# HG changeset patch
# User shade
# Date 1409821885 -14400
# Node ID efb5379034ee29ca69cb4f62023fc2dc1b86f2af
# Parent  1a3bdc233bda70f93102af2550244f1a4faa16b4
8050147: StoreLoad barrier interferes with stack usages
Reviewed-by: jrose, kvn, drchase

diff -r 1a3bdc233bda -r efb5379034ee src/cpu/x86/vm/assembler_x86.hpp
--- a/src/cpu/x86/vm/assembler_x86.hpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/cpu/x86/vm/assembler_x86.hpp	Thu Sep 04 13:11:25 2014 +0400
@@ -26,6 +26,7 @@
 #define CPU_X86_VM_ASSEMBLER_X86_HPP
 
 #include "asm/register.hpp"
+#include "vm_version_x86.hpp"
 
 class BiasedLockingCounters;
 
@@ -1283,14 +1284,34 @@
       if (order_constraint & StoreLoad) {
         // All usable chips support "locked" instructions which suffice
         // as barriers, and are much faster than the alternative of
-        // using cpuid instruction. We use here a locked add [esp],0.
+        // using cpuid instruction. We use here a locked add [esp-C],0.
         // This is conveniently otherwise a no-op except for blowing
-        // flags.
+        // flags, and introducing a false dependency on target memory
+        // location. We can't do anything with flags, but we can avoid
+        // memory dependencies in the current method by locked-adding
+        // somewhere else on the stack. Doing [esp+C] will collide with
+        // something on stack in current method, hence we go for [esp-C].
+        // It is convenient since it is almost always in data cache, for
+        // any small C.  We need to step back from SP to avoid data
+        // dependencies with other things on below SP (callee-saves, for
+        // example). Without a clear way to figure out the minimal safe
+        // distance from SP, it makes sense to step back the complete
+        // cache line, as this will also avoid possible second-order effects
+        // with locked ops against the cache line. Our choice of offset
+        // is bounded by x86 operand encoding, which should stay within
+        // [-128; +127] to have the 8-byte displacement encoding.
+        //
         // Any change to this code may need to revisit other places in
         // the code where this idiom is used, in particular the
         // orderAccess code.
+
+        int offset = -VM_Version::L1_line_size();
+        if (offset < -128) {
+          offset = -128;
+        }
+
         lock();
-        addl(Address(rsp, 0), 0);// Assert the lock# signal here
+        addl(Address(rsp, offset), 0);// Assert the lock# signal here
       }
     }
   }
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/aix_ppc/vm/os_aix_ppc.cpp
--- a/src/os_cpu/aix_ppc/vm/os_aix_ppc.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/aix_ppc/vm/os_aix_ppc.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -563,3 +563,8 @@
   assert(((intptr_t)os::current_stack_pointer() & (StackAlignmentInBytes-1)) == 0, "incorrect stack alignment");
 }
 #endif
+
+void os::extra_bang_size_in_bytes() {
+  // PPC does not require the additional stack bang.
+  return 0;
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/bsd_x86/vm/os_bsd_x86.cpp
--- a/src/os_cpu/bsd_x86/vm/os_bsd_x86.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/bsd_x86/vm/os_bsd_x86.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -1030,3 +1030,8 @@
 void os::verify_stack_alignment() {
 }
 #endif
+
+int os::extra_bang_size_in_bytes() {
+  // JDK-8050147 requires the full cache line bang for x86.
+  return VM_Version::L1_line_size();
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/bsd_zero/vm/os_bsd_zero.cpp
--- a/src/os_cpu/bsd_zero/vm/os_bsd_zero.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/bsd_zero/vm/os_bsd_zero.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -465,3 +465,8 @@
 void os::verify_stack_alignment() {
 }
 #endif
+
+int os::extra_bang_size_in_bytes() {
+  // Zero does not require an additional stack bang.
+  return 0;
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/linux_ppc/vm/os_linux_ppc.cpp
--- a/src/os_cpu/linux_ppc/vm/os_linux_ppc.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/linux_ppc/vm/os_linux_ppc.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -612,3 +612,8 @@
   assert(((intptr_t)os::current_stack_pointer() & (StackAlignmentInBytes-1)) == 0, "incorrect stack alignment");
 }
 #endif
+
+int os:extra_bang_size_in_bytes() {
+  // PPC does not require the additional stack bang.
+  return 0;
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/linux_sparc/vm/os_linux_sparc.cpp
--- a/src/os_cpu/linux_sparc/vm/os_linux_sparc.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/linux_sparc/vm/os_linux_sparc.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -752,3 +752,8 @@
 void os::verify_stack_alignment() {
 }
 #endif
+
+int os::extra_bang_size_in_bytes() {
+  // SPARC does not require the additional stack bang.
+  return 0;
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/linux_x86/vm/os_linux_x86.cpp
--- a/src/os_cpu/linux_x86/vm/os_linux_x86.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/linux_x86/vm/os_linux_x86.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -930,3 +930,8 @@
   // keep the page mapped so CS limit isn't reduced.
 #endif
 }
+
+int os::extra_bang_size_in_bytes() {
+  // JDK-8050147 requires the full cache line bang for x86.
+  return VM_Version::L1_line_size();
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/linux_zero/vm/os_linux_zero.cpp
--- a/src/os_cpu/linux_zero/vm/os_linux_zero.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/linux_zero/vm/os_linux_zero.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -495,3 +495,8 @@
 void os::verify_stack_alignment() {
 }
 #endif
+
+int os::extra_bang_size_in_bytes() {
+  // Zero does not require an additional stack banging.
+  return 0;
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp
--- a/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -774,3 +774,8 @@
 void os::verify_stack_alignment() {
 }
 #endif
+
+int os::extra_bang_size_in_bytes() {
+  // SPARC does not require an additional stack bang.
+  return 0;
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp
--- a/src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -918,3 +918,8 @@
 #endif
 }
 #endif
+
+int os::extra_bang_size_in_bytes() {
+  // JDK-8050147 requires the full cache line bang for x86.
+  return VM_Version::L1_line_size();
+}
diff -r 1a3bdc233bda -r efb5379034ee src/os_cpu/windows_x86/vm/os_windows_x86.cpp
--- a/src/os_cpu/windows_x86/vm/os_windows_x86.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/os_cpu/windows_x86/vm/os_windows_x86.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -639,3 +639,8 @@
 #endif
 }
 #endif
+
+int os::extra_bang_size_in_bytes() {
+  // JDK-8050147 requires the full cache line bang for x86.
+  return VM_Version::L1_line_size();
+}
diff -r 1a3bdc233bda -r efb5379034ee src/share/vm/c1/c1_LIRAssembler.cpp
--- a/src/share/vm/c1/c1_LIRAssembler.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/share/vm/c1/c1_LIRAssembler.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -170,7 +170,7 @@
 // removes the need to bang the stack in the deoptimization blob which
 // in turn simplifies stack overflow handling.
 int LIR_Assembler::bang_size_in_bytes() const {
-  return MAX2(initial_frame_size_in_bytes(), _compilation->interpreter_frame_size());
+  return MAX2(initial_frame_size_in_bytes() + os::extra_bang_size_in_bytes(), _compilation->interpreter_frame_size());
 }
 
 void LIR_Assembler::emit_exception_entries(ExceptionInfoList* info_list) {
diff -r 1a3bdc233bda -r efb5379034ee src/share/vm/opto/compile.cpp
--- a/src/share/vm/opto/compile.cpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/share/vm/opto/compile.cpp	Thu Sep 04 13:11:25 2014 +0400
@@ -430,7 +430,7 @@
 // removes the need to bang the stack in the deoptimization blob which
 // in turn simplifies stack overflow handling.
 int Compile::bang_size_in_bytes() const {
-  return MAX2(_interpreter_frame_size, frame_size_in_bytes());
+  return MAX2(frame_size_in_bytes() + os::extra_bang_size_in_bytes(), _interpreter_frame_size);
 }
 
 // ============================================================================
diff -r 1a3bdc233bda -r efb5379034ee src/share/vm/runtime/os.hpp
--- a/src/share/vm/runtime/os.hpp	Fri Aug 29 08:14:19 2014 -0700
+++ b/src/share/vm/runtime/os.hpp	Thu Sep 04 13:11:25 2014 +0400
@@ -761,6 +761,9 @@
   // Hook for os specific jvm options that we don't want to abort on seeing
   static bool obsolete_option(const JavaVMOption *option);
 
+  // Amount beyond the callee frame size that we bang the stack.
+  static int extra_bang_size_in_bytes();
+
   // Extensions
 #include "runtime/os_ext.hpp"