X-Git-Url: http://git.lttng.org/?p=urcu.git;a=blobdiff_plain;f=include%2Furcu%2Fstatic%2Fpointer.h;h=891daed2994ba8a58c0b7f1bdfe10d886d37f91c;hp=732f8c078d7dcf585bde531cdacd51154818a211;hb=HEAD;hpb=7afcc7685532e13f39fc13d505c96a3e5581d4d3

diff --git a/include/urcu/static/pointer.h b/include/urcu/static/pointer.h
index 732f8c0..9dc0d3e 100644
--- a/include/urcu/static/pointer.h
+++ b/include/urcu/static/pointer.h
@@ -1,31 +1,17 @@
+// SPDX-FileCopyrightText: 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+// SPDX-FileCopyrightText: 2009 Paul E. McKenney, IBM Corporation.
+//
+// SPDX-License-Identifier: LGPL-2.1-or-later
+
 #ifndef _URCU_POINTER_STATIC_H
 #define _URCU_POINTER_STATIC_H
 
 /*
- * urcu/static/pointer.h
- *
  * Userspace RCU header. Operations on pointers.
  *
  * TO BE INCLUDED ONLY IN CODE THAT IS TO BE RECOMPILED ON EACH LIBURCU
  * RELEASE. See urcu.h for linking dynamically with the userspace rcu library.
  *
- * Copyright (c) 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
- * Copyright (c) 2009 Paul E. McKenney, IBM Corporation.
- *
- * This library is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * This library is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- *
  * IBM's contributions to this file may be relicensed under LGPLv2 or later.
  */
 
@@ -49,14 +35,33 @@ extern "C" {
  * Inserts memory barriers on architectures that require them (currently only
  * Alpha) and documents which pointers are protected by RCU.
  *
- * The compiler memory barrier in CMM_LOAD_SHARED() ensures that value-speculative
- * optimizations (e.g. VSS: Value Speculation Scheduling) does not perform the
- * data read before the pointer read by speculating the value of the pointer.
- * Correct ordering is ensured because the pointer is read as a volatile access.
- * This acts as a global side-effect operation, which forbids reordering of
- * dependent memory operations. Note that such concern about dependency-breaking
- * optimizations will eventually be taken care of by the "memory_order_consume"
- * addition to forthcoming C++ standard.
+ * With C standards prior to C11/C++11, the compiler memory barrier in
+ * CMM_LOAD_SHARED() ensures that value-speculative optimizations (e.g.
+ * VSS: Value Speculation Scheduling) does not perform the data read
+ * before the pointer read by speculating the value of the pointer.
+ * Correct ordering is ensured because the pointer is read as a volatile
+ * access. This acts as a global side-effect operation, which forbids
+ * reordering of dependent memory operations.
+ *
+ * With C standards C11/C++11, concerns about dependency-breaking
+ * optimizations are taken care of by the "memory_order_consume" atomic
+ * load.
+ *
+ * Use the gcc __atomic_load() rather than C11/C++11 atomic load
+ * explicit because the pointer used as input argument is a pointer,
+ * not an _Atomic type as required by C11/C++11.
+ *
+ * By defining URCU_DEREFERENCE_USE_VOLATILE, the user requires use of
+ * volatile access to implement rcu_dereference rather than
+ * memory_order_consume load from the C11/C++11 standards.
+ *
+ * This may improve performance on weakly-ordered architectures where
+ * the compiler implements memory_order_consume as a
+ * memory_order_acquire, which is stricter than required by the
+ * standard.
+ *
+ * Note that using volatile accesses for rcu_dereference may cause
+ * LTO to generate incorrectly ordered code starting from C11/C++11.
  *
  * Should match rcu_assign_pointer() or rcu_xchg_pointer().
  *
@@ -64,14 +69,21 @@ extern "C" {
  * meets the 10-line criterion in LGPL, allowing this function to be
  * expanded directly in non-LGPL code.
  */
-#define _rcu_dereference(p)						\
-				__extension__				\
-				({					\
-				__typeof__(p) _________p1 = CMM_LOAD_SHARED(p); \
-				cmm_smp_read_barrier_depends();		\
-				(_________p1);				\
-				})
 
+#if !defined (URCU_DEREFERENCE_USE_VOLATILE) &&		\
+	((defined (__cplusplus) && __cplusplus >= 201103L) ||	\
+	(defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L))
+# define __URCU_DEREFERENCE_USE_ATOMIC_CONSUME
+#endif
+
+/*
+ * If p is const (the pointer itself, not what it points to), using
+ * __typeof__(p) would declare a const variable, leading to
+ * -Wincompatible-pointer-types errors.  Using the statement expression
+ * makes it an rvalue and gets rid of the const-ness.
+ */
+# define _rcu_dereference(p)			\
+	uatomic_load(&(p), CMM_CONSUME)
 /**
  * _rcu_cmpxchg_pointer - same as rcu_assign_pointer, but tests if the pointer
  * is as expected by "old". If succeeds, returns the previous pointer to the
@@ -79,7 +91,7 @@ extern "C" {
  * using synchronize_rcu(). If fails (unexpected value), returns old (which
  * should not be freed !).
  *
- * uatomic_cmpxchg() acts as both release and acquire barriers.
+ * uatomic_cmpxchg() acts as both release and acquire barriers on success.
  *
  * This macro is less than 10 lines long.  The intent is that this macro
  * meets the 10-line criterion in LGPL, allowing this function to be
@@ -90,8 +102,9 @@ extern "C" {
 	({								\
 		__typeof__(*p) _________pold = (old);			\
 		__typeof__(*p) _________pnew = (_new);			\
-		uatomic_cmpxchg(p, _________pold, _________pnew);	\
-	})
+		uatomic_cmpxchg_mo(p, _________pold, _________pnew,	\
+				   CMM_SEQ_CST, CMM_RELAXED);		\
+	});
 
 /**
  * _rcu_xchg_pointer - same as rcu_assign_pointer, but returns the previous
@@ -108,17 +121,17 @@ extern "C" {
 	__extension__					\
 	({						\
 		__typeof__(*p) _________pv = (v);	\
-		uatomic_xchg(p, _________pv);		\
+		uatomic_xchg_mo(p, _________pv,		\
+				CMM_SEQ_CST);		\
 	})
 
 
-#define _rcu_set_pointer(p, v)				\
-	do {						\
-		__typeof__(*p) _________pv = (v);	\
-		if (!__builtin_constant_p(v) || 	\
-		    ((v) != NULL))			\
-			cmm_wmb();				\
-		uatomic_set(p, _________pv);		\
+#define _rcu_set_pointer(p, v)						\
+	do {								\
+		__typeof__(*p) _________pv = (v);			\
+		uatomic_store(p, _________pv,				\
+			__builtin_constant_p(v) && (v) == NULL ?	\
+			CMM_RELAXED : CMM_RELEASE);			\
 	} while (0)
 
 /**