replace: add checks for atomic_thread_fence(memory_order_seq_cst) and add possible fallbacks

This implements a full memory barrier.
On ubuntu amd64 with results in an 'mfence' instruction.

This is required to syncronization between threads, where
there's typically only one write of a memory that should be
synced between all threads with the barrier.

Much more details can be found here:
https://gcc.gnu.org/onlinedocs/gcc-7.3.0/gcc/_005f_005fatomic-Builtins.html#g_t_005f_005fatomic-Builtins
https://gcc.gnu.org/onlinedocs/gcc-7.3.0/gcc/_005f_005fsync-Builtins.html#g_t_005f_005fsync-Builtins

The main one we use seems to be in C11 via stdatomic.h,
the oldest fallback is __sync_synchronize(), which is available
since 2005 in gcc.

Signed-off-by: Stefan Metzmacher <metze@samba.org>
Reviewed-by: Ralph Boehme <slow@samba.org>

2 files changed, 45 insertions, 1 deletions

diff --git a/lib/replace/system/threads.h b/lib/replace/system/threads.h
index fe6d0fbac54..d189ed620c5 100644
--- a/lib/replace/system/threads.h
+++ b/lib/replace/system/threads.h
@@ -42,4 +42,31 @@
 #define pthread_mutex_consistent pthread_mutex_consistent_np
 #endif
 
+#ifdef HAVE_STDATOMIC_H
+#include <stdatomic.h>
+#endif
+
+#ifndef HAVE_ATOMIC_THREAD_FENCE
+#ifdef HAVE___ATOMIC_THREAD_FENCE
+#define atomic_thread_fence(__ignore_order) __atomic_thread_fence(__ATOMIC_SEQ_CST)
+#define HAVE_ATOMIC_THREAD_FENCE 1
+#endif /* HAVE___ATOMIC_THREAD_FENCE */
+#endif /* not HAVE_ATOMIC_THREAD_FENCE */
+
+#ifndef HAVE_ATOMIC_THREAD_FENCE
+#ifdef HAVE___SYNC_SYNCHRONIZE
+#define atomic_thread_fence(__ignore_order) __sync_synchronize()
+#define HAVE_ATOMIC_THREAD_FENCE 1
+#endif /* HAVE___SYNC_SYNCHRONIZE */
+#endif /* not HAVE_ATOMIC_THREAD_FENCE */
+
+#ifndef HAVE_ATOMIC_THREAD_FENCE
+#ifdef HAVE_ATOMIC_THREAD_FENCE_SUPPORT
+#error mismatch_error_between_configure_test_and_header
+#endif
+/* make sure the build fails if someone uses it without checking the define */
+#define atomic_thread_fence(__order) \
+        __function__atomic_thread_fence_not_available_on_this_platform__()
+#endif /* not HAVE_ATOMIC_THREAD_FENCE */
+
 #endif
diff --git a/lib/replace/wscript b/lib/replace/wscript
index 02d98c59e47..5b53461f844 100644
--- a/lib/replace/wscript
+++ b/lib/replace/wscript
@@ -113,7 +113,7 @@ def configure(conf):
     conf.CHECK_HEADERS('sys/extattr.h sys/ea.h sys/proplist.h sys/cdefs.h')
     conf.CHECK_HEADERS('utmp.h utmpx.h lastlog.h')
     conf.CHECK_HEADERS('syscall.h sys/syscall.h inttypes.h')
-    conf.CHECK_HEADERS('sys/atomic.h')
+    conf.CHECK_HEADERS('sys/atomic.h stdatomic.h')
     conf.CHECK_HEADERS('libgen.h')
 
     if conf.CHECK_CFLAGS('-Wno-format-truncation'):
@@ -260,6 +260,23 @@ def configure(conf):
                     headers='stdint.h sys/atomic.h',
                     msg='Checking for atomic_add_32 compiler builtin')
 
+    # Check for thread fence. */
+    tf = conf.CHECK_CODE('atomic_thread_fence(memory_order_seq_cst);',
+                         'HAVE_ATOMIC_THREAD_FENCE',
+                         headers='stdatomic.h',
+                         msg='Checking for atomic_thread_fence(memory_order_seq_cst) in stdatomic.h')
+    if not tf:
+        tf = conf.CHECK_CODE('__atomic_thread_fence(__ATOMIC_SEQ_CST);',
+                             'HAVE___ATOMIC_THREAD_FENCE',
+                             msg='Checking for __atomic_thread_fence(__ATOMIC_SEQ_CST)')
+    if not tf:
+        # __sync_synchronize() is available since 2005 in gcc.
+        tf = conf.CHECK_CODE('__sync_synchronize();',
+                             'HAVE___SYNC_SYNCHRONIZE',
+                             msg='Checking for __sync_synchronize')
+    if tf:
+        conf.DEFINE('HAVE_ATOMIC_THREAD_FENCE_SUPPORT', 1)
+
     conf.CHECK_CODE('''
                     #define FALL_THROUGH __attribute__((fallthrough))