diff options
Diffstat (limited to 'src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp')
-rw-r--r-- | src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp | 16 |
1 files changed, 12 insertions, 4 deletions
diff --git a/src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp b/src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp index 33ce50dc7..b7aca02e0 100644 --- a/src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp +++ b/src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp @@ -1,6 +1,6 @@ /* - * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved. - * Copyright 2007, 2008, 2011 Red Hat, Inc. + * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved. + * Copyright 2007, 2008, 2011, 2015, Red Hat, Inc. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -232,7 +232,13 @@ inline jint Atomic::xchg(jint exchange_value, volatile jint* dest) { // operation. Note that some platforms only support this with the // limitation that the only valid value to store is the immediate // constant 1. There is a test for this in JNI_CreateJavaVM(). - return __sync_lock_test_and_set (dest, exchange_value); + jint result = __sync_lock_test_and_set (dest, exchange_value); + // All atomic operations are expected to be full memory barriers + // (see atomic.hpp). However, __sync_lock_test_and_set is not + // a full memory barrier, but an acquire barrier. Hence, this added + // barrier. + __sync_synchronize(); + return result; #endif // M68K #endif // ARM } @@ -245,7 +251,9 @@ inline intptr_t Atomic::xchg_ptr(intptr_t exchange_value, #ifdef M68K return m68k_lock_test_and_set(dest, exchange_value); #else - return __sync_lock_test_and_set (dest, exchange_value); + intptr_t result = __sync_lock_test_and_set (dest, exchange_value); + __sync_synchronize(); + return result; #endif // M68K #endif // ARM } |