8320892: AArch64: Restore FPU control state after JNI

Reviewed-by: adinn, stuefe

src/hotspot/cpu/aarch64/downcallLinker_aarch64.cpp

@@ -278,6 +278,9 @@ void DowncallLinker::StubGenerator::generate() {
   Label L_reguard;
   Label L_after_reguard;
   if (_needs_transition) {
+    // Restore cpu control state after JNI call
+    __ restore_cpu_control_state_after_jni(rscratch1, tmp1);
+
     __ mov(tmp1, _thread_in_native_trans);
     __ strw(tmp1, Address(rthread, JavaThread::thread_state_offset()));
 

src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp

@@ -4428,6 +4428,23 @@ void MacroAssembler::load_klass(Register dst, Register src) {
   }
 }
 
+void MacroAssembler::restore_cpu_control_state_after_jni(Register tmp1, Register tmp2) {
+  if (RestoreMXCSROnJNICalls) {
+    Label OK;
+    get_fpcr(tmp1);
+    mov(tmp2, tmp1);
+    // Set FPCR to the state we need. We do want Round to Nearest. We
+    // don't want non-IEEE rounding modes or floating-point traps.
+    bfi(tmp1, zr, 22, 4); // Clear DN, FZ, and Rmode
+    bfi(tmp1, zr, 8, 5);  // Clear exception-control bits (8-12)
+    bfi(tmp1, zr, 0, 2);  // Clear AH:FIZ
+    eor(tmp2, tmp1, tmp2);
+    cbz(tmp2, OK);        // Only reset FPCR if it's wrong
+    set_fpcr(tmp1);
+    bind(OK);
+  }
+}
+
 // ((OopHandle)result).resolve();
 void MacroAssembler::resolve_oop_handle(Register result, Register tmp1, Register tmp2) {
   // OopHandle::resolve is an indirection.

src/hotspot/cpu/aarch64/macroAssembler_aarch64.hpp

@@ -1047,8 +1047,8 @@ public:
 #define verify_method_ptr(reg) _verify_method_ptr(reg, "broken method " #reg, __FILE__, __LINE__)
 #define verify_klass_ptr(reg) _verify_klass_ptr(reg, "broken klass " #reg, __FILE__, __LINE__)
 
-  // only if +VerifyFPU
-  void verify_FPU(int stack_depth, const char* s = "illegal FPU state");
+  // Restore cpu control state after JNI call
+  void restore_cpu_control_state_after_jni(Register tmp1, Register tmp2);
 
   // prints msg, dumps registers and stops execution
   void stop(const char* msg);

src/hotspot/cpu/aarch64/sharedRuntime_aarch64.cpp

@@ -1838,6 +1838,9 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
   intptr_t return_pc = (intptr_t) __ pc();
   oop_maps->add_gc_map(return_pc - start, map);
 
+  // Verify or restore cpu control state after JNI call
+  __ restore_cpu_control_state_after_jni(rscratch1, rscratch2);
+
   // Unpack native results.
   switch (ret_type) {
   case T_BOOLEAN: __ c2bool(r0);                     break;

src/hotspot/cpu/aarch64/templateInterpreterGenerator_aarch64.cpp

@@ -1383,6 +1383,9 @@ address TemplateInterpreterGenerator::generate_native_entry(bool synchronized) {
   __ get_method(rmethod);
   // result potentially in r0 or v0
 
+  // Restore cpu control state after JNI call
+  __ restore_cpu_control_state_after_jni(rscratch1, rscratch2);
+
   // make room for the pushes we're about to do
   __ sub(rscratch1, esp, 4 * wordSize);
   __ andr(sp, rscratch1, -16);