Merge

.hgtags-top-repo

@@ -61,3 +61,4 @@ e1176f86805fe07fd9fb9da065dc51b47712ce76 jdk7-b82
 2f3ea057d1ad56cf3b269cdc4de2741411151982 jdk7-b84
 cf26288a114be67c39f2758959ce50b60f5ae330 jdk7-b85
 433a60a9c0bf1b26ee7e65cebaa89c541f497aed jdk7-b86
+6b1069f53fbc30663ccef49d78c31bb7d6967bde jdk7-b87

corba/.hgtags

@@ -61,3 +61,4 @@ fde0df7a2384f7fe33204a79678989807d9c2b98 jdk7-b83
 68c8961a82e4a3ad2a67991e5d834192a81eb4cd jdk7-b84
 c67a9df7bc0ca291f08f9a9cc05cb78ea15d25e6 jdk7-b85
 6253e28826d16cf1aecc39ce04c8de1f6bf2df5f jdk7-b86
+09a41111a401d327f65e453384d976a10154d9ea jdk7-b87

hotspot/.hgtags

@@ -85,3 +85,4 @@ ffc8d176b84bcfb5ac21302b4feb3b0c0d69b97c jdk7-b84
 418bc80ce13995149eadc9eecbba21d7a9fa02ae hs17-b10
 bf823ef06b4f211e66988d76a2e2669be5c0820e jdk7-b86
 07226e9eab8f74b37346b32715f829a2ef2c3188 hs18-b01
+e7e7e36ccdb5d56edd47e5744351202d38f3b7ad jdk7-b87

hotspot/make/hotspot_version

@@ -35,7 +35,7 @@ HOTSPOT_VM_COPYRIGHT=Copyright 2010
 
 HS_MAJOR_VER=18
 HS_MINOR_VER=0
-HS_BUILD_NUMBER=01
+HS_BUILD_NUMBER=02
 
 JDK_MAJOR_VER=1
 JDK_MINOR_VER=7

hotspot/make/windows/build.bat

@@ -28,6 +28,9 @@ REM
 REM Since we don't have uname and we could be cross-compiling,
 REM Use the compiler to determine which ARCH we are building
 REM 
+REM Note: Running this batch file from the Windows command shell requires
+REM that "grep" be accessible on the PATH. An MKS install does this.
+REM 
 cl 2>&1 | grep "IA-64" >NUL
 if %errorlevel% == 0 goto isia64
 cl 2>&1 | grep "AMD64" >NUL
@@ -57,6 +60,7 @@ if not "%7" == ""      goto usage
 if "%1" == "product"   goto test1
 if "%1" == "debug"     goto test1
 if "%1" == "fastdebug" goto test1
+if "%1" == "tree"      goto test1
 goto usage
 
 :test1
@@ -70,6 +74,7 @@ if "%2" == "adlc"      goto build_adlc
 goto usage
 
 :test2
+if "%1" == "tree"      goto build_tree
 REM check_j2se_version
 REM jvmti.make requires J2SE 1.4.x or newer.
 REM If not found then fail fast.
@@ -93,6 +98,10 @@ goto end
 nmake -f %3/make/windows/build.make Variant=compiler2 WorkSpace=%3 BootStrapDir=%4 BuildUser="%USERNAME%" HOTSPOT_BUILD_VERSION=%5 ADLC_ONLY=1 %1
 goto end
 
+:build_tree
+nmake -f %3/make/windows/build.make Variant=%2 WorkSpace=%3 BootStrapDir=%4 BuildUser="%USERNAME%" HOTSPOT_BUILD_VERSION="%5" %1
+goto end
+
 :usage
 echo Usage: build flavor version workspace bootstrap_dir [build_id] [windbg_home]
 echo.
@@ -100,8 +109,10 @@ echo where:
 echo flavor is "product", "debug" or "fastdebug",
 echo version is "core", "kernel", "compiler1", "compiler2", or "tiered",
 echo workspace is source directory without trailing slash, 
-echo bootstrap_dir is a full path to echo a JDK in which bin/java 
+echo bootstrap_dir is a full path to a JDK in which bin/java 
-echo   and bin/javac are present and working, and echo build_id is an 
+echo   and bin/javac are present and working, and build_id is an 
 echo   optional build identifier displayed by java -version
+exit /b 1
 
 :end
+exit /b %errorlevel%

hotspot/make/windows/build.make

@@ -27,6 +27,9 @@
 # environment variables (Variant, WorkSpace, BootStrapDir, BuildUser, HOTSPOT_BUILD_VERSION)
 # are passed in as command line arguments.
 
+# Note: Running nmake or build.bat from the Windows command shell requires
+# that "sh" be accessible on the PATH. An MKS install does this.
+
 # SA components are built if BUILD_WIN_SA=1 is specified.
 # See notes in README. This produces files:
 #  1. sa-jdi.jar       - This is built before building jvm.dll
@@ -233,6 +236,12 @@ develop: checks $(variantDir) $(variantDir)\local.make sanity
 	cd $(variantDir)
 	nmake -nologo -f $(WorkSpace)\make\windows\makefiles\top.make BUILD_FLAVOR=product DEVELOP=1 ARCH=$(ARCH)
 
+# target to create just the directory structure
+tree: checks $(variantDir) $(variantDir)\local.make sanity
+	mkdir $(variantDir)\product
+	mkdir $(variantDir)\debug
+	mkdir $(variantDir)\fastdebug
+
 sanity:
 	@ echo;
 	@ cd $(variantDir)

hotspot/make/windows/create.bat

@@ -36,6 +36,9 @@ REM
 REM Since we don't have uname and we could be cross-compiling,
 REM Use the compiler to determine which ARCH we are building
 REM 
+REM Note: Running this batch file from the Windows command shell requires
+REM that "grep" be accessible on the PATH. An MKS install does this.
+REM 
 cl 2>&1 | grep "IA-64" >NUL
 if %errorlevel% == 0 goto isia64
 cl 2>&1 | grep "AMD64" >NUL

hotspot/make/windows/get_msc_ver.sh

@@ -22,6 +22,8 @@
 #  
 #
 
+set -e
+
 # This shell script echoes "MSC_VER=<munged version of cl>"
 # It ignores the micro version component.
 # Examples:
@@ -38,17 +40,20 @@
 # sh, and it has been found that sometimes `which sh` fails.
 
 if [ "x$HotSpotMksHome" != "x" ]; then
- MKS_HOME="$HotSpotMksHome"
+  TOOL_DIR="$HotSpotMksHome"
 else
+  # HotSpotMksHome is not set so use the directory that contains "sh".
+  # This works with both MKS and Cygwin.
   SH=`which sh`
- MKS_HOME=`dirname "$SH"`
+  TOOL_DIR=`dirname "$SH"`
 fi
 
-HEAD="$MKS_HOME/head"
+DIRNAME="$TOOL_DIR/dirname"
-ECHO="$MKS_HOME/echo"
+HEAD="$TOOL_DIR/head"
-EXPR="$MKS_HOME/expr"
+ECHO="$TOOL_DIR/echo"
-CUT="$MKS_HOME/cut"
+EXPR="$TOOL_DIR/expr"
-SED="$MKS_HOME/sed"
+CUT="$TOOL_DIR/cut"
+SED="$TOOL_DIR/sed"
 
 if [ "x$FORCE_MSC_VER" != "x" ]; then
   echo "MSC_VER=$FORCE_MSC_VER"
@@ -70,7 +75,15 @@ fi
 if [ "x$FORCE_LINK_VER" != "x" ]; then
   echo "LINK_VER=$FORCE_LINK_VER"
 else
-  LINK_VER_RAW=`link 2>&1 | "$HEAD" -n 1 | "$SED" 's/.*Version[\ ]*\([0-9][0-9.]*\).*/\1/'`
+  # use the "link" command that is co-located with the "cl" command
+  cl_cmd=`which cl`
+  if [ "x$cl_cmd" != "x" ]; then
+    link_cmd=`$DIRNAME "$cl_cmd"`/link
+  else
+    # which can't find "cl" so just use which ever "link" we find
+    link_cmd="link"
+  fi
+  LINK_VER_RAW=`"$link_cmd" 2>&1 | "$HEAD" -n 1 | "$SED" 's/.*Version[\ ]*\([0-9][0-9.]*\).*/\1/'`
   LINK_VER_MAJOR=`"$ECHO" $LINK_VER_RAW | "$CUT" -d'.' -f1`
   LINK_VER_MINOR=`"$ECHO" $LINK_VER_RAW | "$CUT" -d'.' -f2`
   LINK_VER_MICRO=`"$ECHO" $LINK_VER_RAW | "$CUT" -d'.' -f3`

hotspot/src/os/linux/vm/os_linux.cpp

@@ -22,6 +22,8 @@
  *
  */
 
+# define __STDC_FORMAT_MACROS
+
 // do not include  precompiled  header file
 # include "incls/_os_linux.cpp.incl"
 
@@ -53,6 +55,8 @@
 # include <sys/ipc.h>
 # include <sys/shm.h>
 # include <link.h>
+# include <stdint.h>
+# include <inttypes.h>
 
 #define MAX_PATH    (2 * K)
 
@@ -2492,6 +2496,91 @@ bool os::uncommit_memory(char* addr, size_t size) {
     != MAP_FAILED;
 }
 
+// Linux uses a growable mapping for the stack, and if the mapping for
+// the stack guard pages is not removed when we detach a thread the
+// stack cannot grow beyond the pages where the stack guard was
+// mapped.  If at some point later in the process the stack expands to
+// that point, the Linux kernel cannot expand the stack any further
+// because the guard pages are in the way, and a segfault occurs.
+//
+// However, it's essential not to split the stack region by unmapping
+// a region (leaving a hole) that's already part of the stack mapping,
+// so if the stack mapping has already grown beyond the guard pages at
+// the time we create them, we have to truncate the stack mapping.
+// So, we need to know the extent of the stack mapping when
+// create_stack_guard_pages() is called.
+
+// Find the bounds of the stack mapping.  Return true for success.
+//
+// We only need this for stacks that are growable: at the time of
+// writing thread stacks don't use growable mappings (i.e. those
+// creeated with MAP_GROWSDOWN), and aren't marked "[stack]", so this
+// only applies to the main thread.
+static bool
+get_stack_bounds(uintptr_t *bottom, uintptr_t *top)
+{
+  FILE *f = fopen("/proc/self/maps", "r");
+  if (f == NULL)
+    return false;
+
+  while (!feof(f)) {
+    size_t dummy;
+    char *str = NULL;
+    ssize_t len = getline(&str, &dummy, f);
+    if (len == -1) {
+      fclose(f);
+      return false;
+    }
+
+    if (len > 0 && str[len-1] == '\n') {
+      str[len-1] = 0;
+      len--;
+    }
+
+    static const char *stack_str = "[stack]";
+    if (len > (ssize_t)strlen(stack_str)
+       && (strcmp(str + len - strlen(stack_str), stack_str) == 0)) {
+      if (sscanf(str, "%" SCNxPTR "-%" SCNxPTR, bottom, top) == 2) {
+        uintptr_t sp = (uintptr_t)__builtin_frame_address(0);
+        if (sp >= *bottom && sp <= *top) {
+          free(str);
+          fclose(f);
+          return true;
+        }
+      }
+    }
+    free(str);
+  }
+  fclose(f);
+  return false;
+}
+
+// If the (growable) stack mapping already extends beyond the point
+// where we're going to put our guard pages, truncate the mapping at
+// that point by munmap()ping it.  This ensures that when we later
+// munmap() the guard pages we don't leave a hole in the stack
+// mapping.
+bool os::create_stack_guard_pages(char* addr, size_t size) {
+  uintptr_t stack_extent, stack_base;
+  if (get_stack_bounds(&stack_extent, &stack_base)) {
+    if (stack_extent < (uintptr_t)addr)
+      ::munmap((void*)stack_extent, (uintptr_t)addr - stack_extent);
+  }
+
+  return os::commit_memory(addr, size);
+}
+
+// If this is a growable mapping, remove the guard pages entirely by
+// munmap()ping them.  If not, just call uncommit_memory().
+bool os::remove_stack_guard_pages(char* addr, size_t size) {
+  uintptr_t stack_extent, stack_base;
+  if (get_stack_bounds(&stack_extent, &stack_base)) {
+    return ::munmap(addr, size) == 0;
+  }
+
+  return os::uncommit_memory(addr, size);
+}
+
 static address _highest_vm_reserved_address = NULL;
 
 // If 'fixed' is true, anon_mmap() will attempt to reserve anonymous memory

hotspot/src/os/solaris/dtrace/hotspot.d

@@ -25,9 +25,20 @@
 provider hotspot {
   probe class__loaded(char*, uintptr_t, void*, uintptr_t);
   probe class__unloaded(char*, uintptr_t, void*, uintptr_t);
+  probe class__initialization__required(char*, uintptr_t, void*, intptr_t,int);
+  probe class__initialization__recursive(char*, uintptr_t, void*, intptr_t,int);
+  probe class__initialization__concurrent(char*, uintptr_t, void*, intptr_t,int);
+  probe class__initialization__erroneous(char*, uintptr_t, void*, intptr_t, int);
+  probe class__initialization__super__failed(char*, uintptr_t, void*, intptr_t,int);
+  probe class__initialization__clinit(char*, uintptr_t, void*, intptr_t,int);
+  probe class__initialization__error(char*, uintptr_t, void*, intptr_t,int);
+  probe class__initialization__end(char*, uintptr_t, void*, intptr_t,int);
   probe vm__init__begin();
   probe vm__init__end();
   probe vm__shutdown();
+  probe vmops__request(char*, uintptr_t, int);
+  probe vmops__begin(char*, uintptr_t, int);
+  probe vmops__end(char*, uintptr_t, int);
   probe gc__begin(uintptr_t);
   probe gc__end();
   probe mem__pool__gc__begin(
@@ -38,6 +49,12 @@ provider hotspot {
     uintptr_t, uintptr_t, uintptr_t, uintptr_t);
   probe thread__start(char*, uintptr_t, uintptr_t, uintptr_t, uintptr_t);
   probe thread__stop(char*, uintptr_t, uintptr_t, uintptr_t, uintptr_t);
+  probe thread__sleep__begin(long long);
+  probe thread__sleep__end(int);
+  probe thread__yield();
+  probe thread__park__begin(uintptr_t, int, long long);
+  probe thread__park__end(uintptr_t);
+  probe thread__unpark(uintptr_t);
   probe method__compile__begin(
     char*, uintptr_t, char*, uintptr_t, char*, uintptr_t, char*, uintptr_t); 
   probe method__compile__end(

hotspot/src/os/solaris/vm/attachListener_solaris.cpp

@@ -668,13 +668,18 @@ jint AttachListener::pd_set_flag(AttachOperation* op, outputStream* out) {
     }
   }
 
-  if (strcmp(name, "ExtendedDTraceProbes") != 0) {
+  if (strcmp(name, "ExtendedDTraceProbes") == 0) {
-    out->print_cr("flag '%s' cannot be changed", name);
-    return JNI_ERR;
-  }
-
     DTrace::set_extended_dprobes(flag);
     return JNI_OK;
+  }
+
+  if (strcmp(name, "DTraceMonitorProbes") == 0) {
+    DTrace::set_monitor_dprobes(flag);
+    return JNI_OK;
+  }
+
+  out->print_cr("flag '%s' cannot be changed", name);
+  return JNI_ERR;
 }
 
 void AttachListener::pd_detachall() {

hotspot/src/os/solaris/vm/os_solaris.cpp

@@ -2698,6 +2698,14 @@ void os::free_memory(char* addr, size_t bytes) {
   }
 }
 
+bool os::create_stack_guard_pages(char* addr, size_t size) {
+  return os::commit_memory(addr, size);
+}
+
+bool os::remove_stack_guard_pages(char* addr, size_t size) {
+  return os::uncommit_memory(addr, size);
+}
+
 // Change the page size in a given range.
 void os::realign_memory(char *addr, size_t bytes, size_t alignment_hint) {
   assert((intptr_t)addr % alignment_hint == 0, "Address should be aligned.");

hotspot/src/os/windows/vm/os_windows.cpp

@@ -2803,6 +2803,14 @@ bool os::release_memory(char* addr, size_t bytes) {
   return VirtualFree(addr, 0, MEM_RELEASE) != 0;
 }
 
+bool os::create_stack_guard_pages(char* addr, size_t size) {
+  return os::commit_memory(addr, size);
+}
+
+bool os::remove_stack_guard_pages(char* addr, size_t size) {
+  return os::uncommit_memory(addr, size);
+}
+
 // Set protections specified
 bool os::protect_memory(char* addr, size_t bytes, ProtType prot,
                         bool is_committed) {

hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp

@@ -760,7 +760,10 @@ void ConcurrentMark::checkpointRootsInitialPost() {
   rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
 
   SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
-  satb_mq_set.set_active_all_threads(true);
+  // This is the start of  the marking cycle, we're expected all
+  // threads to have SATB queues with active set to false.
+  satb_mq_set.set_active_all_threads(true, /* new active value */
+                                     false /* expected_active */);
 
   // update_g1_committed() will be called at the end of an evac pause
   // when marking is on. So, it's also called at the end of the
@@ -1079,7 +1082,11 @@ void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
       gclog_or_tty->print_cr("\nRemark led to restart for overflow.");
   } else {
     // We're done with marking.
-    JavaThread::satb_mark_queue_set().set_active_all_threads(false);
+    // This is the end of  the marking cycle, we're expected all
+    // threads to have SATB queues with active set to true.
+    JavaThread::satb_mark_queue_set().set_active_all_threads(
+                                                  false, /* new active value */
+                                                  true /* expected_active */);
 
     if (VerifyDuringGC) {
       HandleMark hm;  // handle scope
@@ -2586,7 +2593,11 @@ void ConcurrentMark::abort() {
 
   SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
   satb_mq_set.abandon_partial_marking();
-  satb_mq_set.set_active_all_threads(false);
+  // This can be called either during or outside marking, we'll read
+  // the expected_active value from the SATB queue set.
+  satb_mq_set.set_active_all_threads(
+                                 false, /* new active value */
+                                 satb_mq_set.is_active() /* expected_active */);
 }
 
 static void print_ms_time_info(const char* prefix, const char* name,
@@ -3704,7 +3715,14 @@ void CMTask::do_marking_step(double time_target_ms) {
         // enough to point to the next possible object header (the
         // bitmap knows by how much we need to move it as it knows its
         // granularity).
-        move_finger_to(_nextMarkBitMap->nextWord(_finger));
+        assert(_finger < _region_limit, "invariant");
+        HeapWord* new_finger = _nextMarkBitMap->nextWord(_finger);
+        // Check if bitmap iteration was aborted while scanning the last object
+        if (new_finger >= _region_limit) {
+            giveup_current_region();
+        } else {
+            move_finger_to(new_finger);
+        }
       }
     }
     // At this point we have either completed iterating over the

hotspot/src/share/vm/gc_implementation/g1/concurrentMark.hpp

@@ -25,7 +25,7 @@
 class G1CollectedHeap;
 class CMTask;
 typedef GenericTaskQueue<oop>            CMTaskQueue;
-typedef GenericTaskQueueSet<oop> CMTaskQueueSet;
+typedef GenericTaskQueueSet<CMTaskQueue> CMTaskQueueSet;
 
 // A generic CM bit map.  This is essentially a wrapper around the BitMap
 // class, with one bit per (1<<_shifter) HeapWords.

hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp

@@ -2102,18 +2102,21 @@ size_t G1CollectedHeap::tlab_capacity(Thread* ignored) const {
 size_t G1CollectedHeap::unsafe_max_tlab_alloc(Thread* ignored) const {
   // Return the remaining space in the cur alloc region, but not less than
   // the min TLAB size.
-  // Also, no more than half the region size, since we can't allow tlabs to
-  // grow big enough to accomodate humongous objects.
 
-  // We need to story it locally, since it might change between when we
+  // Also, this value can be at most the humongous object threshold,
-  // test for NULL and when we use it later.
+  // since we can't allow tlabs to grow big enough to accomodate
+  // humongous objects.
+
+  // We need to store the cur alloc region locally, since it might change
+  // between when we test for NULL and when we use it later.
   ContiguousSpace* cur_alloc_space = _cur_alloc_region;
+  size_t max_tlab_size = _humongous_object_threshold_in_words * wordSize;
+
   if (cur_alloc_space == NULL) {
-    return HeapRegion::GrainBytes/2;
+    return max_tlab_size;
   } else {
-    return MAX2(MIN2(cur_alloc_space->free(),
+    return MIN2(MAX2(cur_alloc_space->free(), (size_t)MinTLABSize),
-                     (size_t)(HeapRegion::GrainBytes/2)),
+                max_tlab_size);
-                (size_t)MinTLABSize);
   }
 }
 

hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp

@@ -57,7 +57,7 @@ class ConcurrentZFThread;
 #endif
 
 typedef GenericTaskQueue<StarTask>          RefToScanQueue;
-typedef GenericTaskQueueSet<StarTask> RefToScanQueueSet;
+typedef GenericTaskQueueSet<RefToScanQueue> RefToScanQueueSet;
 
 typedef int RegionIdx_t;   // needs to hold [ 0..max_regions() )
 typedef int CardIdx_t;     // needs to hold [ 0..CardsPerRegion )
@@ -1055,7 +1055,12 @@ public:
 
   // Returns "true" iff the given word_size is "very large".
   static bool isHumongous(size_t word_size) {
-    return word_size >= _humongous_object_threshold_in_words;
+    // Note this has to be strictly greater-than as the TLABs
+    // are capped at the humongous thresold and we want to
+    // ensure that we don't try to allocate a TLAB as
+    // humongous and that we don't allocate a humongous
+    // object in a TLAB.
+    return word_size > _humongous_object_threshold_in_words;
   }
 
   // Update mod union table with the set of dirty cards.

hotspot/src/share/vm/gc_implementation/g1/g1MarkSweep.cpp

@@ -101,6 +101,8 @@ void G1MarkSweep::allocate_stacks() {
 
   GenMarkSweep::_marking_stack =
     new (ResourceObj::C_HEAP) GrowableArray<oop>(4000, true);
+  GenMarkSweep::_objarray_stack =
+    new (ResourceObj::C_HEAP) GrowableArray<ObjArrayTask>(50, true);
 
   int size = SystemDictionary::number_of_classes() * 2;
   GenMarkSweep::_revisit_klass_stack =

hotspot/src/share/vm/gc_implementation/g1/g1SATBCardTableModRefBS.cpp

@@ -35,7 +35,7 @@ G1SATBCardTableModRefBS::G1SATBCardTableModRefBS(MemRegion whole_heap,
 
 void G1SATBCardTableModRefBS::enqueue(oop pre_val) {
   assert(pre_val->is_oop_or_null(true), "Error");
-  if (!JavaThread::satb_mark_queue_set().active()) return;
+  if (!JavaThread::satb_mark_queue_set().is_active()) return;
   Thread* thr = Thread::current();
   if (thr->is_Java_thread()) {
     JavaThread* jt = (JavaThread*)thr;
@@ -51,7 +51,7 @@ template <class T> void
 G1SATBCardTableModRefBS::write_ref_field_pre_static(T* field,
                                                     oop new_val,
                                                     JavaThread* jt) {
-  if (!JavaThread::satb_mark_queue_set().active()) return;
+  if (!JavaThread::satb_mark_queue_set().is_active()) return;
   T heap_oop = oopDesc::load_heap_oop(field);
   if (!oopDesc::is_null(heap_oop)) {
     oop pre_val = oopDesc::decode_heap_oop_not_null(heap_oop);
@@ -62,7 +62,7 @@ G1SATBCardTableModRefBS::write_ref_field_pre_static(T* field,
 
 template <class T> void
 G1SATBCardTableModRefBS::write_ref_array_pre_work(T* dst, int count) {
-  if (!JavaThread::satb_mark_queue_set().active()) return;
+  if (!JavaThread::satb_mark_queue_set().is_active()) return;
   T* elem_ptr = dst;
   for (int i = 0; i < count; i++, elem_ptr++) {
     T heap_oop = oopDesc::load_heap_oop(elem_ptr);

hotspot/src/share/vm/gc_implementation/g1/ptrQueue.cpp

@@ -25,8 +25,8 @@
 # include "incls/_precompiled.incl"
 # include "incls/_ptrQueue.cpp.incl"
 
-PtrQueue::PtrQueue(PtrQueueSet* qset_, bool perm) :
+PtrQueue::PtrQueue(PtrQueueSet* qset_, bool perm, bool active) :
-  _qset(qset_), _buf(NULL), _index(0), _active(false),
+  _qset(qset_), _buf(NULL), _index(0), _active(active),
   _perm(perm), _lock(NULL)
 {}
 

hotspot/src/share/vm/gc_implementation/g1/ptrQueue.hpp

@@ -62,7 +62,7 @@ protected:
 public:
   // Initialize this queue to contain a null buffer, and be part of the
   // given PtrQueueSet.
-  PtrQueue(PtrQueueSet*, bool perm = false);
+  PtrQueue(PtrQueueSet*, bool perm = false, bool active = false);
   // Release any contained resources.
   void flush();
   // Calls flush() when destroyed.
@@ -101,6 +101,8 @@ public:
     }
   }
 
+  bool is_active() { return _active; }
+
   static int byte_index_to_index(int ind) {
     assert((ind % oopSize) == 0, "Invariant.");
     return ind / oopSize;
@@ -257,7 +259,7 @@ public:
   bool process_completed_buffers() { return _process_completed; }
   void set_process_completed(bool x) { _process_completed = x; }
 
-  bool active() { return _all_active; }
+  bool is_active() { return _all_active; }
 
   // Set the buffer size.  Should be called before any "enqueue" operation
   // can be called.  And should only be called once.

hotspot/src/share/vm/gc_implementation/g1/satbQueue.cpp

@@ -82,9 +82,57 @@ void SATBMarkQueueSet::handle_zero_index_for_thread(JavaThread* t) {
   t->satb_mark_queue().handle_zero_index();
 }
 
-void SATBMarkQueueSet::set_active_all_threads(bool b) {
+#ifdef ASSERT
+void SATBMarkQueueSet::dump_active_values(JavaThread* first,
+                                          bool expected_active) {
+  gclog_or_tty->print_cr("SATB queue active values for Java Threads");
+  gclog_or_tty->print_cr(" SATB queue set: active is %s",
+                         (is_active()) ? "TRUE" : "FALSE");
+  gclog_or_tty->print_cr(" expected_active is %s",
+                         (expected_active) ? "TRUE" : "FALSE");
+  for (JavaThread* t = first; t; t = t->next()) {
+    bool active = t->satb_mark_queue().is_active();
+    gclog_or_tty->print_cr("  thread %s, active is %s",
+                           t->name(), (active) ? "TRUE" : "FALSE");
+  }
+}
+#endif // ASSERT
+
+void SATBMarkQueueSet::set_active_all_threads(bool b,
+                                              bool expected_active) {
+  assert(SafepointSynchronize::is_at_safepoint(), "Must be at safepoint.");
+  JavaThread* first = Threads::first();
+
+#ifdef ASSERT
+  if (_all_active != expected_active) {
+    dump_active_values(first, expected_active);
+
+    // I leave this here as a guarantee, instead of an assert, so
+    // that it will still be compiled in if we choose to uncomment
+    // the #ifdef ASSERT in a product build. The whole block is
+    // within an #ifdef ASSERT so the guarantee will not be compiled
+    // in a product build anyway.
+    guarantee(false,
+              "SATB queue set has an unexpected active value");
+  }
+#endif // ASSERT
   _all_active = b;
-  for(JavaThread* t = Threads::first(); t; t = t->next()) {
+
+  for (JavaThread* t = first; t; t = t->next()) {
+#ifdef ASSERT
+    bool active = t->satb_mark_queue().is_active();
+    if (active != expected_active) {
+      dump_active_values(first, expected_active);
+
+      // I leave this here as a guarantee, instead of an assert, so
+      // that it will still be compiled in if we choose to uncomment
+      // the #ifdef ASSERT in a product build. The whole block is
+      // within an #ifdef ASSERT so the guarantee will not be compiled
+      // in a product build anyway.
+      guarantee(false,
+                "thread has an unexpected active value in its SATB queue");
+    }
+#endif // ASSERT
     t->satb_mark_queue().set_active(b);
   }
 }

hotspot/src/share/vm/gc_implementation/g1/satbQueue.hpp

@@ -29,8 +29,7 @@ class JavaThread;
 class ObjPtrQueue: public PtrQueue {
 public:
   ObjPtrQueue(PtrQueueSet* qset_, bool perm = false) :
-    PtrQueue(qset_, perm)
+    PtrQueue(qset_, perm, qset_->is_active()) { }
-  {}
   // Apply the closure to all elements, and reset the index to make the
   // buffer empty.
   void apply_closure(ObjectClosure* cl);
@@ -55,6 +54,9 @@ class SATBMarkQueueSet: public PtrQueueSet {
   // is ignored.
   bool apply_closure_to_completed_buffer_work(bool par, int worker);
 
+#ifdef ASSERT
+  void dump_active_values(JavaThread* first, bool expected_active);
+#endif // ASSERT
 
 public:
   SATBMarkQueueSet();
@@ -65,9 +67,11 @@ public:
 
   static void handle_zero_index_for_thread(JavaThread* t);
 
-  // Apply "set_active(b)" to all thread tloq's.  Should be called only
+  // Apply "set_active(b)" to all Java threads' SATB queues. It should be
-  // with the world stopped.
+  // called only with the world stopped. The method will assert that the
-  void set_active_all_threads(bool b);
+  // SATB queues of all threads it visits, as well as the SATB queue
+  // set itself, has an active value same as expected_active.
+  void set_active_all_threads(bool b, bool expected_active);
 
   // Register "blk" as "the closure" for all queues.  Only one such closure
   // is allowed.  The "apply_closure_to_completed_buffer" method will apply

hotspot/src/share/vm/gc_implementation/includeDB_gc_parallelScavenge

@@ -175,6 +175,7 @@ psAdaptiveSizePolicy.hpp                gcUtil.hpp
 psAdaptiveSizePolicy.hpp		adaptiveSizePolicy.hpp
 
 psCompactionManager.cpp                 gcTaskManager.hpp
+psCompactionManager.cpp                 objArrayKlass.inline.hpp
 psCompactionManager.cpp                 objectStartArray.hpp
 psCompactionManager.cpp                 oop.hpp
 psCompactionManager.cpp                 oop.inline.hpp
@@ -189,6 +190,9 @@ psCompactionManager.cpp                 systemDictionary.hpp
 psCompactionManager.hpp                 allocation.hpp
 psCompactionManager.hpp                 taskqueue.hpp
 
+psCompactionManager.inline.hpp		psCompactionManager.hpp
+psCompactionManager.inline.hpp		psParallelCompact.hpp
+
 psGCAdaptivePolicyCounters.hpp		gcAdaptivePolicyCounters.hpp
 psGCAdaptivePolicyCounters.hpp          gcPolicyCounters.hpp
 psGCAdaptivePolicyCounters.hpp          psAdaptiveSizePolicy.hpp
@@ -379,12 +383,12 @@ pcTasks.cpp                             fprofiler.hpp
 pcTasks.cpp                             jniHandles.hpp
 pcTasks.cpp                             jvmtiExport.hpp
 pcTasks.cpp                             management.hpp
+pcTasks.cpp                             objArrayKlass.inline.hpp
 pcTasks.cpp                             psParallelCompact.hpp
 pcTasks.cpp                             pcTasks.hpp
 pcTasks.cpp                             oop.inline.hpp
 pcTasks.cpp                             oop.pcgc.inline.hpp
 pcTasks.cpp                             systemDictionary.hpp
-pcTasks.cpp                             taskqueue.hpp
 pcTasks.cpp                             thread.hpp
 pcTasks.cpp                             universe.hpp
 pcTasks.cpp                             vmThread.hpp

hotspot/src/share/vm/gc_implementation/parallelScavenge/pcTasks.cpp

@@ -48,7 +48,7 @@ void ThreadRootsMarkingTask::do_it(GCTaskManager* manager, uint which) {
     _vm_thread->oops_do(&mark_and_push_closure, &mark_and_push_in_blobs);
 
   // Do the real work
-  cm->drain_marking_stacks(&mark_and_push_closure);
+  cm->follow_marking_stacks();
 }
 
 
@@ -118,7 +118,7 @@ void MarkFromRootsTask::do_it(GCTaskManager* manager, uint which) {
   }
 
   // Do the real work
-  cm->drain_marking_stacks(&mark_and_push_closure);
+  cm->follow_marking_stacks();
   // cm->deallocate_stacks();
 }
 
@@ -196,17 +196,19 @@ void StealMarkingTask::do_it(GCTaskManager* manager, uint which) {
   PSParallelCompact::MarkAndPushClosure mark_and_push_closure(cm);
 
   oop obj = NULL;
+  ObjArrayTask task;
   int random_seed = 17;
-  while(true) {
+  do {
-    if (ParCompactionManager::steal(which, &random_seed, obj)) {
+    while (ParCompactionManager::steal_objarray(which, &random_seed, task)) {
+      objArrayKlass* const k = (objArrayKlass*)task.obj()->blueprint();
+      k->oop_follow_contents(cm, task.obj(), task.index());
+      cm->follow_marking_stacks();
+    }
+    while (ParCompactionManager::steal(which, &random_seed, obj)) {
       obj->follow_contents(cm);
-      cm->drain_marking_stacks(&mark_and_push_closure);
+      cm->follow_marking_stacks();
-    } else {
-      if (terminator()->offer_termination()) {
-        break;
-      }
-    }
     }
+  } while (!terminator()->offer_termination());
 }
 
 //

hotspot/src/share/vm/gc_implementation/parallelScavenge/psCompactionManager.cpp

@@ -28,6 +28,8 @@
 PSOldGen*            ParCompactionManager::_old_gen = NULL;
 ParCompactionManager**  ParCompactionManager::_manager_array = NULL;
 OopTaskQueueSet*     ParCompactionManager::_stack_array = NULL;
+ParCompactionManager::ObjArrayTaskQueueSet*
+  ParCompactionManager::_objarray_queues = NULL;
 ObjectStartArray*    ParCompactionManager::_start_array = NULL;
 ParMarkBitMap*       ParCompactionManager::_mark_bitmap = NULL;
 RegionTaskQueueSet*   ParCompactionManager::_region_array = NULL;
@@ -46,6 +48,11 @@ ParCompactionManager::ParCompactionManager() :
 
   // We want the overflow stack to be permanent
   _overflow_stack = new (ResourceObj::C_HEAP) GrowableArray<oop>(10, true);
+
+  _objarray_queue.initialize();
+  _objarray_overflow_stack =
+    new (ResourceObj::C_HEAP) ObjArrayOverflowStack(10, true);
+
 #ifdef USE_RegionTaskQueueWithOverflow
   region_stack()->initialize();
 #else
@@ -69,6 +76,7 @@ ParCompactionManager::ParCompactionManager() :
 
 ParCompactionManager::~ParCompactionManager() {
   delete _overflow_stack;
+  delete _objarray_overflow_stack;
   delete _revisit_klass_stack;
   delete _revisit_mdo_stack;
   // _manager_array and _stack_array are statics
@@ -86,18 +94,21 @@ void ParCompactionManager::initialize(ParMarkBitMap* mbm) {
 
   assert(_manager_array == NULL, "Attempt to initialize twice");
   _manager_array = NEW_C_HEAP_ARRAY(ParCompactionManager*, parallel_gc_threads+1 );
-  guarantee(_manager_array != NULL, "Could not initialize promotion manager");
+  guarantee(_manager_array != NULL, "Could not allocate manager_array");
 
   _stack_array = new OopTaskQueueSet(parallel_gc_threads);
-  guarantee(_stack_array != NULL, "Count not initialize promotion manager");
+  guarantee(_stack_array != NULL, "Could not allocate stack_array");
+  _objarray_queues = new ObjArrayTaskQueueSet(parallel_gc_threads);
+  guarantee(_objarray_queues != NULL, "Could not allocate objarray_queues");
   _region_array = new RegionTaskQueueSet(parallel_gc_threads);
-  guarantee(_region_array != NULL, "Count not initialize promotion manager");
+  guarantee(_region_array != NULL, "Could not allocate region_array");
 
   // Create and register the ParCompactionManager(s) for the worker threads.
   for(uint i=0; i<parallel_gc_threads; i++) {
     _manager_array[i] = new ParCompactionManager();
     guarantee(_manager_array[i] != NULL, "Could not create ParCompactionManager");
     stack_array()->register_queue(i, _manager_array[i]->marking_stack());
+    _objarray_queues->register_queue(i, &_manager_array[i]->_objarray_queue);
 #ifdef USE_RegionTaskQueueWithOverflow
     region_array()->register_queue(i, _manager_array[i]->region_stack()->task_queue());
 #else
@@ -203,36 +214,30 @@ void ParCompactionManager::reset() {
   }
 }
 
-void ParCompactionManager::drain_marking_stacks(OopClosure* blk) {
+void ParCompactionManager::follow_marking_stacks() {
-#ifdef ASSERT
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-  MutableSpace* to_space = heap->young_gen()->to_space();
-  MutableSpace* old_space = heap->old_gen()->object_space();
-  MutableSpace* perm_space = heap->perm_gen()->object_space();
-#endif /* ASSERT */
-
-
   do {
-
+    // Drain the overflow stack first, to allow stealing from the marking stack.
-    // Drain overflow stack first, so other threads can steal from
-    // claimed stack while we work.
-    while(!overflow_stack()->is_empty()) {
-      oop obj = overflow_stack()->pop();
-      obj->follow_contents(this);
-    }
-
     oop obj;
-    // obj is a reference!!!
+    while (!overflow_stack()->is_empty()) {
+      overflow_stack()->pop()->follow_contents(this);
+    }
     while (marking_stack()->pop_local(obj)) {
-      // It would be nice to assert about the type of objects we might
-      // pop, but they can come from anywhere, unfortunately.
       obj->follow_contents(this);
     }
-  } while((marking_stack()->size() != 0) || (overflow_stack()->length() != 0));
 
-  assert(marking_stack()->size() == 0, "Sanity");
+    // Process ObjArrays one at a time to avoid marking stack bloat.
-  assert(overflow_stack()->length() == 0, "Sanity");
+    ObjArrayTask task;
+    if (!_objarray_overflow_stack->is_empty()) {
+      task = _objarray_overflow_stack->pop();
+      objArrayKlass* const k = (objArrayKlass*)task.obj()->blueprint();
+      k->oop_follow_contents(this, task.obj(), task.index());
+    } else if (_objarray_queue.pop_local(task)) {
+      objArrayKlass* const k = (objArrayKlass*)task.obj()->blueprint();
+      k->oop_follow_contents(this, task.obj(), task.index());
+    }
+  } while (!marking_stacks_empty());
+
+  assert(marking_stacks_empty(), "Sanity");
 }
 
 void ParCompactionManager::drain_region_overflow_stack() {

hotspot/src/share/vm/gc_implementation/parallelScavenge/psCompactionManager.hpp

@@ -22,18 +22,6 @@
  *
  */
 
-//
-// psPromotionManager is used by a single thread to manage object survival
-// during a scavenge. The promotion manager contains thread local data only.
-//
-// NOTE! Be carefull when allocating the stacks on cheap. If you are going
-// to use a promotion manager in more than one thread, the stacks MUST be
-// on cheap. This can lead to memory leaks, though, as they are not auto
-// deallocated.
-//
-// FIX ME FIX ME Add a destructor, and don't rely on the user to drain/flush/deallocate!
-//
-
 // Move to some global location
 #define HAS_BEEN_MOVED 0x1501d01d
 // End move to some global location
@@ -46,8 +34,6 @@ class ObjectStartArray;
 class ParallelCompactData;
 class ParMarkBitMap;
 
-// Move to it's own file if this works out.
-
 class ParCompactionManager : public CHeapObj {
   friend class ParallelTaskTerminator;
   friend class ParMarkBitMap;
@@ -72,14 +58,27 @@ class ParCompactionManager : public CHeapObj {
 // ------------------------  End don't putback if not needed
 
  private:
+  // 32-bit:  4K * 8 = 32KiB; 64-bit:  8K * 16 = 128KiB
+  #define OBJARRAY_QUEUE_SIZE (1 << NOT_LP64(12) LP64_ONLY(13))
+  typedef GenericTaskQueue<ObjArrayTask, OBJARRAY_QUEUE_SIZE> ObjArrayTaskQueue;
+  typedef GenericTaskQueueSet<ObjArrayTaskQueue> ObjArrayTaskQueueSet;
+  #undef OBJARRAY_QUEUE_SIZE
+
   static ParCompactionManager** _manager_array;
   static OopTaskQueueSet*       _stack_array;
+  static ObjArrayTaskQueueSet*  _objarray_queues;
   static ObjectStartArray*      _start_array;
   static RegionTaskQueueSet*    _region_array;
   static PSOldGen*              _old_gen;
 
+private:
   OopTaskQueue                  _marking_stack;
   GrowableArray<oop>*           _overflow_stack;
+
+  typedef GrowableArray<ObjArrayTask> ObjArrayOverflowStack;
+  ObjArrayTaskQueue             _objarray_queue;
+  ObjArrayOverflowStack*        _objarray_overflow_stack;
+
   // Is there a way to reuse the _marking_stack for the
   // saving empty regions?  For now just create a different
   // type of TaskQueue.
@@ -128,8 +127,8 @@ class ParCompactionManager : public CHeapObj {
   // Pushes onto the region stack.  If the region stack is full,
   // pushes onto the region overflow stack.
   void region_stack_push(size_t region_index);
- public:
 
+public:
   Action action() { return _action; }
   void set_action(Action v) { _action = v; }
 
@@ -163,6 +162,8 @@ class ParCompactionManager : public CHeapObj {
   // Get a oop for scanning.  If returns null, no oop were found.
   oop retrieve_for_scanning();
 
+  inline void push_objarray(oop obj, size_t index);
+
   // Save region for later processing.  Must not fail.
   void save_for_processing(size_t region_index);
   // Get a region for processing.  If returns null, no region were found.
@@ -175,12 +176,17 @@ class ParCompactionManager : public CHeapObj {
     return stack_array()->steal(queue_num, seed, t);
   }
 
+  static bool steal_objarray(int queue_num, int* seed, ObjArrayTask& t) {
+    return _objarray_queues->steal(queue_num, seed, t);
+  }
+
   static bool steal(int queue_num, int* seed, RegionTask& t) {
     return region_array()->steal(queue_num, seed, t);
   }
 
-  // Process tasks remaining on any stack
+  // Process tasks remaining on any marking stack
-  void drain_marking_stacks(OopClosure *blk);
+  void follow_marking_stacks();
+  inline bool marking_stacks_empty() const;
 
   // Process tasks remaining on any stack
   void drain_region_stacks();
@@ -200,3 +206,8 @@ inline ParCompactionManager* ParCompactionManager::manager_array(int index) {
     "out of range manager_array access");
   return _manager_array[index];
 }
+
+bool ParCompactionManager::marking_stacks_empty() const {
+  return _marking_stack.size() == 0 && _overflow_stack->is_empty() &&
+    _objarray_queue.size() == 0 && _objarray_overflow_stack->is_empty();
+}

hotspot/src/share/vm/gc_implementation/parallelScavenge/psCompactionManager.inline.hpp

@@ -0,0 +1,32 @@
+/*
+ * Copyright 2010 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ */
+
+void ParCompactionManager::push_objarray(oop obj, size_t index)
+{
+  ObjArrayTask task(obj, index);
+  assert(task.is_valid(), "bad ObjArrayTask");
+  if (!_objarray_queue.push(task)) {
+    _objarray_overflow_stack->push(task);
+  }
+}

hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp

@@ -479,6 +479,7 @@ void PSMarkSweep::allocate_stacks() {
   _preserved_oop_stack = NULL;
 
   _marking_stack = new (ResourceObj::C_HEAP) GrowableArray<oop>(4000, true);
+  _objarray_stack = new (ResourceObj::C_HEAP) GrowableArray<ObjArrayTask>(50, true);
 
   int size = SystemDictionary::number_of_classes() * 2;
   _revisit_klass_stack = new (ResourceObj::C_HEAP) GrowableArray<Klass*>(size, true);
@@ -497,6 +498,7 @@ void PSMarkSweep::deallocate_stacks() {
   }
 
   delete _marking_stack;
+  delete _objarray_stack;
   delete _revisit_klass_stack;
   delete _revisit_mdo_stack;
 }

hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp

@@ -785,7 +785,7 @@ PSParallelCompact::AdjustPointerClosure PSParallelCompact::_adjust_pointer_closu
 void PSParallelCompact::AdjustPointerClosure::do_oop(oop* p)       { adjust_pointer(p, _is_root); }
 void PSParallelCompact::AdjustPointerClosure::do_oop(narrowOop* p) { adjust_pointer(p, _is_root); }
 
-void PSParallelCompact::FollowStackClosure::do_void() { follow_stack(_compaction_manager); }
+void PSParallelCompact::FollowStackClosure::do_void() { _compaction_manager->follow_marking_stacks(); }
 
 void PSParallelCompact::MarkAndPushClosure::do_oop(oop* p)       { mark_and_push(_compaction_manager, p); }
 void PSParallelCompact::MarkAndPushClosure::do_oop(narrowOop* p) { mark_and_push(_compaction_manager, p); }
@@ -2376,7 +2376,7 @@ void PSParallelCompact::marking_phase(ParCompactionManager* cm,
   // Follow code cache roots.
   CodeCache::do_unloading(is_alive_closure(), &mark_and_push_closure,
                           purged_class);
-  follow_stack(cm); // Flush marking stack.
+  cm->follow_marking_stacks(); // Flush marking stack.
 
   // Update subklass/sibling/implementor links of live klasses
   // revisit_klass_stack is used in follow_weak_klass_links().
@@ -2389,8 +2389,7 @@ void PSParallelCompact::marking_phase(ParCompactionManager* cm,
   SymbolTable::unlink(is_alive_closure());
   StringTable::unlink(is_alive_closure());
 
-  assert(cm->marking_stack()->size() == 0, "stack should be empty by now");
+  assert(cm->marking_stacks_empty(), "marking stacks should be empty");
-  assert(cm->overflow_stack()->is_empty(), "stack should be empty by now");
 }
 
 // This should be moved to the shared markSweep code!
@@ -2709,22 +2708,6 @@ void PSParallelCompact::compact_serial(ParCompactionManager* cm) {
   young_gen->move_and_update(cm);
 }
 
-
-void PSParallelCompact::follow_stack(ParCompactionManager* cm) {
-  while(!cm->overflow_stack()->is_empty()) {
-    oop obj = cm->overflow_stack()->pop();
-    obj->follow_contents(cm);
-  }
-
-  oop obj;
-  // obj is a reference!!!
-  while (cm->marking_stack()->pop_local(obj)) {
-    // It would be nice to assert about the type of objects we might
-    // pop, but they can come from anywhere, unfortunately.
-    obj->follow_contents(cm);
-  }
-}
-
 void
 PSParallelCompact::follow_weak_klass_links() {
   // All klasses on the revisit stack are marked at this point.
@@ -2745,7 +2728,7 @@ PSParallelCompact::follow_weak_klass_links() {
         &keep_alive_closure);
     }
     // revisit_klass_stack is cleared in reset()
-    follow_stack(cm);
+    cm->follow_marking_stacks();
   }
 }
 
@@ -2776,7 +2759,7 @@ void PSParallelCompact::follow_mdo_weak_refs() {
       rms->at(j)->follow_weak_refs(is_alive_closure());
     }
     // revisit_mdo_stack is cleared in reset()
-    follow_stack(cm);
+    cm->follow_marking_stacks();
   }
 }
 

hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.hpp

@@ -901,7 +901,6 @@ class PSParallelCompact : AllStatic {
   // Mark live objects
   static void marking_phase(ParCompactionManager* cm,
                             bool maximum_heap_compaction);
-  static void follow_stack(ParCompactionManager* cm);
   static void follow_weak_klass_links();
   static void follow_mdo_weak_refs();
 
@@ -1276,7 +1275,7 @@ inline void PSParallelCompact::follow_root(ParCompactionManager* cm, T* p) {
       }
     }
   }
-  follow_stack(cm);
+  cm->follow_marking_stacks();
 }
 
 template <class T>

hotspot/src/share/vm/gc_implementation/shared/markSweep.cpp

@@ -26,6 +26,7 @@
 #include "incls/_markSweep.cpp.incl"
 
 GrowableArray<oop>*          MarkSweep::_marking_stack = NULL;
+GrowableArray<ObjArrayTask>* MarkSweep::_objarray_stack = NULL;
 GrowableArray<Klass*>*       MarkSweep::_revisit_klass_stack = NULL;
 GrowableArray<DataLayout*>*  MarkSweep::_revisit_mdo_stack = NULL;
 
@@ -104,11 +105,19 @@ void MarkSweep::MarkAndPushClosure::do_oop(oop* p)       { mark_and_push(p); }
 void MarkSweep::MarkAndPushClosure::do_oop(narrowOop* p) { mark_and_push(p); }
 
 void MarkSweep::follow_stack() {
+  do {
     while (!_marking_stack->is_empty()) {
       oop obj = _marking_stack->pop();
       assert (obj->is_gc_marked(), "p must be marked");
       obj->follow_contents();
     }
+    // Process ObjArrays one at a time to avoid marking stack bloat.
+    if (!_objarray_stack->is_empty()) {
+      ObjArrayTask task = _objarray_stack->pop();
+      objArrayKlass* const k = (objArrayKlass*)task.obj()->blueprint();
+      k->oop_follow_contents(task.obj(), task.index());
+    }
+  } while (!_marking_stack->is_empty() || !_objarray_stack->is_empty());
 }
 
 MarkSweep::FollowStackClosure MarkSweep::follow_stack_closure;

hotspot/src/share/vm/gc_implementation/shared/markSweep.hpp

@@ -110,8 +110,9 @@ class MarkSweep : AllStatic {
   // Vars
   //
  protected:
-  // Traversal stack used during phase1
+  // Traversal stacks used during phase1
   static GrowableArray<oop>*             _marking_stack;
+  static GrowableArray<ObjArrayTask>*    _objarray_stack;
   // Stack for live klasses to revisit at end of marking phase
   static GrowableArray<Klass*>*          _revisit_klass_stack;
   // Set (stack) of MDO's to revisit at end of marking phase
@@ -188,6 +189,7 @@ class MarkSweep : AllStatic {
   template <class T> static inline void mark_and_follow(T* p);
   // Check mark and maybe push on marking stack
   template <class T> static inline void mark_and_push(T* p);
+  static inline void push_objarray(oop obj, size_t index);
 
   static void follow_stack();   // Empty marking stack.
 

hotspot/src/share/vm/gc_implementation/shared/markSweep.inline.hpp

@@ -77,6 +77,12 @@ template <class T> inline void MarkSweep::mark_and_push(T* p) {
   }
 }
 
+void MarkSweep::push_objarray(oop obj, size_t index) {
+  ObjArrayTask task(obj, index);
+  assert(task.is_valid(), "bad ObjArrayTask");
+  _objarray_stack->push(task);
+}
+
 template <class T> inline void MarkSweep::adjust_pointer(T* p, bool isroot) {
   T heap_oop = oopDesc::load_heap_oop(p);
   if (!oopDesc::is_null(heap_oop)) {

hotspot/src/share/vm/includeDB_core

@@ -2026,6 +2026,7 @@ init.hpp                                top.hpp
 
 instanceKlass.cpp                       collectedHeap.inline.hpp
 instanceKlass.cpp                       compileBroker.hpp
+instanceKlass.cpp                       dtrace.hpp
 instanceKlass.cpp                       fieldDescriptor.hpp
 instanceKlass.cpp                       genOopClosures.inline.hpp
 instanceKlass.cpp                       handles.inline.hpp
@@ -2485,6 +2486,7 @@ jvm.cpp                                 classLoader.hpp
 jvm.cpp                                 collectedHeap.inline.hpp
 jvm.cpp                                 copy.hpp
 jvm.cpp                                 defaultStream.hpp
+jvm.cpp                                 dtrace.hpp
 jvm.cpp                                 dtraceJSDT.hpp
 jvm.cpp                                 events.hpp
 jvm.cpp                                 handles.inline.hpp
@@ -2726,8 +2728,10 @@ markOop.inline.hpp                      markOop.hpp
 
 markSweep.cpp                           compileBroker.hpp
 markSweep.cpp                           methodDataOop.hpp
+markSweep.cpp				objArrayKlass.inline.hpp
 
 markSweep.hpp                           collectedHeap.hpp
+markSweep.hpp				taskqueue.hpp
 
 memRegion.cpp                           globals.hpp
 memRegion.cpp                           memRegion.hpp
@@ -3057,8 +3061,10 @@ objArrayKlass.cpp                       copy.hpp
 objArrayKlass.cpp                       genOopClosures.inline.hpp
 objArrayKlass.cpp                       handles.inline.hpp
 objArrayKlass.cpp                       instanceKlass.hpp
+objArrayKlass.cpp                       markSweep.inline.hpp
 objArrayKlass.cpp                       mutexLocker.hpp
 objArrayKlass.cpp                       objArrayKlass.hpp
+objArrayKlass.cpp                       objArrayKlass.inline.hpp
 objArrayKlass.cpp                       objArrayKlassKlass.hpp
 objArrayKlass.cpp                       objArrayOop.hpp
 objArrayKlass.cpp                       oop.inline.hpp
@@ -3069,11 +3075,12 @@ objArrayKlass.cpp                       systemDictionary.hpp
 objArrayKlass.cpp                       universe.inline.hpp
 objArrayKlass.cpp                       vmSymbols.hpp
 
-
 objArrayKlass.hpp                       arrayKlass.hpp
 objArrayKlass.hpp                       instanceKlass.hpp
 objArrayKlass.hpp                       specialized_oop_closures.hpp
 
+objArrayKlass.inline.hpp		objArrayKlass.hpp
+
 objArrayKlassKlass.cpp                  collectedHeap.inline.hpp
 objArrayKlassKlass.cpp                  instanceKlass.hpp
 objArrayKlassKlass.cpp                  javaClasses.hpp
@@ -4099,6 +4106,7 @@ task.cpp                                timer.hpp
 task.hpp                                top.hpp
 
 taskqueue.cpp                           debug.hpp
+taskqueue.cpp				oop.inline.hpp
 taskqueue.cpp                           os.hpp
 taskqueue.cpp                           taskqueue.hpp
 taskqueue.cpp                           thread_<os_family>.inline.hpp
@@ -4452,6 +4460,7 @@ universe.inline.hpp                     universe.hpp
 
 unsafe.cpp                              allocation.inline.hpp
 unsafe.cpp                              copy.hpp
+unsafe.cpp                              dtrace.hpp
 unsafe.cpp                              globals.hpp
 unsafe.cpp                              interfaceSupport.hpp
 unsafe.cpp                              jni.h
@@ -4623,6 +4632,7 @@ vmSymbols.hpp                           symbolOop.hpp
 
 vmThread.cpp                            collectedHeap.hpp
 vmThread.cpp                            compileBroker.hpp
+vmThread.cpp                            dtrace.hpp
 vmThread.cpp                            events.hpp
 vmThread.cpp                            interfaceSupport.hpp
 vmThread.cpp                            methodOop.hpp

hotspot/src/share/vm/includeDB_gc_parallel

@@ -115,10 +115,14 @@ objArrayKlass.cpp                       heapRegionSeq.inline.hpp
 objArrayKlass.cpp                       g1CollectedHeap.inline.hpp
 objArrayKlass.cpp                       g1OopClosures.inline.hpp
 objArrayKlass.cpp                       oop.pcgc.inline.hpp
+objArrayKlass.cpp                       psCompactionManager.hpp
 objArrayKlass.cpp                       psPromotionManager.inline.hpp
 objArrayKlass.cpp                       psScavenge.inline.hpp
 objArrayKlass.cpp                       parOopClosures.inline.hpp
 
+objArrayKlass.inline.hpp		psCompactionManager.inline.hpp
+objArrayKlass.inline.hpp		psParallelCompact.hpp
+
 oop.pcgc.inline.hpp                     parNewGeneration.hpp
 oop.pcgc.inline.hpp                     parallelScavengeHeap.hpp
 oop.pcgc.inline.hpp                     psCompactionManager.hpp

hotspot/src/share/vm/memory/genMarkSweep.cpp

@@ -159,6 +159,7 @@ void GenMarkSweep::allocate_stacks() {
   _preserved_oop_stack = NULL;
 
   _marking_stack       = new (ResourceObj::C_HEAP) GrowableArray<oop>(4000, true);
+  _objarray_stack      = new (ResourceObj::C_HEAP) GrowableArray<ObjArrayTask>(50, true);
 
   int size = SystemDictionary::number_of_classes() * 2;
   _revisit_klass_stack = new (ResourceObj::C_HEAP) GrowableArray<Klass*>(size, true);
@@ -194,7 +195,6 @@ void GenMarkSweep::allocate_stacks() {
 
 
 void GenMarkSweep::deallocate_stacks() {
-
   if (!UseG1GC) {
     GenCollectedHeap* gch = GenCollectedHeap::heap();
     gch->release_scratch();
@@ -208,6 +208,7 @@ void GenMarkSweep::deallocate_stacks() {
   }
 
   delete _marking_stack;
+  delete _objarray_stack;
   delete _revisit_klass_stack;
   delete _revisit_mdo_stack;
 

hotspot/src/share/vm/memory/genOopClosures.hpp

@@ -28,10 +28,10 @@ class CardTableRS;
 class CardTableModRefBS;
 class DefNewGeneration;
 
-template<class E> class GenericTaskQueue;
+template<class E, unsigned int N> class GenericTaskQueue;
-typedef GenericTaskQueue<oop> OopTaskQueue;
+typedef GenericTaskQueue<oop, TASKQUEUE_SIZE> OopTaskQueue;
-template<class E> class GenericTaskQueueSet;
+template<class T> class GenericTaskQueueSet;
-typedef GenericTaskQueueSet<oop> OopTaskQueueSet;
+typedef GenericTaskQueueSet<OopTaskQueue> OopTaskQueueSet;
 
 // Closure for iterating roots from a particular generation
 // Note: all classes deriving from this MUST call this do_barrier

hotspot/src/share/vm/oops/instanceKlass.cpp

@@ -25,6 +25,58 @@
 # include "incls/_precompiled.incl"
 # include "incls/_instanceKlass.cpp.incl"
 
+#ifdef DTRACE_ENABLED
+
+HS_DTRACE_PROBE_DECL4(hotspot, class__initialization__required,
+  char*, intptr_t, oop, intptr_t);
+HS_DTRACE_PROBE_DECL5(hotspot, class__initialization__recursive,
+  char*, intptr_t, oop, intptr_t, int);
+HS_DTRACE_PROBE_DECL5(hotspot, class__initialization__concurrent,
+  char*, intptr_t, oop, intptr_t, int);
+HS_DTRACE_PROBE_DECL5(hotspot, class__initialization__erroneous,
+  char*, intptr_t, oop, intptr_t, int);
+HS_DTRACE_PROBE_DECL5(hotspot, class__initialization__super__failed,
+  char*, intptr_t, oop, intptr_t, int);
+HS_DTRACE_PROBE_DECL5(hotspot, class__initialization__clinit,
+  char*, intptr_t, oop, intptr_t, int);
+HS_DTRACE_PROBE_DECL5(hotspot, class__initialization__error,
+  char*, intptr_t, oop, intptr_t, int);
+HS_DTRACE_PROBE_DECL5(hotspot, class__initialization__end,
+  char*, intptr_t, oop, intptr_t, int);
+
+#define DTRACE_CLASSINIT_PROBE(type, clss, thread_type)          \
+  {                                                              \
+    char* data = NULL;                                           \
+    int len = 0;                                                 \
+    symbolOop name = (clss)->name();                             \
+    if (name != NULL) {                                          \
+      data = (char*)name->bytes();                               \
+      len = name->utf8_length();                                 \
+    }                                                            \
+    HS_DTRACE_PROBE4(hotspot, class__initialization__##type,     \
+      data, len, (clss)->class_loader(), thread_type);           \
+  }
+
+#define DTRACE_CLASSINIT_PROBE_WAIT(type, clss, thread_type, wait) \
+  {                                                              \
+    char* data = NULL;                                           \
+    int len = 0;                                                 \
+    symbolOop name = (clss)->name();                             \
+    if (name != NULL) {                                          \
+      data = (char*)name->bytes();                               \
+      len = name->utf8_length();                                 \
+    }                                                            \
+    HS_DTRACE_PROBE5(hotspot, class__initialization__##type,     \
+      data, len, (clss)->class_loader(), thread_type, wait);     \
+  }
+
+#else //  ndef DTRACE_ENABLED
+
+#define DTRACE_CLASSINIT_PROBE(type, clss, thread_type)
+#define DTRACE_CLASSINIT_PROBE_WAIT(type, clss, thread_type, wait)
+
+#endif //  ndef DTRACE_ENABLED
+
 bool instanceKlass::should_be_initialized() const {
   return !is_initialized();
 }
@@ -292,6 +344,10 @@ void instanceKlass::initialize_impl(instanceKlassHandle this_oop, TRAPS) {
   // A class could already be verified, since it has been reflected upon.
   this_oop->link_class(CHECK);
 
+  DTRACE_CLASSINIT_PROBE(required, instanceKlass::cast(this_oop()), -1);
+
+  bool wait = false;
+
   // refer to the JVM book page 47 for description of steps
   // Step 1
   { ObjectLocker ol(this_oop, THREAD);
@@ -303,19 +359,25 @@ void instanceKlass::initialize_impl(instanceKlassHandle this_oop, TRAPS) {
     // we might end up throwing IE from link/symbol resolution sites
     // that aren't expected to throw.  This would wreak havoc.  See 6320309.
     while(this_oop->is_being_initialized() && !this_oop->is_reentrant_initialization(self)) {
+        wait = true;
       ol.waitUninterruptibly(CHECK);
     }
 
     // Step 3
-    if (this_oop->is_being_initialized() && this_oop->is_reentrant_initialization(self))
+    if (this_oop->is_being_initialized() && this_oop->is_reentrant_initialization(self)) {
+      DTRACE_CLASSINIT_PROBE_WAIT(recursive, instanceKlass::cast(this_oop()), -1,wait);
       return;
+    }
 
     // Step 4
-    if (this_oop->is_initialized())
+    if (this_oop->is_initialized()) {
+      DTRACE_CLASSINIT_PROBE_WAIT(concurrent, instanceKlass::cast(this_oop()), -1,wait);
       return;
+    }
 
     // Step 5
     if (this_oop->is_in_error_state()) {
+      DTRACE_CLASSINIT_PROBE_WAIT(erroneous, instanceKlass::cast(this_oop()), -1,wait);
       ResourceMark rm(THREAD);
       const char* desc = "Could not initialize class ";
       const char* className = this_oop->external_name();
@@ -348,6 +410,7 @@ void instanceKlass::initialize_impl(instanceKlassHandle this_oop, TRAPS) {
         this_oop->set_initialization_state_and_notify(initialization_error, THREAD); // Locks object, set state, and notify all waiting threads
         CLEAR_PENDING_EXCEPTION;   // ignore any exception thrown, superclass initialization error is thrown below
       }
+      DTRACE_CLASSINIT_PROBE_WAIT(super__failed, instanceKlass::cast(this_oop()), -1,wait);
       THROW_OOP(e());
     }
   }
@@ -356,6 +419,7 @@ void instanceKlass::initialize_impl(instanceKlassHandle this_oop, TRAPS) {
   {
     assert(THREAD->is_Java_thread(), "non-JavaThread in initialize_impl");
     JavaThread* jt = (JavaThread*)THREAD;
+    DTRACE_CLASSINIT_PROBE_WAIT(clinit, instanceKlass::cast(this_oop()), -1,wait);
     // Timer includes any side effects of class initialization (resolution,
     // etc), but not recursive entry into call_class_initializer().
     PerfClassTraceTime timer(ClassLoader::perf_class_init_time(),
@@ -383,6 +447,7 @@ void instanceKlass::initialize_impl(instanceKlassHandle this_oop, TRAPS) {
       this_oop->set_initialization_state_and_notify(initialization_error, THREAD);
       CLEAR_PENDING_EXCEPTION;   // ignore any exception thrown, class initialization error is thrown below
     }
+    DTRACE_CLASSINIT_PROBE_WAIT(error, instanceKlass::cast(this_oop()), -1,wait);
     if (e->is_a(SystemDictionary::Error_klass())) {
       THROW_OOP(e());
     } else {
@@ -392,6 +457,7 @@ void instanceKlass::initialize_impl(instanceKlassHandle this_oop, TRAPS) {
                 &args);
     }
   }
+  DTRACE_CLASSINIT_PROBE_WAIT(end, instanceKlass::cast(this_oop()), -1,wait);
 }
 
 

hotspot/src/share/vm/oops/objArrayKlass.cpp

@@ -314,24 +314,24 @@ void objArrayKlass::initialize(TRAPS) {
 
 void objArrayKlass::oop_follow_contents(oop obj) {
   assert (obj->is_array(), "obj must be array");
-  objArrayOop a = objArrayOop(obj);
+  objArrayOop(obj)->follow_header();
-  a->follow_header();
+  if (UseCompressedOops) {
-  ObjArrayKlass_OOP_ITERATE( \
+    objarray_follow_contents<narrowOop>(obj, 0);
-    a, p, \
+  } else {
-    /* we call mark_and_follow here to avoid excessive marking stack usage */ \
+    objarray_follow_contents<oop>(obj, 0);
-    MarkSweep::mark_and_follow(p))
+  }
 }
 
 #ifndef SERIALGC
 void objArrayKlass::oop_follow_contents(ParCompactionManager* cm,
                                         oop obj) {
-  assert (obj->is_array(), "obj must be array");
+  assert(obj->is_array(), "obj must be array");
-  objArrayOop a = objArrayOop(obj);
+  objArrayOop(obj)->follow_header(cm);
-  a->follow_header(cm);
+  if (UseCompressedOops) {
-  ObjArrayKlass_OOP_ITERATE( \
+    objarray_follow_contents<narrowOop>(cm, obj, 0);
-    a, p, \
+  } else {
-    /* we call mark_and_follow here to avoid excessive marking stack usage */ \
+    objarray_follow_contents<oop>(cm, obj, 0);
-    PSParallelCompact::mark_and_follow(cm, p))
+  }
 }
 #endif // SERIALGC
 

hotspot/src/share/vm/oops/objArrayKlass.hpp

@@ -91,10 +91,18 @@ class objArrayKlass : public arrayKlass {
 
   // Garbage collection
   void oop_follow_contents(oop obj);
+  inline void oop_follow_contents(oop obj, int index);
+  template <class T> inline void objarray_follow_contents(oop obj, int index);
+
   int  oop_adjust_pointers(oop obj);
 
   // Parallel Scavenge and Parallel Old
   PARALLEL_GC_DECLS
+#ifndef SERIALGC
+  inline void oop_follow_contents(ParCompactionManager* cm, oop obj, int index);
+  template <class T> inline void
+    objarray_follow_contents(ParCompactionManager* cm, oop obj, int index);
+#endif // !SERIALGC
 
   // Iterators
   int oop_oop_iterate(oop obj, OopClosure* blk) {
@@ -131,5 +139,4 @@ class objArrayKlass : public arrayKlass {
   void oop_verify_on(oop obj, outputStream* st);
   void oop_verify_old_oop(oop obj, oop* p, bool allow_dirty);
   void oop_verify_old_oop(oop obj, narrowOop* p, bool allow_dirty);
-
 };

hotspot/src/share/vm/oops/objArrayKlass.inline.hpp

@@ -0,0 +1,89 @@
+/*
+ * Copyright 2010 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ */
+
+void objArrayKlass::oop_follow_contents(oop obj, int index) {
+  if (UseCompressedOops) {
+    objarray_follow_contents<narrowOop>(obj, index);
+  } else {
+    objarray_follow_contents<oop>(obj, index);
+  }
+}
+
+template <class T>
+void objArrayKlass::objarray_follow_contents(oop obj, int index) {
+  objArrayOop a = objArrayOop(obj);
+  const size_t len = size_t(a->length());
+  const size_t beg_index = size_t(index);
+  assert(beg_index < len || len == 0, "index too large");
+
+  const size_t stride = MIN2(len - beg_index, ObjArrayMarkingStride);
+  const size_t end_index = beg_index + stride;
+  T* const base = (T*)a->base();
+  T* const beg = base + beg_index;
+  T* const end = base + end_index;
+
+  // Push the non-NULL elements of the next stride on the marking stack.
+  for (T* e = beg; e < end; e++) {
+    MarkSweep::mark_and_push<T>(e);
+  }
+
+  if (end_index < len) {
+    MarkSweep::push_objarray(a, end_index); // Push the continuation.
+  }
+}
+
+#ifndef SERIALGC
+void objArrayKlass::oop_follow_contents(ParCompactionManager* cm, oop obj,
+                                        int index) {
+  if (UseCompressedOops) {
+    objarray_follow_contents<narrowOop>(cm, obj, index);
+  } else {
+    objarray_follow_contents<oop>(cm, obj, index);
+  }
+}
+
+template <class T>
+void objArrayKlass::objarray_follow_contents(ParCompactionManager* cm, oop obj,
+                                             int index) {
+  objArrayOop a = objArrayOop(obj);
+  const size_t len = size_t(a->length());
+  const size_t beg_index = size_t(index);
+  assert(beg_index < len || len == 0, "index too large");
+
+  const size_t stride = MIN2(len - beg_index, ObjArrayMarkingStride);
+  const size_t end_index = beg_index + stride;
+  T* const base = (T*)a->base();
+  T* const beg = base + beg_index;
+  T* const end = base + end_index;
+
+  // Push the non-NULL elements of the next stride on the marking stack.
+  for (T* e = beg; e < end; e++) {
+    PSParallelCompact::mark_and_push<T>(cm, e);
+  }
+
+  if (end_index < len) {
+    cm->push_objarray(a, end_index); // Push the continuation.
+  }
+}
+#endif // #ifndef SERIALGC

hotspot/src/share/vm/prims/jvm.cpp

@@ -26,6 +26,10 @@
 #include "incls/_jvm.cpp.incl"
 #include <errno.h>
 
+HS_DTRACE_PROBE_DECL1(hotspot, thread__sleep__begin, long long);
+HS_DTRACE_PROBE_DECL1(hotspot, thread__sleep__end, int);
+HS_DTRACE_PROBE_DECL0(hotspot, thread__yield);
+
 /*
   NOTE about use of any ctor or function call that can trigger a safepoint/GC:
   such ctors and calls MUST NOT come between an oop declaration/init and its
@@ -2762,6 +2766,7 @@ JVM_END
 JVM_ENTRY(void, JVM_Yield(JNIEnv *env, jclass threadClass))
   JVMWrapper("JVM_Yield");
   if (os::dont_yield()) return;
+  HS_DTRACE_PROBE0(hotspot, thread__yield);
   // When ConvertYieldToSleep is off (default), this matches the classic VM use of yield.
   // Critical for similar threading behaviour
   if (ConvertYieldToSleep) {
@@ -2787,6 +2792,8 @@ JVM_ENTRY(void, JVM_Sleep(JNIEnv* env, jclass threadClass, jlong millis))
   // And set new thread state to SLEEPING.
   JavaThreadSleepState jtss(thread);
 
+  HS_DTRACE_PROBE1(hotspot, thread__sleep__begin, millis);
+
   if (millis == 0) {
     // When ConvertSleepToYield is on, this matches the classic VM implementation of
     // JVM_Sleep. Critical for similar threading behaviour (Win32)
@@ -2807,6 +2814,7 @@ JVM_ENTRY(void, JVM_Sleep(JNIEnv* env, jclass threadClass, jlong millis))
       // An asynchronous exception (e.g., ThreadDeathException) could have been thrown on
       // us while we were sleeping. We do not overwrite those.
       if (!HAS_PENDING_EXCEPTION) {
+        HS_DTRACE_PROBE1(hotspot, thread__sleep__end,1);
         // TODO-FIXME: THROW_MSG returns which means we will not call set_state()
         // to properly restore the thread state.  That's likely wrong.
         THROW_MSG(vmSymbols::java_lang_InterruptedException(), "sleep interrupted");
@@ -2814,6 +2822,7 @@ JVM_ENTRY(void, JVM_Sleep(JNIEnv* env, jclass threadClass, jlong millis))
     }
     thread->osthread()->set_state(old_state);
   }
+  HS_DTRACE_PROBE1(hotspot, thread__sleep__end,0);
 JVM_END
 
 JVM_ENTRY(jobject, JVM_CurrentThread(JNIEnv* env, jclass threadClass))

hotspot/src/share/vm/prims/unsafe.cpp

@@ -29,6 +29,10 @@
 #include "incls/_precompiled.incl"
 #include "incls/_unsafe.cpp.incl"
 
+HS_DTRACE_PROBE_DECL3(hotspot, thread__park__begin, uintptr_t, int, long long);
+HS_DTRACE_PROBE_DECL1(hotspot, thread__park__end, uintptr_t);
+HS_DTRACE_PROBE_DECL1(hotspot, thread__unpark, uintptr_t);
+
 #define MAX_OBJECT_SIZE \
   ( arrayOopDesc::header_size(T_DOUBLE) * HeapWordSize \
     + ((julong)max_jint * sizeof(double)) )
@@ -1083,8 +1087,10 @@ UNSAFE_END
 
 UNSAFE_ENTRY(void, Unsafe_Park(JNIEnv *env, jobject unsafe, jboolean isAbsolute, jlong time))
   UnsafeWrapper("Unsafe_Park");
+  HS_DTRACE_PROBE3(hotspot, thread__park__begin, thread->parker(), (int) isAbsolute, time);
   JavaThreadParkedState jtps(thread, time != 0);
   thread->parker()->park(isAbsolute != 0, time);
+  HS_DTRACE_PROBE1(hotspot, thread__park__end, thread->parker());
 UNSAFE_END
 
 UNSAFE_ENTRY(void, Unsafe_Unpark(JNIEnv *env, jobject unsafe, jobject jthread))
@@ -1116,6 +1122,7 @@ UNSAFE_ENTRY(void, Unsafe_Unpark(JNIEnv *env, jobject unsafe, jobject jthread))
     }
   }
   if (p != NULL) {
+    HS_DTRACE_PROBE1(hotspot, thread__unpark, p);
     p->unpark();
   }
 UNSAFE_END

hotspot/src/share/vm/runtime/arguments.cpp

@@ -1346,9 +1346,7 @@ void Arguments::set_g1_gc_flags() {
   }
 
   if (FLAG_IS_DEFAULT(MarkStackSize)) {
-    // Size as a multiple of TaskQueueSuper::N which is larger
+    FLAG_SET_DEFAULT(MarkStackSize, 128 * TASKQUEUE_SIZE);
-    // for 64-bit.
-    FLAG_SET_DEFAULT(MarkStackSize, 128 * TaskQueueSuper::total_size());
   }
   if (PrintGCDetails && Verbose) {
     tty->print_cr("MarkStackSize: %uk  MarkStackSizeMax: %uk",

hotspot/src/share/vm/runtime/globals.hpp

@@ -1795,6 +1795,10 @@ class CommandLineFlags {
   product(uintx, PreserveMarkStackSize, 1024,                               \
           "Size for stack used in promotion failure handling")              \
                                                                             \
+  develop(uintx, ObjArrayMarkingStride, 512,                                \
+          "Number of ObjArray elements to push onto the marking stack"      \
+          "before pushing a continuation entry")                            \
+                                                                            \
   product_pd(bool, UseTLAB, "Use thread-local object allocation")           \
                                                                             \
   product_pd(bool, ResizeTLAB,                                              \
@@ -2289,6 +2293,10 @@ class CommandLineFlags {
           "print safepoint statistics only when safepoint takes"            \
           " more than PrintSafepointSatisticsTimeout in millis")            \
                                                                             \
+  product(bool, TraceSafepointCleanupTime, false,                           \
+          "print the break down of clean up tasks performed during"         \
+          " safepoint")                                                     \
+                                                                            \
   develop(bool, InlineAccessors, true,                                      \
           "inline accessor methods (get/set)")                              \
                                                                             \

hotspot/src/share/vm/runtime/os.hpp

@@ -218,6 +218,9 @@ class os: AllStatic {
 
   static bool   guard_memory(char* addr, size_t bytes);
   static bool   unguard_memory(char* addr, size_t bytes);
+  static bool   create_stack_guard_pages(char* addr, size_t bytes);
+  static bool   remove_stack_guard_pages(char* addr, size_t bytes);
+
   static char*  map_memory(int fd, const char* file_name, size_t file_offset,
                            char *addr, size_t bytes, bool read_only = false,
                            bool allow_exec = false);

hotspot/src/share/vm/runtime/safepoint.cpp

@@ -457,21 +457,21 @@ void CounterDecay::decay() {
 // Various cleaning tasks that should be done periodically at safepoints
 void SafepointSynchronize::do_cleanup_tasks() {
   {
-    TraceTime t1("deflating idle monitors", TraceSafepoint);
+    TraceTime t1("deflating idle monitors", TraceSafepointCleanupTime);
     ObjectSynchronizer::deflate_idle_monitors();
   }
 
   {
-    TraceTime t2("updating inline caches", TraceSafepoint);
+    TraceTime t2("updating inline caches", TraceSafepointCleanupTime);
     InlineCacheBuffer::update_inline_caches();
   }
 
   if(UseCounterDecay && CounterDecay::is_decay_needed()) {
-    TraceTime t3("decaying counter", TraceSafepoint);
+    TraceTime t3("decaying counter", TraceSafepointCleanupTime);
     CounterDecay::decay();
   }
 
-  TraceTime t4("sweeping nmethods", TraceSafepoint);
+  TraceTime t4("sweeping nmethods", TraceSafepointCleanupTime);
   NMethodSweeper::sweep();
 }
 

hotspot/src/share/vm/runtime/thread.cpp

@@ -2137,7 +2137,7 @@ void JavaThread::create_stack_guard_pages() {
   int allocate = os::allocate_stack_guard_pages();
   // warning("Guarding at " PTR_FORMAT " for len " SIZE_FORMAT "\n", low_addr, len);
 
-  if (allocate && !os::commit_memory((char *) low_addr, len)) {
+  if (allocate && !os::create_stack_guard_pages((char *) low_addr, len)) {
     warning("Attempt to allocate stack guard pages failed.");
     return;
   }
@@ -2158,7 +2158,7 @@ void JavaThread::remove_stack_guard_pages() {
   size_t len = (StackYellowPages + StackRedPages) * os::vm_page_size();
 
   if (os::allocate_stack_guard_pages()) {
-    if (os::uncommit_memory((char *) low_addr, len)) {
+    if (os::remove_stack_guard_pages((char *) low_addr, len)) {
       _stack_guard_state = stack_guard_unused;
     } else {
       warning("Attempt to deallocate stack guard pages failed.");

hotspot/src/share/vm/runtime/vmThread.cpp

@@ -25,6 +25,10 @@
 # include "incls/_precompiled.incl"
 # include "incls/_vmThread.cpp.incl"
 
+HS_DTRACE_PROBE_DECL3(hotspot, vmops__request, char *, uintptr_t, int);
+HS_DTRACE_PROBE_DECL3(hotspot, vmops__begin, char *, uintptr_t, int);
+HS_DTRACE_PROBE_DECL3(hotspot, vmops__end, char *, uintptr_t, int);
+
 // Dummy VM operation to act as first element in our circular double-linked list
 class VM_Dummy: public VM_Operation {
   VMOp_Type type() const { return VMOp_Dummy; }
@@ -132,6 +136,10 @@ void VMOperationQueue::drain_list_oops_do(OopClosure* f) {
 //-----------------------------------------------------------------
 // High-level interface
 bool VMOperationQueue::add(VM_Operation *op) {
+
+  HS_DTRACE_PROBE3(hotspot, vmops__request, op->name(), strlen(op->name()),
+                   op->evaluation_mode());
+
   // Encapsulates VM queue policy. Currently, that
   // only involves putting them on the right list
   if (op->evaluate_at_safepoint()) {
@@ -325,7 +333,11 @@ void VMThread::evaluate_operation(VM_Operation* op) {
 
   {
     PerfTraceTime vm_op_timer(perf_accumulated_vm_operation_time());
+    HS_DTRACE_PROBE3(hotspot, vmops__begin, op->name(), strlen(op->name()),
+                     op->evaluation_mode());
     op->evaluate();
+    HS_DTRACE_PROBE3(hotspot, vmops__end, op->name(), strlen(op->name()),
+                     op->evaluation_mode());
   }
 
   // Last access of info in _cur_vm_operation!

hotspot/src/share/vm/services/dtraceAttacher.cpp

@@ -135,4 +135,9 @@ void DTrace::set_extended_dprobes(bool flag) {
   }
 }
 
+void DTrace::set_monitor_dprobes(bool flag) {
+  // explicit setting of DTraceMonitorProbes flag
+  set_bool_flag("DTraceMonitorProbes", flag);
+}
+
 #endif /* SOLARIS */

hotspot/src/share/vm/services/dtraceAttacher.hpp

@@ -41,4 +41,6 @@ class DTrace : public AllStatic {
   static void detach_all_clients();
   // set ExtendedDTraceProbes flag
   static void set_extended_dprobes(bool value);
+  // set DTraceMonitorProbes flag
+  static void set_monitor_dprobes(bool value);
 };

hotspot/src/share/vm/services/management.cpp

@@ -1537,7 +1537,6 @@ bool add_global_entry(JNIEnv* env, Handle name, jmmVMGlobal *global, Flag *flag,
     global->type = JMM_VMGLOBAL_TYPE_JSTRING;
   } else {
     global->type = JMM_VMGLOBAL_TYPE_UNKNOWN;
-    assert(false, "Unsupported VMGlobal Type");
     return false;
   }
 

hotspot/src/share/vm/utilities/globalDefinitions.hpp

@@ -827,6 +827,8 @@ const int      badCodeHeapFreeVal = 0xDD;                   // value used to zap
 #define       badHeapWord       (::badHeapWordVal)
 #define       badJNIHandle      ((oop)::badJNIHandleVal)
 
+// Default TaskQueue size is 16K (32-bit) or 128K (64-bit)
+#define TASKQUEUE_SIZE (NOT_LP64(1<<14) LP64_ONLY(1<<17))
 
 //----------------------------------------------------------------------------------------------------
 // Utility functions for bitfield manipulations

hotspot/src/share/vm/utilities/taskqueue.cpp

@@ -31,10 +31,6 @@ uint ParallelTaskTerminator::_total_spins = 0;
 uint ParallelTaskTerminator::_total_peeks = 0;
 #endif
 
-bool TaskQueueSuper::peek() {
-  return _bottom != _age.top();
-}
-
 int TaskQueueSetSuper::randomParkAndMiller(int *seed0) {
   const int a =      16807;
   const int m = 2147483647;
@@ -180,6 +176,13 @@ void ParallelTaskTerminator::reset_for_reuse() {
   }
 }
 
+#ifdef ASSERT
+bool ObjArrayTask::is_valid() const {
+  return _obj != NULL && _obj->is_objArray() && _index > 0 &&
+    _index < objArrayOop(_obj)->length();
+}
+#endif // ASSERT
+
 bool RegionTaskQueueWithOverflow::is_empty() {
   return (_region_queue.size() == 0) &&
          (_overflow_stack->length() == 0);

hotspot/src/share/vm/utilities/taskqueue.hpp

@@ -22,6 +22,7 @@
  *
  */
 
+template <unsigned int N>
 class TaskQueueSuper: public CHeapObj {
 protected:
   // Internal type for indexing the queue; also used for the tag.
@@ -30,10 +31,7 @@ protected:
   // The first free element after the last one pushed (mod N).
   volatile uint _bottom;
 
-  enum {
+  enum { MOD_N_MASK = N - 1 };
-    N = 1 << NOT_LP64(14) LP64_ONLY(17), // Queue size: 16K or 128K
-    MOD_N_MASK = N - 1                   // To compute x mod N efficiently.
-  };
 
   class Age {
   public:
@@ -84,12 +82,12 @@ protected:
 
   // Returns a number in the range [0..N).  If the result is "N-1", it should be
   // interpreted as 0.
-  uint dirty_size(uint bot, uint top) {
+  uint dirty_size(uint bot, uint top) const {
     return (bot - top) & MOD_N_MASK;
   }
 
   // Returns the size corresponding to the given "bot" and "top".
-  uint size(uint bot, uint top) {
+  uint size(uint bot, uint top) const {
     uint sz = dirty_size(bot, top);
     // Has the queue "wrapped", so that bottom is less than top?  There's a
     // complicated special case here.  A pair of threads could perform pop_local
@@ -111,17 +109,17 @@ protected:
 public:
   TaskQueueSuper() : _bottom(0), _age() {}
 
-  // Return "true" if the TaskQueue contains any tasks.
+  // Return true if the TaskQueue contains any tasks.
-  bool peek();
+  bool peek() { return _bottom != _age.top(); }
 
   // Return an estimate of the number of elements in the queue.
   // The "careful" version admits the possibility of pop_local/pop_global
   // races.
-  uint size() {
+  uint size() const {
     return size(_bottom, _age.top());
   }
 
-  uint dirty_size() {
+  uint dirty_size() const {
     return dirty_size(_bottom, _age.top());
   }
 
@@ -132,19 +130,36 @@ public:
 
   // Maximum number of elements allowed in the queue.  This is two less
   // than the actual queue size, for somewhat complicated reasons.
-  uint max_elems() { return N - 2; }
+  uint max_elems() const { return N - 2; }
 
   // Total size of queue.
   static const uint total_size() { return N; }
 };
 
-template<class E> class GenericTaskQueue: public TaskQueueSuper {
+template<class E, unsigned int N = TASKQUEUE_SIZE>
+class GenericTaskQueue: public TaskQueueSuper<N> {
+protected:
+  typedef typename TaskQueueSuper<N>::Age Age;
+  typedef typename TaskQueueSuper<N>::idx_t idx_t;
+
+  using TaskQueueSuper<N>::_bottom;
+  using TaskQueueSuper<N>::_age;
+  using TaskQueueSuper<N>::increment_index;
+  using TaskQueueSuper<N>::decrement_index;
+  using TaskQueueSuper<N>::dirty_size;
+
+public:
+  using TaskQueueSuper<N>::max_elems;
+  using TaskQueueSuper<N>::size;
+
 private:
   // Slow paths for push, pop_local.  (pop_global has no fast path.)
   bool push_slow(E t, uint dirty_n_elems);
   bool pop_local_slow(uint localBot, Age oldAge);
 
 public:
+  typedef E element_type;
+
   // Initializes the queue to empty.
   GenericTaskQueue();
 
@@ -175,19 +190,19 @@ private:
   volatile E* _elems;
 };
 
-template<class E>
+template<class E, unsigned int N>
-GenericTaskQueue<E>::GenericTaskQueue():TaskQueueSuper() {
+GenericTaskQueue<E, N>::GenericTaskQueue() {
   assert(sizeof(Age) == sizeof(size_t), "Depends on this.");
 }
 
-template<class E>
+template<class E, unsigned int N>
-void GenericTaskQueue<E>::initialize() {
+void GenericTaskQueue<E, N>::initialize() {
   _elems = NEW_C_HEAP_ARRAY(E, N);
   guarantee(_elems != NULL, "Allocation failed.");
 }
 
-template<class E>
+template<class E, unsigned int N>
-void GenericTaskQueue<E>::oops_do(OopClosure* f) {
+void GenericTaskQueue<E, N>::oops_do(OopClosure* f) {
   // tty->print_cr("START OopTaskQueue::oops_do");
   uint iters = size();
   uint index = _bottom;
@@ -203,21 +218,21 @@ void GenericTaskQueue<E>::oops_do(OopClosure* f) {
   // tty->print_cr("END OopTaskQueue::oops_do");
 }
 
-
+template<class E, unsigned int N>
-template<class E>
+bool GenericTaskQueue<E, N>::push_slow(E t, uint dirty_n_elems) {
-bool GenericTaskQueue<E>::push_slow(E t, uint dirty_n_elems) {
   if (dirty_n_elems == N - 1) {
     // Actually means 0, so do the push.
     uint localBot = _bottom;
-    _elems[localBot] = t;
+    // g++ complains if the volatile result of the assignment is unused.
+    const_cast<E&>(_elems[localBot] = t);
     OrderAccess::release_store(&_bottom, increment_index(localBot));
     return true;
   }
   return false;
 }
 
-template<class E>
+template<class E, unsigned int N>
-bool GenericTaskQueue<E>::
+bool GenericTaskQueue<E, N>::
 pop_local_slow(uint localBot, Age oldAge) {
   // This queue was observed to contain exactly one element; either this
   // thread will claim it, or a competing "pop_global".  In either case,
@@ -249,8 +264,8 @@ pop_local_slow(uint localBot, Age oldAge) {
   return false;
 }
 
-template<class E>
+template<class E, unsigned int N>
-bool GenericTaskQueue<E>::pop_global(E& t) {
+bool GenericTaskQueue<E, N>::pop_global(E& t) {
   Age oldAge = _age.get();
   uint localBot = _bottom;
   uint n_elems = size(localBot, oldAge.top());
@@ -258,7 +273,7 @@ bool GenericTaskQueue<E>::pop_global(E& t) {
     return false;
   }
 
-  t = _elems[oldAge.top()];
+  const_cast<E&>(t = _elems[oldAge.top()]);
   Age newAge(oldAge);
   newAge.increment();
   Age resAge = _age.cmpxchg(newAge, oldAge);
@@ -269,8 +284,8 @@ bool GenericTaskQueue<E>::pop_global(E& t) {
   return resAge == oldAge;
 }
 
-template<class E>
+template<class E, unsigned int N>
-GenericTaskQueue<E>::~GenericTaskQueue() {
+GenericTaskQueue<E, N>::~GenericTaskQueue() {
   FREE_C_HEAP_ARRAY(E, _elems);
 }
 
@@ -283,16 +298,18 @@ public:
   virtual bool peek() = 0;
 };
 
-template<class E> class GenericTaskQueueSet: public TaskQueueSetSuper {
+template<class T>
+class GenericTaskQueueSet: public TaskQueueSetSuper {
 private:
   uint _n;
-  GenericTaskQueue<E>** _queues;
+  T** _queues;
 
 public:
+  typedef typename T::element_type E;
+
   GenericTaskQueueSet(int n) : _n(n) {
-    typedef GenericTaskQueue<E>* GenericTaskQueuePtr;
+    typedef T* GenericTaskQueuePtr;
     _queues = NEW_C_HEAP_ARRAY(GenericTaskQueuePtr, n);
-    guarantee(_queues != NULL, "Allocation failure.");
     for (int i = 0; i < n; i++) {
       _queues[i] = NULL;
     }
@@ -302,9 +319,9 @@ public:
   bool steal_best_of_2(uint queue_num, int* seed, E& t);
   bool steal_best_of_all(uint queue_num, int* seed, E& t);
 
-  void register_queue(uint i, GenericTaskQueue<E>* q);
+  void register_queue(uint i, T* q);
 
-  GenericTaskQueue<E>* queue(uint n);
+  T* queue(uint n);
 
   // The thread with queue number "queue_num" (and whose random number seed
   // is at "seed") is trying to steal a task from some other queue.  (It
@@ -316,27 +333,27 @@ public:
   bool peek();
 };
 
-template<class E>
+template<class T> void
-void GenericTaskQueueSet<E>::register_queue(uint i, GenericTaskQueue<E>* q) {
+GenericTaskQueueSet<T>::register_queue(uint i, T* q) {
   assert(i < _n, "index out of range.");
   _queues[i] = q;
 }
 
-template<class E>
+template<class T> T*
-GenericTaskQueue<E>* GenericTaskQueueSet<E>::queue(uint i) {
+GenericTaskQueueSet<T>::queue(uint i) {
   return _queues[i];
 }
 
-template<class E>
+template<class T> bool
-bool GenericTaskQueueSet<E>::steal(uint queue_num, int* seed, E& t) {
+GenericTaskQueueSet<T>::steal(uint queue_num, int* seed, E& t) {
   for (uint i = 0; i < 2 * _n; i++)
     if (steal_best_of_2(queue_num, seed, t))
       return true;
   return false;
 }
 
-template<class E>
+template<class T> bool
-bool GenericTaskQueueSet<E>::steal_best_of_all(uint queue_num, int* seed, E& t) {
+GenericTaskQueueSet<T>::steal_best_of_all(uint queue_num, int* seed, E& t) {
   if (_n > 2) {
     int best_k;
     uint best_sz = 0;
@@ -359,8 +376,8 @@ bool GenericTaskQueueSet<E>::steal_best_of_all(uint queue_num, int* seed, E& t)
   }
 }
 
-template<class E>
+template<class T> bool
-bool GenericTaskQueueSet<E>::steal_1_random(uint queue_num, int* seed, E& t) {
+GenericTaskQueueSet<T>::steal_1_random(uint queue_num, int* seed, E& t) {
   if (_n > 2) {
     uint k = queue_num;
     while (k == queue_num) k = randomParkAndMiller(seed) % _n;
@@ -375,8 +392,8 @@ bool GenericTaskQueueSet<E>::steal_1_random(uint queue_num, int* seed, E& t) {
   }
 }
 
-template<class E>
+template<class T> bool
-bool GenericTaskQueueSet<E>::steal_best_of_2(uint queue_num, int* seed, E& t) {
+GenericTaskQueueSet<T>::steal_best_of_2(uint queue_num, int* seed, E& t) {
   if (_n > 2) {
     uint k1 = queue_num;
     while (k1 == queue_num) k1 = randomParkAndMiller(seed) % _n;
@@ -397,8 +414,8 @@ bool GenericTaskQueueSet<E>::steal_best_of_2(uint queue_num, int* seed, E& t) {
   }
 }
 
-template<class E>
+template<class T>
-bool GenericTaskQueueSet<E>::peek() {
+bool GenericTaskQueueSet<T>::peek() {
   // Try all the queues.
   for (uint j = 0; j < _n; j++) {
     if (_queues[j]->peek())
@@ -468,14 +485,16 @@ public:
 #endif
 };
 
-template<class E> inline bool GenericTaskQueue<E>::push(E t) {
+template<class E, unsigned int N> inline bool
+GenericTaskQueue<E, N>::push(E t) {
   uint localBot = _bottom;
   assert((localBot >= 0) && (localBot < N), "_bottom out of range.");
   idx_t top = _age.top();
   uint dirty_n_elems = dirty_size(localBot, top);
-  assert((dirty_n_elems >= 0) && (dirty_n_elems < N), "n_elems out of range.");
+  assert(dirty_n_elems < N, "n_elems out of range.");
   if (dirty_n_elems < max_elems()) {
-    _elems[localBot] = t;
+    // g++ complains if the volatile result of the assignment is unused.
+    const_cast<E&>(_elems[localBot] = t);
     OrderAccess::release_store(&_bottom, increment_index(localBot));
     return true;
   } else {
@@ -483,7 +502,8 @@ template<class E> inline bool GenericTaskQueue<E>::push(E t) {
   }
 }
 
-template<class E> inline bool GenericTaskQueue<E>::pop_local(E& t) {
+template<class E, unsigned int N> inline bool
+GenericTaskQueue<E, N>::pop_local(E& t) {
   uint localBot = _bottom;
   // This value cannot be N-1.  That can only occur as a result of
   // the assignment to bottom in this method.  If it does, this method
@@ -497,7 +517,7 @@ template<class E> inline bool GenericTaskQueue<E>::pop_local(E& t) {
   // This is necessary to prevent any read below from being reordered
   // before the store just above.
   OrderAccess::fence();
-  t = _elems[localBot];
+  const_cast<E&>(t = _elems[localBot]);
   // This is a second read of "age"; the "size()" above is the first.
   // If there's still at least one element in the queue, based on the
   // "_bottom" and "age" we've read, then there can be no interference with
@@ -515,16 +535,22 @@ template<class E> inline bool GenericTaskQueue<E>::pop_local(E& t) {
 
 typedef oop Task;
 typedef GenericTaskQueue<Task>            OopTaskQueue;
-typedef GenericTaskQueueSet<Task>      OopTaskQueueSet;
+typedef GenericTaskQueueSet<OopTaskQueue> OopTaskQueueSet;
 
-
+#ifdef _MSC_VER
-#define COMPRESSED_OOP_MASK  1
+#pragma warning(push)
+// warning C4522: multiple assignment operators specified
+#pragma warning(disable:4522)
+#endif
 
 // This is a container class for either an oop* or a narrowOop*.
 // Both are pushed onto a task queue and the consumer will test is_narrow()
 // to determine which should be processed.
 class StarTask {
   void*  _holder;        // either union oop* or narrowOop*
+
+  enum { COMPRESSED_OOP_MASK = 1 };
+
  public:
   StarTask(narrowOop* p) {
     assert(((uintptr_t)p & COMPRESSED_OOP_MASK) == 0, "Information loss!");
@@ -540,20 +566,61 @@ class StarTask {
     return (narrowOop*)((uintptr_t)_holder & ~COMPRESSED_OOP_MASK);
   }
 
-  // Operators to preserve const/volatile in assignments required by gcc
+  StarTask& operator=(const StarTask& t) {
-  void operator=(const volatile StarTask& t) volatile { _holder = t._holder; }
+    _holder = t._holder;
+    return *this;
+  }
+  volatile StarTask& operator=(const volatile StarTask& t) volatile {
+    _holder = t._holder;
+    return *this;
+  }
 
   bool is_narrow() const {
     return (((uintptr_t)_holder & COMPRESSED_OOP_MASK) != 0);
   }
 };
 
+class ObjArrayTask
+{
+public:
+  ObjArrayTask(oop o = NULL, int idx = 0): _obj(o), _index(idx) { }
+  ObjArrayTask(oop o, size_t idx): _obj(o), _index(int(idx)) {
+    assert(idx <= size_t(max_jint), "too big");
+  }
+  ObjArrayTask(const ObjArrayTask& t): _obj(t._obj), _index(t._index) { }
+
+  ObjArrayTask& operator =(const ObjArrayTask& t) {
+    _obj = t._obj;
+    _index = t._index;
+    return *this;
+  }
+  volatile ObjArrayTask&
+  operator =(const volatile ObjArrayTask& t) volatile {
+    _obj = t._obj;
+    _index = t._index;
+    return *this;
+  }
+
+  inline oop obj()   const { return _obj; }
+  inline int index() const { return _index; }
+
+  DEBUG_ONLY(bool is_valid() const); // Tasks to be pushed/popped must be valid.
+
+private:
+  oop _obj;
+  int _index;
+};
+
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif
+
 typedef GenericTaskQueue<StarTask>            OopStarTaskQueue;
-typedef GenericTaskQueueSet<StarTask>  OopStarTaskQueueSet;
+typedef GenericTaskQueueSet<OopStarTaskQueue> OopStarTaskQueueSet;
 
 typedef size_t RegionTask;  // index for region
 typedef GenericTaskQueue<RegionTask>         RegionTaskQueue;
-typedef GenericTaskQueueSet<RegionTask> RegionTaskQueueSet;
+typedef GenericTaskQueueSet<RegionTaskQueue> RegionTaskQueueSet;
 
 class RegionTaskQueueWithOverflow: public CHeapObj {
  protected:

hotspot/test/runtime/6929067/T.java

@@ -0,0 +1,12 @@
+public class T
+{
+  public static boolean foo(boolean bar)
+  {
+    return bar;
+  }
+
+  public static void printIt()
+  {
+    System.out.println("Hello");
+  }
+}

hotspot/test/runtime/6929067/Test6929067.sh

@@ -0,0 +1,60 @@
+#!/bin/sh
+
+##
+## @test Test6929067.sh
+## @bug 6929067
+## @summary Stack guard pages should be removed when thread is detached
+## @run shell Test6929067.sh
+##
+
+if [ "${TESTSRC}" = "" ]
+then TESTSRC=.
+fi
+
+if [ "${TESTJAVA}" = "" ]
+then
+  PARENT=`dirname \`which java\``
+  TESTJAVA=`dirname ${PARENT}`
+  echo "TESTJAVA not set, selecting " ${TESTJAVA}
+  echo "If this is incorrect, try setting the variable manually."
+fi
+
+BIT_FLAG=""
+
+# set platform-dependent variables
+OS=`uname -s`
+case "$OS" in
+  Linux)
+    NULL=/dev/null
+    PS=":"
+    FS="/"
+    ;;
+  SunOS | Windows_* )
+    NULL=NUL
+    PS=";"
+    FS="\\"
+    echo "Test passed; only valid for Linux"
+    exit 0;
+    ;;
+  * )
+    echo "Unrecognized system!"
+    exit 1;
+    ;;
+esac
+
+LD_LIBRARY_PATH=.:${TESTJAVA}/jre/lib/i386/client:/usr/openwin/lib:/usr/dt/lib:/usr/lib:$LD_LIBRARY_PATH
+export LD_LIBRARY_PATH
+
+THIS_DIR=`pwd`
+
+cp ${TESTSRC}${FS}invoke.c ${THIS_DIR}
+cp ${TESTSRC}${FS}T.java ${THIS_DIR}
+
+
+${TESTJAVA}${FS}bin${FS}java ${BIT_FLAG} -fullversion
+
+${TESTJAVA}${FS}bin${FS}javac T.java
+
+gcc -o invoke -I${TESTJAVA}/include -I${TESTJAVA}/include/linux invoke.c ${TESTJAVA}/jre/lib/i386/client/libjvm.so
+./invoke
+exit $?

hotspot/test/runtime/6929067/invoke.c

@@ -0,0 +1,90 @@
+#include <assert.h>
+#include <jni.h>
+#include <alloca.h>
+
+#include <pthread.h>
+
+union env_union
+{
+  void *void_env;
+  JNIEnv *jni_env;
+};
+
+union env_union tmp;
+JNIEnv* env;
+JavaVM* jvm;
+JavaVMInitArgs vm_args;
+JavaVMOption options[1];
+jclass class_id;
+jmethodID method_id;
+jint result;
+
+long product(unsigned long n, unsigned long m) {
+    if (m == 1) {
+      return n;
+    } else {
+      int *p = alloca(sizeof (int));
+      *p = n;
+      return product (n, m-1) + *p;
+    }
+}
+
+void *
+floobydust (void *p)
+{
+  (*jvm)->AttachCurrentThread(jvm, &tmp.void_env, NULL);
+  env = tmp.jni_env;
+
+  class_id = (*env)->FindClass (env, "T");
+  assert (class_id);
+
+  method_id = (*env)->GetStaticMethodID (env, class_id, "printIt", "()V");
+  assert (method_id);
+
+  (*env)->CallStaticVoidMethod (env, class_id, method_id, NULL);
+
+  (*jvm)->DetachCurrentThread(jvm);
+
+  printf("%ld\n", product(5000,5000));
+
+  (*jvm)->AttachCurrentThread(jvm, &tmp.void_env, NULL);
+  env = tmp.jni_env;
+
+  class_id = (*env)->FindClass (env, "T");
+  assert (class_id);
+
+  method_id = (*env)->GetStaticMethodID (env, class_id, "printIt", "()V");
+  assert (method_id);
+
+  (*env)->CallStaticVoidMethod (env, class_id, method_id, NULL);
+
+  (*jvm)->DetachCurrentThread(jvm);
+
+  printf("%ld\n", product(5000,5000));
+
+  return NULL;
+}
+
+int
+main (int argc, const char** argv)
+{
+  options[0].optionString = "-Xss320k";
+
+  vm_args.version = JNI_VERSION_1_2;
+  vm_args.ignoreUnrecognized = JNI_TRUE;
+  vm_args.options = options;
+  vm_args.nOptions = 1;
+
+  result = JNI_CreateJavaVM (&jvm, &tmp.void_env, &vm_args);
+  assert (result >= 0);
+
+  env = tmp.jni_env;
+
+  floobydust (NULL);
+
+  pthread_t thr;
+  pthread_create (&thr, NULL, floobydust, NULL);
+  pthread_join (thr, NULL);
+
+  return 0;
+}

jdk/.hgtags

@@ -61,3 +61,4 @@ e6a5d095c356a547cf5b3c8885885aca5e91e09b jdk7-b77
 7cb9388bb1a16365fa5118c5efa38b1cd58be40d jdk7-b84
 b396584a3e64988839cca21ea1f7fbdcc9248783 jdk7-b85
 eae6e9ab26064d9ba0e7665dd646a1fd2506fcc1 jdk7-b86
+2cafbbe9825e911a6ca6c17d9a18eb1f0bf0873c jdk7-b87