host-config.h (pre_seq_barrier, post_seq_barrier,
	pre_post_seq_barrier): Require barrier on __ATOMIC_SEQ_CST.

--- inline copy of patch -- 
diff --git a/libatomic/config/arm/host-config.h b/libatomic/config/arm/host-config.h
index bbf4a3f84c3f3ae21fb2162aab68bdedf3fbdcb4..ef16fad2a35ec9055e918849e69a1a0e23b62838 100644




--
diff --git a/libatomic/config/arm/host-config.h b/libatomic/config/arm/host-config.h
index bbf4a3f84c3f3ae21fb2162aab68bdedf3fbdcb4..ef16fad2a35ec9055e918849e69a1a0e23b62838 100644
--- a/libatomic/config/arm/host-config.h
+++ b/libatomic/config/arm/host-config.h
@@ -1,4 +1,23 @@
 /* Avoiding the DMB (or kernel helper) can be a good thing.  */
 #define WANT_SPECIALCASE_RELAXED
 
+/* Glibc, at least, uses acq_rel in its pthread mutex
+   implementation.  If the user is asking for seq_cst,
+   this is insufficient.  */
+
+static inline void __attribute__((always_inline, artificial))
+pre_seq_barrier(int model)
+{
+  if (model == __ATOMIC_SEQ_CST)
+    __atomic_thread_fence (__ATOMIC_SEQ_CST);
+}
+
+static inline void __attribute__((always_inline, artificial))
+post_seq_barrier(int model)
+{
+  pre_seq_barrier(model);
+}
+
+#define pre_post_seq_barrier 1
+
 #include_next <host-config.h>

Comments

Tamar Christina June 16, 2022, 9:14 a.m. UTC | #1 
ping

> -----Original Message-----
> From: Tamar Christina <tamar.christina@arm.com>
> Sent: Wednesday, June 8, 2022 3:50 PM
> To: gcc-patches@gcc.gnu.org
> Cc: nd <nd@arm.com>; Ramana Radhakrishnan
> <Ramana.Radhakrishnan@arm.com>; Richard Earnshaw
> <Richard.Earnshaw@arm.com>; nickc@redhat.com; Kyrylo Tkachov
> <Kyrylo.Tkachov@arm.com>
> Subject: [PATCH 2/2][AArch32] Fix 128-bit sequential consistency atomic
> operations.
> 
> Hi All,
> 
> Similar to AArch64 the Arm implementation of 128-bit atomics is broken.
> 
> For 128-bit atomics we rely on pthread barriers to correct guard the address
> in the pointer to get correct memory ordering.  However for 128-bit atomics
> the address under the lock is different from the original pointer.
> 
> This means that one of the values under the atomic operation is not
> protected properly and so we fail during when the user has requested
> sequential consistency as there's no barrier to enforce this requirement.
> 
> As such users have resorted to adding an
> 
> #ifdef GCC
> <emit barrier>
> #endif
> 
> around the use of these atomics.
> 
> This corrects the issue by issuing a barrier only when __ATOMIC_SEQ_CST
> was requested.  I have hand verified that the barriers are inserted for atomic
> seq cst.
> 
> 
> Bootstrapped Regtested on arm-none-linux-gnueabihf and no issues.
> 
> Ok for master? and for backporting to GCC 12, 11 and 10?
> 
> Thanks,
> Tamar
> 
> libatomic/ChangeLog:
> 
> 	PR target/102218
> 	* config/arm/host-config.h (pre_seq_barrier, post_seq_barrier,
> 	pre_post_seq_barrier): Require barrier on __ATOMIC_SEQ_CST.
> 
> --- inline copy of patch --
> diff --git a/libatomic/config/arm/host-config.h b/libatomic/config/arm/host-
> config.h
> index
> bbf4a3f84c3f3ae21fb2162aab68bdedf3fbdcb4..ef16fad2a35ec9055e918849e6
> 9a1a0e23b62838 100644
> --- a/libatomic/config/arm/host-config.h
> +++ b/libatomic/config/arm/host-config.h
> @@ -1,4 +1,23 @@
>  /* Avoiding the DMB (or kernel helper) can be a good thing.  */  #define
> WANT_SPECIALCASE_RELAXED
> 
> +/* Glibc, at least, uses acq_rel in its pthread mutex
> +   implementation.  If the user is asking for seq_cst,
> +   this is insufficient.  */
> +
> +static inline void __attribute__((always_inline, artificial))
> +pre_seq_barrier(int model) {
> +  if (model == __ATOMIC_SEQ_CST)
> +    __atomic_thread_fence (__ATOMIC_SEQ_CST); }
> +
> +static inline void __attribute__((always_inline, artificial))
> +post_seq_barrier(int model) {
> +  pre_seq_barrier(model);
> +}
> +
> +#define pre_post_seq_barrier 1
> +
>  #include_next <host-config.h>
> 
> 
> 
> 
> --
Kyrylo Tkachov Aug. 8, 2022, 12:53 p.m. UTC | #2 
> -----Original Message-----
> From: Tamar Christina <Tamar.Christina@arm.com>
> Sent: Wednesday, June 8, 2022 3:50 PM
> To: gcc-patches@gcc.gnu.org
> Cc: nd <nd@arm.com>; Ramana Radhakrishnan
> <Ramana.Radhakrishnan@arm.com>; Richard Earnshaw
> <Richard.Earnshaw@arm.com>; nickc@redhat.com; Kyrylo Tkachov
> <Kyrylo.Tkachov@arm.com>
> Subject: [PATCH 2/2][AArch32] Fix 128-bit sequential consistency atomic
> operations.
> 
> Hi All,
> 
> Similar to AArch64 the Arm implementation of 128-bit atomics is broken.
> 
> For 128-bit atomics we rely on pthread barriers to correct guard the address
> in the pointer to get correct memory ordering.  However for 128-bit atomics
> the
> address under the lock is different from the original pointer.
> 
> This means that one of the values under the atomic operation is not
> protected
> properly and so we fail during when the user has requested sequential
> consistency as there's no barrier to enforce this requirement.
> 
> As such users have resorted to adding an
> 
> #ifdef GCC
> <emit barrier>
> #endif
> 
> around the use of these atomics.
> 
> This corrects the issue by issuing a barrier only when __ATOMIC_SEQ_CST
> was
> requested.  I have hand verified that the barriers are inserted
> for atomic seq cst.
> 
> 
> Bootstrapped Regtested on arm-none-linux-gnueabihf and no issues.
> 
> Ok for master? and for backporting to GCC 12, 11 and 10?

Ok, with backports after a couple weeks on master.
Thanks,
Kyrill

> 
> Thanks,
> Tamar
> 
> libatomic/ChangeLog:
> 
> 	PR target/102218
> 	* config/arm/host-config.h (pre_seq_barrier, post_seq_barrier,
> 	pre_post_seq_barrier): Require barrier on __ATOMIC_SEQ_CST.
> 
> --- inline copy of patch --
> diff --git a/libatomic/config/arm/host-config.h b/libatomic/config/arm/host-
> config.h
> index
> bbf4a3f84c3f3ae21fb2162aab68bdedf3fbdcb4..ef16fad2a35ec9055e918849e
> 69a1a0e23b62838 100644
> --- a/libatomic/config/arm/host-config.h
> +++ b/libatomic/config/arm/host-config.h
> @@ -1,4 +1,23 @@
>  /* Avoiding the DMB (or kernel helper) can be a good thing.  */
>  #define WANT_SPECIALCASE_RELAXED
> 
> +/* Glibc, at least, uses acq_rel in its pthread mutex
> +   implementation.  If the user is asking for seq_cst,
> +   this is insufficient.  */
> +
> +static inline void __attribute__((always_inline, artificial))
> +pre_seq_barrier(int model)
> +{
> +  if (model == __ATOMIC_SEQ_CST)
> +    __atomic_thread_fence (__ATOMIC_SEQ_CST);
> +}
> +
> +static inline void __attribute__((always_inline, artificial))
> +post_seq_barrier(int model)
> +{
> +  pre_seq_barrier(model);
> +}
> +
> +#define pre_post_seq_barrier 1
> +
>  #include_next <host-config.h>
> 
> 
> 
> 
> --
diff mbox series

Patch

--- a/libatomic/config/arm/host-config.h
+++ b/libatomic/config/arm/host-config.h
@@ -1,4 +1,23 @@ 
 /* Avoiding the DMB (or kernel helper) can be a good thing.  */
 #define WANT_SPECIALCASE_RELAXED
 
+/* Glibc, at least, uses acq_rel in its pthread mutex
+   implementation.  If the user is asking for seq_cst,
+   this is insufficient.  */
+
+static inline void __attribute__((always_inline, artificial))
+pre_seq_barrier(int model)
+{
+  if (model == __ATOMIC_SEQ_CST)
+    __atomic_thread_fence (__ATOMIC_SEQ_CST);
+}
+
+static inline void __attribute__((always_inline, artificial))
+post_seq_barrier(int model)
+{
+  pre_seq_barrier(model);
+}
+
+#define pre_post_seq_barrier 1
+
 #include_next <host-config.h>