membarrier: add --enable-membarrier

Actually enable the global memory barriers if supported by the OS.
Because only recent versions of Linux include the support, they
are disabled by default. Note that it also has to be disabled
for QEMU to run under Wine.

Before this patch, rcutorture reports 85 ns/read for my machine,
after the patch it reports 12.5 ns/read. On the other hand updates
go from 50 *micro*seconds to 20 *milli*seconds.

Backports commit a40161cbe9ccbcbab798c3e4d257c4bba99d153a from qemu

qemu/configure

@@ -175,6 +175,7 @@ solaris="no"
 softmmu="yes"
 pie=""
 tcg="yes"
+membarrier=""
 cpuid_h="no"
 avx2_opt="no"
 
@@ -562,6 +563,10 @@ for opt do
   ;;
   --disable-stack-protector) stack_protector="no"
   ;;
+  --disable-membarrier) membarrier="no"
+  ;;
+  --enable-membarrier) membarrier="yes"
+  ;;
   *)
       echo "ERROR: unknown option $opt"
       echo "Try '$0 --help' for more information"
@@ -1181,6 +1186,37 @@ if compile_prog "" "" ; then
   vector16=yes
 fi
 
+##########################################
+# check for usable membarrier system call
+if test "$membarrier" = "yes"; then
+    have_membarrier=no
+    if test "$mingw32" = "yes" ; then
+        have_membarrier=yes
+    elif test "$linux" = "yes" ; then
+        cat > $TMPC << EOF
+    #include <linux/membarrier.h>
+    #include <sys/syscall.h>
+    #include <unistd.h>
+    #include <stdlib.h>
+    int main(void) {
+        syscall(__NR_membarrier, MEMBARRIER_CMD_QUERY, 0);
+        syscall(__NR_membarrier, MEMBARRIER_CMD_SHARED, 0);
+  exit(0);
+    }
+EOF
+        if compile_prog "" "" ; then
+            have_membarrier=yes
+        fi
+    fi
+    if test "$have_membarrier" = "no"; then
+      feature_not_found "membarrier" "membarrier system call not available"
+    fi
+else
+    # Do not enable it by default even for Mingw32, because it doesn't
+    # work on Wine.
+    membarrier=no
+fi
+
 #########################################
 # See if 128-bit atomic operations are supported.
 
@@ -1323,6 +1359,7 @@ if test -n "$sparc_cpu"; then
     echo "Target Sparc Arch $sparc_cpu"
 fi
 echo "PIE               $pie"
+echo "membarrier        $membarrier"
 echo "TCG support       $tcg"
 if test "$tcg" = "yes" ; then
     echo "TCG debug enabled $debug_tcg"
@@ -1408,6 +1445,9 @@ fi
 if test "$bswap_h" = "yes" ; then
   echo "CONFIG_MACHINE_BSWAP_H=y" >> $config_host_mak
 fi
+if test "$membarrier" = "yes" ; then
+  echo "CONFIG_MEMBARRIER=y" >> $config_host_mak
+fi
 if test "$tcg" = "yes"; then
   echo "CONFIG_TCG=y" >> $config_host_mak
   if test "$tcg_interpreter" = "yes" ; then

qemu/include/qemu/sys_membarrier.h

@@ -9,9 +9,19 @@
 #ifndef QEMU_SYS_MEMBARRIER_H
 #define QEMU_SYS_MEMBARRIER_H 1
 
+#ifdef CONFIG_MEMBARRIER
+/* Only block reordering at the compiler level in the performance-critical
+ * side.  The slow side forces processor-level ordering on all other cores
+ * through a system call.
+ */
+extern void smp_mb_global_init(void);
+extern void smp_mb_global(void);
+#define smp_mb_placeholder()       barrier()
+#else
 /* Keep it simple, execute a real memory barrier on both sides.  */
 static inline void smp_mb_global_init(void) {}
 #define smp_mb_global()            smp_mb()
 #define smp_mb_placeholder()       smp_mb()
+#endif
 
 #endif

qemu/util/Makefile.objs

@@ -11,3 +11,4 @@ util-obj-y += host-utils.o
 util-obj-y += getauxval.o
 util-obj-y += log.o
 util-obj-y += range.o
+util-obj-$(CONFIG_MEMBARRIER) += sys_membarrier.o

qemu/util/sys_membarrier.c

@@ -0,0 +1,49 @@
+/*
+ * Process-global memory barriers
+ *
+ * Copyright (c) 2018 Red Hat, Inc.
+ *
+ * Author: Paolo Bonzini <pbonzini@redhat.com>
+ */
+
+#include <qemu/osdep.h>
+#include <qemu/sys_membarrier.h>
+
+#ifdef CONFIG_LINUX
+#include <linux/membarrier.h>
+#include <sys/syscall.h>
+
+static int
+membarrier(int cmd, int flags)
+{
+    return syscall(__NR_membarrier, cmd, flags);
+}
+#endif
+
+void smp_mb_global(void)
+{
+#if defined CONFIG_WIN32
+    FlushProcessWriteBuffers();
+#elif defined CONFIG_LINUX
+    membarrier(MEMBARRIER_CMD_SHARED, 0);
+#else
+#error --enable-membarrier is not supported on this operating system.
+#endif
+}
+
+void smp_mb_global_init(void)
+{
+#ifdef CONFIG_LINUX
+    int ret = membarrier(MEMBARRIER_CMD_QUERY, 0);
+    if (ret < 0) {
+        //error_report("This QEMU binary requires the membarrier system call.");
+        //error_report("Please upgrade your system to a newer version of Linux");
+        exit(1);
+    }
+    if (!(ret & MEMBARRIER_CMD_SHARED)) {
+        //error_report("This QEMU binary requires MEMBARRIER_CMD_SHARED support.");
+        //error_report("Please upgrade your system to a newer version of Linux");
+        exit(1);
+    }
+#endif
+}