--- a/ChangeLog	Tue May 19 15:58:51 2009 +0100
+++ b/ChangeLog	Thu Apr 23 16:59:47 2009 +0200
@@ -1,3 +1,17 @@
+2009-04-23 Xerxes Rånby <xerxes@zafena.se>
+
+	* ports/hotspot/src/share/vm/shark/sharkBuilder.cpp
+	(SharkBuilder::init_external_functions): Use memory address to
+	kernel helper __kernel_dmb on arm for memory barrier.
+	* ports/hotspot/src/os_cpu/linux_zero/vm/orderAccess_linux_zero.inline.hpp
+	(__kernel_dmb_t): New type.
+	(__kernel_dmb): New macro.
+	(FULL_MEM_BARRIER): New macro that uses __kernel_dmb on arm
+	and __sync_synchronize on all other architectures.
+	(READ_MEM_BARRIER): Updated to use __kernel_dmb on arm.
+	(WRITE_MEM_BARRIER): Likewise.
+	(OrderAccess::fence): Updated to use FULL_MEM_BARRIER.
+
 2009-04-23 Xerxes Rånby <xerxes@zafena.se>
 
 	* ports/hotspot/src/share/vm/shark/sharkBuilder.cpp

--- a/ports/hotspot/src/os_cpu/linux_zero/vm/orderAccess_linux_zero.inline.hpp	Tue May 19 15:58:51 2009 +0100
+++ b/ports/hotspot/src/os_cpu/linux_zero/vm/orderAccess_linux_zero.inline.hpp	Thu Apr 23 16:59:47 2009 +0200
@@ -23,6 +23,25 @@
  *
  */
 
+#ifdef ARM
+
+/* 
+ * ARM Kernel helper for memory barrier. 
+ * Using __asm __volatile ("":::"memory") does not work reliable on ARM
+ * and gcc __sync_synchronize(); implementation does not use the kernel
+ * helper for all gcc versions so it is unreliable to use as well. 
+ */
+typedef void (__kernel_dmb_t) (void);
+#define __kernel_dmb (*(__kernel_dmb_t *) 0xffff0fa0)
+
+#define FULL_MEM_BARRIER __kernel_dmb()
+#define READ_MEM_BARRIER __kernel_dmb()
+#define WRITE_MEM_BARRIER __kernel_dmb()
+
+#else // ARM
+
+#define FULL_MEM_BARRIER __sync_synchronize()
+
 #ifdef PPC
 
 #define READ_MEM_BARRIER __asm __volatile ("isync":::"memory")
@@ -39,6 +58,7 @@
 
 #endif // PPC
 
+#endif // ARM
 
 
 inline void OrderAccess::loadload()   { acquire(); }
@@ -58,7 +78,7 @@
 
 inline void OrderAccess::fence()
 {
-  __sync_synchronize();
+  FULL_MEM_BARRIER;
 }
 
 inline jbyte    OrderAccess::load_acquire(volatile jbyte*   p) { jbyte data = *p; acquire(); return data; }

--- a/ports/hotspot/src/share/vm/shark/sharkBuilder.cpp	Tue May 19 15:58:51 2009 +0100
+++ b/ports/hotspot/src/share/vm/shark/sharkBuilder.cpp	Thu Apr 23 16:59:47 2009 +0200
@@ -138,12 +138,26 @@
       "llvm.atomic.cmp.swap.i" LP64_ONLY("64") NOT_LP64("32"), type));
 #endif
 
+ /* 
+  *  The five booleans passed to llvm.memory.barrier are used like this:
+  *  The first four arguments enables a specific barrier in this order:
+  *  load-load, load-store, store-load and store-store.
+  *  The fith argument specifies that the barrier applies to io or device
+  *  or uncached memory.
+  */
   params.clear();
   for (int i = 0; i < 5; i++)
     params.push_back(Type::Int1Ty);
   type = FunctionType::get(Type::VoidTy, params, false);
   set_llvm_memory_barrier_fn(
+#ifdef ARM
+    make_function(
+      0xffff0fa0, // __kernel_dmb
+      type,
+      "__kernel_dmb"));
+#else
     module()->getOrInsertFunction("llvm.memory.barrier", type));
+#endif
 
   params.clear();
   params.push_back(SharkType::jdouble_type());