changeset 2617:688202ef6306

Merge
author never
date Thu, 12 May 2011 19:39:58 -0700
parents 3fd6f2d58ef3 153957c9207b
children a7ccd5419f48 2aa9ddbb9e60
files src/share/vm/classfile/vmSymbols.hpp src/share/vm/runtime/globals.hpp
diffstat 33 files changed, 780 insertions(+), 197 deletions(-) [+]
line wrap: on
line diff
--- a/make/solaris/makefiles/saproc.make	Thu May 12 16:24:44 2011 -0700
+++ b/make/solaris/makefiles/saproc.make	Thu May 12 19:39:58 2011 -0700
@@ -57,24 +57,27 @@
 endif
 
 # The libproc Pstack_iter() interface changed in Nevada-B159.
-# This logic needs to match
+# Use 'uname -r -v' to determine the Solaris version as per
+# Solaris Nevada team request. This logic needs to match:
 # agent/src/os/solaris/proc/saproc.cpp: set_has_newer_Pstack_iter():
 #   - skip SunOS 4 or older
 #   - skip Solaris 10 or older
-#   - skip two digit Nevada builds
-#   - skip three digit Nevada builds thru 149
-#   - skip Nevada builds 150-158
+#   - skip two digit internal Nevada builds
+#   - skip three digit internal Nevada builds thru 149
+#   - skip internal Nevada builds 150-158
+#   - if not skipped, print define for Nevada-B159 or later
 SOLARIS_11_B159_OR_LATER := \
 $(shell uname -r -v \
-    | sed -n ' \
-          /^[0-3]\. /b \
-          /^5\.[0-9] /b \
-          /^5\.10 /b \
-          / snv_[0-9][0-9]$/b \
-          / snv_[01][0-4][0-9]$/b \
-          / snv_15[0-8]$/b \
-          s/.*/-DSOLARIS_11_B159_OR_LATER/p \
-          ')
+    | sed -n \
+          -e '/^[0-4]\. /b' \
+          -e '/^5\.[0-9] /b' \
+          -e '/^5\.10 /b' \
+          -e '/ snv_[0-9][0-9]$/b' \
+          -e '/ snv_[01][0-4][0-9]$/b' \
+          -e '/ snv_15[0-8]$/b' \
+          -e 's/.*/-DSOLARIS_11_B159_OR_LATER/' \
+          -e 'p' \
+          )
 
 # Uncomment the following to simulate building on Nevada-B159 or later
 # when actually building on Nevada-B158 or earlier:
--- a/src/os/linux/vm/os_linux.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/os/linux/vm/os_linux.cpp	Thu May 12 19:39:58 2011 -0700
@@ -2850,7 +2850,7 @@
         char chars[257];
         long x = 0;
         if (fgets(chars, sizeof(chars), fp)) {
-          if (sscanf(chars, "%lx-%*lx", &x) == 1
+          if (sscanf(chars, "%lx-%*x", &x) == 1
               && x == (long)p) {
             if (strstr (chars, "hugepage")) {
               result = true;
--- a/src/share/vm/classfile/vmSymbols.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/classfile/vmSymbols.hpp	Thu May 12 19:39:58 2011 -0700
@@ -473,6 +473,13 @@
   template(sun_management_ManagementFactory,           "sun/management/ManagementFactory")                        \
   template(sun_management_Sensor,                      "sun/management/Sensor")                                   \
   template(sun_management_Agent,                       "sun/management/Agent")                                    \
+  template(sun_management_GarbageCollectorImpl,        "sun/management/GarbageCollectorImpl")                     \
+  template(getGcInfoBuilder_name,                      "getGcInfoBuilder")                                        \
+  template(getGcInfoBuilder_signature,                 "()Lsun/management/GcInfoBuilder;")                        \
+  template(com_sun_management_GcInfo,                  "com/sun/management/GcInfo")                               \
+  template(com_sun_management_GcInfo_constructor_signature, "(Lsun/management/GcInfoBuilder;JJJ[Ljava/lang/management/MemoryUsage;[Ljava/lang/management/MemoryUsage;[Ljava/lang/Object;)V") \
+  template(createGCNotification_name,                  "createGCNotification")                                    \
+  template(createGCNotification_signature,             "(JLjava/lang/String;Ljava/lang/String;Ljava/lang/String;Lcom/sun/management/GcInfo;)V") \
   template(createMemoryPoolMBean_name,                 "createMemoryPoolMBean")                                   \
   template(createMemoryManagerMBean_name,              "createMemoryManagerMBean")                                \
   template(createGarbageCollectorMBean_name,           "createGarbageCollectorMBean")                             \
@@ -490,6 +497,7 @@
   template(java_lang_management_MemoryPoolMXBean,      "java/lang/management/MemoryPoolMXBean")                   \
   template(java_lang_management_MemoryManagerMXBean,   "java/lang/management/MemoryManagerMXBean")                \
   template(java_lang_management_GarbageCollectorMXBean,"java/lang/management/GarbageCollectorMXBean")             \
+  template(gcInfoBuilder_name,                         "gcInfoBuilder")                                           \
   template(createMemoryPool_name,                      "createMemoryPool")                                        \
   template(createMemoryManager_name,                   "createMemoryManager")                                     \
   template(createGarbageCollector_name,                "createGarbageCollector")                                  \
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Thu May 12 19:39:58 2011 -0700
@@ -2026,7 +2026,7 @@
   }
 
   {
-    TraceCMSMemoryManagerStats();
+    TraceCMSMemoryManagerStats tmms(gch->gc_cause());
   }
   GenMarkSweep::invoke_at_safepoint(_cmsGen->level(),
     ref_processor(), clear_all_soft_refs);
@@ -3479,7 +3479,7 @@
 void CMSCollector::checkpointRootsInitial(bool asynch) {
   assert(_collectorState == InitialMarking, "Wrong collector state");
   check_correct_thread_executing();
-  TraceCMSMemoryManagerStats tms(_collectorState);
+  TraceCMSMemoryManagerStats tms(_collectorState,GenCollectedHeap::heap()->gc_cause());
 
   ReferenceProcessor* rp = ref_processor();
   SpecializationStats::clear();
@@ -4858,7 +4858,8 @@
   // world is stopped at this checkpoint
   assert(SafepointSynchronize::is_at_safepoint(),
          "world should be stopped");
-  TraceCMSMemoryManagerStats tms(_collectorState);
+  TraceCMSMemoryManagerStats tms(_collectorState,GenCollectedHeap::heap()->gc_cause());
+
   verify_work_stacks_empty();
   verify_overflow_empty();
 
@@ -5993,7 +5994,7 @@
   verify_work_stacks_empty();
   verify_overflow_empty();
   increment_sweep_count();
-  TraceCMSMemoryManagerStats tms(_collectorState);
+  TraceCMSMemoryManagerStats tms(_collectorState,GenCollectedHeap::heap()->gc_cause());
 
   _inter_sweep_timer.stop();
   _inter_sweep_estimate.sample(_inter_sweep_timer.seconds());
@@ -9235,11 +9236,12 @@
   return res;
 }
 
-TraceCMSMemoryManagerStats::TraceCMSMemoryManagerStats(CMSCollector::CollectorState phase): TraceMemoryManagerStats() {
+TraceCMSMemoryManagerStats::TraceCMSMemoryManagerStats(CMSCollector::CollectorState phase, GCCause::Cause cause): TraceMemoryManagerStats() {
 
   switch (phase) {
     case CMSCollector::InitialMarking:
       initialize(true  /* fullGC */ ,
+                 cause /* cause of the GC */,
                  true  /* recordGCBeginTime */,
                  true  /* recordPreGCUsage */,
                  false /* recordPeakUsage */,
@@ -9251,6 +9253,7 @@
 
     case CMSCollector::FinalMarking:
       initialize(true  /* fullGC */ ,
+                 cause /* cause of the GC */,
                  false /* recordGCBeginTime */,
                  false /* recordPreGCUsage */,
                  false /* recordPeakUsage */,
@@ -9262,6 +9265,7 @@
 
     case CMSCollector::Sweeping:
       initialize(true  /* fullGC */ ,
+                 cause /* cause of the GC */,
                  false /* recordGCBeginTime */,
                  false /* recordPreGCUsage */,
                  true  /* recordPeakUsage */,
@@ -9277,8 +9281,9 @@
 }
 
 // when bailing out of cms in concurrent mode failure
-TraceCMSMemoryManagerStats::TraceCMSMemoryManagerStats(): TraceMemoryManagerStats() {
+TraceCMSMemoryManagerStats::TraceCMSMemoryManagerStats(GCCause::Cause cause): TraceMemoryManagerStats() {
   initialize(true /* fullGC */ ,
+             cause /* cause of the GC */,
              true /* recordGCBeginTime */,
              true /* recordPreGCUsage */,
              true /* recordPeakUsage */,
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.hpp	Thu May 12 19:39:58 2011 -0700
@@ -1895,8 +1895,8 @@
 class TraceCMSMemoryManagerStats : public TraceMemoryManagerStats {
 
  public:
-  TraceCMSMemoryManagerStats(CMSCollector::CollectorState phase);
-  TraceCMSMemoryManagerStats();
+  TraceCMSMemoryManagerStats(CMSCollector::CollectorState phase, GCCause::Cause cause);
+  TraceCMSMemoryManagerStats(GCCause::Cause cause);
 };
 
 
--- a/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Thu May 12 19:39:58 2011 -0700
@@ -1162,7 +1162,7 @@
                 PrintGC, true, gclog_or_tty);
 
     TraceCollectorStats tcs(g1mm()->full_collection_counters());
-    TraceMemoryManagerStats tms(true /* fullGC */);
+    TraceMemoryManagerStats tms(true /* fullGC */, gc_cause());
 
     double start = os::elapsedTime();
     g1_policy()->record_full_collection_start();
@@ -3202,7 +3202,7 @@
     TraceTime t(verbose_str, PrintGC && !PrintGCDetails, true, gclog_or_tty);
 
     TraceCollectorStats tcs(g1mm()->incremental_collection_counters());
-    TraceMemoryManagerStats tms(false /* fullGC */);
+    TraceMemoryManagerStats tms(false /* fullGC */, gc_cause());
 
     // If the secondary_free_list is not empty, append it to the
     // free_list. No need to wait for the cleanup operation to finish;
--- a/src/share/vm/gc_implementation/parNew/parCardTableModRefBS.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/parNew/parCardTableModRefBS.cpp	Thu May 12 19:39:58 2011 -0700
@@ -29,13 +29,14 @@
 #include "memory/sharedHeap.hpp"
 #include "memory/space.inline.hpp"
 #include "memory/universe.hpp"
+#include "oops/oop.inline.hpp"
 #include "runtime/java.hpp"
 #include "runtime/mutexLocker.hpp"
 #include "runtime/virtualspace.hpp"
 
 void CardTableModRefBS::non_clean_card_iterate_parallel_work(Space* sp, MemRegion mr,
-                                                             DirtyCardToOopClosure* dcto_cl,
-                                                             ClearNoncleanCardWrapper* cl,
+                                                             OopsInGenClosure* cl,
+                                                             CardTableRS* ct,
                                                              int n_threads) {
   assert(n_threads > 0, "Error: expected n_threads > 0");
   assert((n_threads == 1 && ParallelGCThreads == 0) ||
@@ -49,14 +50,14 @@
                           lowest_non_clean_base_chunk_index,
                           lowest_non_clean_chunk_size);
 
-  int n_strides = n_threads * StridesPerThread;
+  int n_strides = n_threads * ParGCStridesPerThread;
   SequentialSubTasksDone* pst = sp->par_seq_tasks();
   pst->set_n_threads(n_threads);
   pst->set_n_tasks(n_strides);
 
   int stride = 0;
   while (!pst->is_task_claimed(/* reference */ stride)) {
-    process_stride(sp, mr, stride, n_strides, dcto_cl, cl,
+    process_stride(sp, mr, stride, n_strides, cl, ct,
                    lowest_non_clean,
                    lowest_non_clean_base_chunk_index,
                    lowest_non_clean_chunk_size);
@@ -79,13 +80,13 @@
 process_stride(Space* sp,
                MemRegion used,
                jint stride, int n_strides,
-               DirtyCardToOopClosure* dcto_cl,
-               ClearNoncleanCardWrapper* cl,
+               OopsInGenClosure* cl,
+               CardTableRS* ct,
                jbyte** lowest_non_clean,
                uintptr_t lowest_non_clean_base_chunk_index,
                size_t    lowest_non_clean_chunk_size) {
-  // We don't have to go downwards here; it wouldn't help anyway,
-  // because of parallelism.
+  // We go from higher to lower addresses here; it wouldn't help that much
+  // because of the strided parallelism pattern used here.
 
   // Find the first card address of the first chunk in the stride that is
   // at least "bottom" of the used region.
@@ -98,25 +99,35 @@
   if ((uintptr_t)stride >= start_chunk_stride_num) {
     chunk_card_start = (jbyte*)(start_card +
                                 (stride - start_chunk_stride_num) *
-                                CardsPerStrideChunk);
+                                ParGCCardsPerStrideChunk);
   } else {
     // Go ahead to the next chunk group boundary, then to the requested stride.
     chunk_card_start = (jbyte*)(start_card +
                                 (n_strides - start_chunk_stride_num + stride) *
-                                CardsPerStrideChunk);
+                                ParGCCardsPerStrideChunk);
   }
 
   while (chunk_card_start < end_card) {
-    // We don't have to go downwards here; it wouldn't help anyway,
-    // because of parallelism.  (We take care with "min_done"; see below.)
+    // Even though we go from lower to higher addresses below, the
+    // strided parallelism can interleave the actual processing of the
+    // dirty pages in various ways. For a specific chunk within this
+    // stride, we take care to avoid double scanning or missing a card
+    // by suitably initializing the "min_done" field in process_chunk_boundaries()
+    // below, together with the dirty region extension accomplished in
+    // DirtyCardToOopClosure::do_MemRegion().
+    jbyte*    chunk_card_end = chunk_card_start + ParGCCardsPerStrideChunk;
     // Invariant: chunk_mr should be fully contained within the "used" region.
-    jbyte*    chunk_card_end = chunk_card_start + CardsPerStrideChunk;
     MemRegion chunk_mr       = MemRegion(addr_for(chunk_card_start),
                                          chunk_card_end >= end_card ?
                                            used.end() : addr_for(chunk_card_end));
     assert(chunk_mr.word_size() > 0, "[chunk_card_start > used_end)");
     assert(used.contains(chunk_mr), "chunk_mr should be subset of used");
 
+    DirtyCardToOopClosure* dcto_cl = sp->new_dcto_cl(cl, precision(),
+                                                     cl->gen_boundary());
+    ClearNoncleanCardWrapper clear_cl(dcto_cl, ct);
+
+
     // Process the chunk.
     process_chunk_boundaries(sp,
                              dcto_cl,
@@ -126,17 +137,30 @@
                              lowest_non_clean_base_chunk_index,
                              lowest_non_clean_chunk_size);
 
+    // We want the LNC array updates above in process_chunk_boundaries
+    // to be visible before any of the card table value changes as a
+    // result of the dirty card iteration below.
+    OrderAccess::storestore();
+
     // We do not call the non_clean_card_iterate_serial() version because
-    // we want to clear the cards, and the ClearNoncleanCardWrapper closure
-    // itself does the work of finding contiguous dirty ranges of cards to
-    // process (and clear).
-    cl->do_MemRegion(chunk_mr);
+    // we want to clear the cards: clear_cl here does the work of finding
+    // contiguous dirty ranges of cards to process and clear.
+    clear_cl.do_MemRegion(chunk_mr);
 
     // Find the next chunk of the stride.
-    chunk_card_start += CardsPerStrideChunk * n_strides;
+    chunk_card_start += ParGCCardsPerStrideChunk * n_strides;
   }
 }
 
+
+// If you want a talkative process_chunk_boundaries,
+// then #define NOISY(x) x
+#ifdef NOISY
+#error "Encountered a global preprocessor flag, NOISY, which might clash with local definition to follow"
+#else
+#define NOISY(x)
+#endif
+
 void
 CardTableModRefBS::
 process_chunk_boundaries(Space* sp,
@@ -147,126 +171,232 @@
                          uintptr_t lowest_non_clean_base_chunk_index,
                          size_t    lowest_non_clean_chunk_size)
 {
-  // We must worry about the chunk boundaries.
+  // We must worry about non-array objects that cross chunk boundaries,
+  // because such objects are both precisely and imprecisely marked:
+  // .. if the head of such an object is dirty, the entire object
+  //    needs to be scanned, under the interpretation that this
+  //    was an imprecise mark
+  // .. if the head of such an object is not dirty, we can assume
+  //    precise marking and it's efficient to scan just the dirty
+  //    cards.
+  // In either case, each scanned reference must be scanned precisely
+  // once so as to avoid cloning of a young referent. For efficiency,
+  // our closures depend on this property and do not protect against
+  // double scans.
 
-  // First, set our max_to_do:
-  HeapWord* max_to_do = NULL;
   uintptr_t cur_chunk_index = addr_to_chunk_index(chunk_mr.start());
   cur_chunk_index           = cur_chunk_index - lowest_non_clean_base_chunk_index;
 
+  NOISY(tty->print_cr("===========================================================================");)
+  NOISY(tty->print_cr(" process_chunk_boundary: Called with [" PTR_FORMAT "," PTR_FORMAT ")",
+                      chunk_mr.start(), chunk_mr.end());)
+
+  // First, set "our" lowest_non_clean entry, which would be
+  // used by the thread scanning an adjoining left chunk with
+  // a non-array object straddling the mutual boundary.
+  // Find the object that spans our boundary, if one exists.
+  // first_block is the block possibly straddling our left boundary.
+  HeapWord* first_block = sp->block_start(chunk_mr.start());
+  assert((chunk_mr.start() != used.start()) || (first_block == chunk_mr.start()),
+         "First chunk should always have a co-initial block");
+  // Does the block straddle the chunk's left boundary, and is it
+  // a non-array object?
+  if (first_block < chunk_mr.start()        // first block straddles left bdry
+      && sp->block_is_obj(first_block)      // first block is an object
+      && !(oop(first_block)->is_objArray()  // first block is not an array (arrays are precisely dirtied)
+           || oop(first_block)->is_typeArray())) {
+    // Find our least non-clean card, so that a left neighbour
+    // does not scan an object straddling the mutual boundary
+    // too far to the right, and attempt to scan a portion of
+    // that object twice.
+    jbyte* first_dirty_card = NULL;
+    jbyte* last_card_of_first_obj =
+        byte_for(first_block + sp->block_size(first_block) - 1);
+    jbyte* first_card_of_cur_chunk = byte_for(chunk_mr.start());
+    jbyte* last_card_of_cur_chunk = byte_for(chunk_mr.last());
+    jbyte* last_card_to_check =
+      (jbyte*) MIN2((intptr_t) last_card_of_cur_chunk,
+                    (intptr_t) last_card_of_first_obj);
+    // Note that this does not need to go beyond our last card
+    // if our first object completely straddles this chunk.
+    for (jbyte* cur = first_card_of_cur_chunk;
+         cur <= last_card_to_check; cur++) {
+      jbyte val = *cur;
+      if (card_will_be_scanned(val)) {
+        first_dirty_card = cur; break;
+      } else {
+        assert(!card_may_have_been_dirty(val), "Error");
+      }
+    }
+    if (first_dirty_card != NULL) {
+      NOISY(tty->print_cr(" LNC: Found a dirty card at " PTR_FORMAT " in current chunk",
+                    first_dirty_card);)
+      assert(0 <= cur_chunk_index && cur_chunk_index < lowest_non_clean_chunk_size,
+             "Bounds error.");
+      assert(lowest_non_clean[cur_chunk_index] == NULL,
+             "Write exactly once : value should be stable hereafter for this round");
+      lowest_non_clean[cur_chunk_index] = first_dirty_card;
+    } NOISY(else {
+      tty->print_cr(" LNC: Found no dirty card in current chunk; leaving LNC entry NULL");
+      // In the future, we could have this thread look for a non-NULL value to copy from its
+      // right neighbour (up to the end of the first object).
+      if (last_card_of_cur_chunk < last_card_of_first_obj) {
+        tty->print_cr(" LNC: BEWARE!!! first obj straddles past right end of chunk:\n"
+                      "   might be efficient to get value from right neighbour?");
+      }
+    })
+  } else {
+    // In this case we can help our neighbour by just asking them
+    // to stop at our first card (even though it may not be dirty).
+    NOISY(tty->print_cr(" LNC: first block is not a non-array object; setting LNC to first card of current chunk");)
+    assert(lowest_non_clean[cur_chunk_index] == NULL, "Write once : value should be stable hereafter");
+    jbyte* first_card_of_cur_chunk = byte_for(chunk_mr.start());
+    lowest_non_clean[cur_chunk_index] = first_card_of_cur_chunk;
+  }
+  NOISY(tty->print_cr(" process_chunk_boundary: lowest_non_clean[" INTPTR_FORMAT "] = " PTR_FORMAT
+                "   which corresponds to the heap address " PTR_FORMAT,
+                cur_chunk_index, lowest_non_clean[cur_chunk_index],
+                (lowest_non_clean[cur_chunk_index] != NULL)
+                ? addr_for(lowest_non_clean[cur_chunk_index])
+                : NULL);)
+  NOISY(tty->print_cr("---------------------------------------------------------------------------");)
+
+  // Next, set our own max_to_do, which will strictly/exclusively bound
+  // the highest address that we will scan past the right end of our chunk.
+  HeapWord* max_to_do = NULL;
   if (chunk_mr.end() < used.end()) {
-    // This is not the last chunk in the used region.  What is the last
-    // object?
-    HeapWord* last_block = sp->block_start(chunk_mr.end());
+    // This is not the last chunk in the used region.
+    // What is our last block? We check the first block of
+    // the next (right) chunk rather than strictly check our last block
+    // because it's potentially more efficient to do so.
+    HeapWord* const last_block = sp->block_start(chunk_mr.end());
     assert(last_block <= chunk_mr.end(), "In case this property changes.");
-    if (last_block == chunk_mr.end()
-        || !sp->block_is_obj(last_block)) {
+    if ((last_block == chunk_mr.end())     // our last block does not straddle boundary
+        || !sp->block_is_obj(last_block)   // last_block isn't an object
+        || oop(last_block)->is_objArray()  // last_block is an array (precisely marked)
+        || oop(last_block)->is_typeArray()) {
       max_to_do = chunk_mr.end();
-
+      NOISY(tty->print_cr(" process_chunk_boundary: Last block on this card is not a non-array object;\n"
+                         "   max_to_do left at " PTR_FORMAT, max_to_do);)
     } else {
-      // It is an object and starts before the end of the current chunk.
+      assert(last_block < chunk_mr.end(), "Tautology");
+      // It is a non-array object that straddles the right boundary of this chunk.
       // last_obj_card is the card corresponding to the start of the last object
       // in the chunk.  Note that the last object may not start in
       // the chunk.
-      jbyte* last_obj_card = byte_for(last_block);
-      if (!card_may_have_been_dirty(*last_obj_card)) {
-        // The card containing the head is not dirty.  Any marks in
+      jbyte* const last_obj_card = byte_for(last_block);
+      const jbyte val = *last_obj_card;
+      if (!card_will_be_scanned(val)) {
+        assert(!card_may_have_been_dirty(val), "Error");
+        // The card containing the head is not dirty.  Any marks on
         // subsequent cards still in this chunk must have been made
-        // precisely; we can cap processing at the end.
+        // precisely; we can cap processing at the end of our chunk.
         max_to_do = chunk_mr.end();
+        NOISY(tty->print_cr(" process_chunk_boundary: Head of last object on this card is not dirty;\n"
+                            "   max_to_do left at " PTR_FORMAT,
+                            max_to_do);)
       } else {
         // The last object must be considered dirty, and extends onto the
         // following chunk.  Look for a dirty card in that chunk that will
         // bound our processing.
         jbyte* limit_card = NULL;
-        size_t last_block_size = sp->block_size(last_block);
-        jbyte* last_card_of_last_obj =
+        const size_t last_block_size = sp->block_size(last_block);
+        jbyte* const last_card_of_last_obj =
           byte_for(last_block + last_block_size - 1);
-        jbyte* first_card_of_next_chunk = byte_for(chunk_mr.end());
+        jbyte* const first_card_of_next_chunk = byte_for(chunk_mr.end());
         // This search potentially goes a long distance looking
-        // for the next card that will be scanned.  For example,
-        // an object that is an array of primitives will not
-        // have any cards covering regions interior to the array
-        // that will need to be scanned. The scan can be terminated
-        // at the last card of the next chunk.  That would leave
-        // limit_card as NULL and would result in "max_to_do"
-        // being set with the LNC value or with the end
-        // of the last block.
-        jbyte* last_card_of_next_chunk = first_card_of_next_chunk +
-          CardsPerStrideChunk;
-        assert(byte_for(chunk_mr.end()) - byte_for(chunk_mr.start())
-          == CardsPerStrideChunk, "last card of next chunk may be wrong");
-        jbyte* last_card_to_check = (jbyte*) MIN2(last_card_of_last_obj,
-                                                  last_card_of_next_chunk);
+        // for the next card that will be scanned, terminating
+        // at the end of the last_block, if no earlier dirty card
+        // is found.
+        assert(byte_for(chunk_mr.end()) - byte_for(chunk_mr.start()) == ParGCCardsPerStrideChunk,
+               "last card of next chunk may be wrong");
         for (jbyte* cur = first_card_of_next_chunk;
-             cur <= last_card_to_check; cur++) {
-          if (card_will_be_scanned(*cur)) {
+             cur <= last_card_of_last_obj; cur++) {
+          const jbyte val = *cur;
+          if (card_will_be_scanned(val)) {
+            NOISY(tty->print_cr(" Found a non-clean card " PTR_FORMAT " with value 0x%x",
+                                cur, (int)val);)
             limit_card = cur; break;
+          } else {
+            assert(!card_may_have_been_dirty(val), "Error: card can't be skipped");
           }
         }
-        assert(0 <= cur_chunk_index+1 &&
-               cur_chunk_index+1 < lowest_non_clean_chunk_size,
+        if (limit_card != NULL) {
+          max_to_do = addr_for(limit_card);
+          assert(limit_card != NULL && max_to_do != NULL, "Error");
+          NOISY(tty->print_cr(" process_chunk_boundary: Found a dirty card at " PTR_FORMAT
+                        "   max_to_do set at " PTR_FORMAT " which is before end of last block in chunk: "
+                        PTR_FORMAT " + " PTR_FORMAT " = " PTR_FORMAT,
+                        limit_card, max_to_do, last_block, last_block_size, (last_block+last_block_size));)
+        } else {
+          // The following is a pessimistic value, because it's possible
+          // that a dirty card on a subsequent chunk has been cleared by
+          // the time we get to look at it; we'll correct for that further below,
+          // using the LNC array which records the least non-clean card
+          // before cards were cleared in a particular chunk.
+          limit_card = last_card_of_last_obj;
+          max_to_do = last_block + last_block_size;
+          assert(limit_card != NULL && max_to_do != NULL, "Error");
+          NOISY(tty->print_cr(" process_chunk_boundary: Found no dirty card before end of last block in chunk\n"
+                              "   Setting limit_card to " PTR_FORMAT
+                              " and max_to_do " PTR_FORMAT " + " PTR_FORMAT " = " PTR_FORMAT,
+                              limit_card, last_block, last_block_size, max_to_do);)
+        }
+        assert(0 < cur_chunk_index+1 && cur_chunk_index+1 < lowest_non_clean_chunk_size,
                "Bounds error.");
-        // LNC for the next chunk
-        jbyte* lnc_card = lowest_non_clean[cur_chunk_index+1];
-        if (limit_card == NULL) {
-          limit_card = lnc_card;
+        // It is possible that a dirty card for the last object may have been
+        // cleared before we had a chance to examine it. In that case, the value
+        // will have been logged in the LNC for that chunk.
+        // We need to examine as many chunks to the right as this object
+        // covers.
+        const uintptr_t last_chunk_index_to_check = addr_to_chunk_index(last_block + last_block_size - 1)
+                                                    - lowest_non_clean_base_chunk_index;
+        DEBUG_ONLY(const uintptr_t last_chunk_index = addr_to_chunk_index(used.last())
+                                                      - lowest_non_clean_base_chunk_index;)
+        assert(last_chunk_index_to_check <= last_chunk_index,
+               err_msg("Out of bounds: last_chunk_index_to_check " INTPTR_FORMAT
+                       " exceeds last_chunk_index " INTPTR_FORMAT,
+                       last_chunk_index_to_check, last_chunk_index));
+        for (uintptr_t lnc_index = cur_chunk_index + 1;
+             lnc_index <= last_chunk_index_to_check;
+             lnc_index++) {
+          jbyte* lnc_card = lowest_non_clean[lnc_index];
+          if (lnc_card != NULL) {
+            // we can stop at the first non-NULL entry we find
+            if (lnc_card <= limit_card) {
+              NOISY(tty->print_cr(" process_chunk_boundary: LNC card " PTR_FORMAT " is lower than limit_card " PTR_FORMAT,
+                                  "   max_to_do will be lowered to " PTR_FORMAT " from " PTR_FORMAT,
+                                  lnc_card, limit_card, addr_for(lnc_card), max_to_do);)
+              limit_card = lnc_card;
+              max_to_do = addr_for(limit_card);
+              assert(limit_card != NULL && max_to_do != NULL, "Error");
+            }
+            // In any case, we break now
+            break;
+          }  // else continue to look for a non-NULL entry if any
         }
-        if (limit_card != NULL) {
-          if (lnc_card != NULL) {
-            limit_card = (jbyte*)MIN2((intptr_t)limit_card,
-                                      (intptr_t)lnc_card);
-          }
-          max_to_do = addr_for(limit_card);
-        } else {
-          max_to_do = last_block + last_block_size;
-        }
+        assert(limit_card != NULL && max_to_do != NULL, "Error");
       }
+      assert(max_to_do != NULL, "OOPS 1 !");
     }
-    assert(max_to_do != NULL, "OOPS!");
+    assert(max_to_do != NULL, "OOPS 2!");
   } else {
     max_to_do = used.end();
+    NOISY(tty->print_cr(" process_chunk_boundary: Last chunk of this space;\n"
+                  "   max_to_do left at " PTR_FORMAT,
+                  max_to_do);)
   }
+  assert(max_to_do != NULL, "OOPS 3!");
   // Now we can set the closure we're using so it doesn't to beyond
   // max_to_do.
   dcto_cl->set_min_done(max_to_do);
 #ifndef PRODUCT
   dcto_cl->set_last_bottom(max_to_do);
 #endif
+  NOISY(tty->print_cr("===========================================================================\n");)
+}
 
-  // Now we set *our" lowest_non_clean entry.
-  // Find the object that spans our boundary, if one exists.
-  // Nothing to do on the first chunk.
-  if (chunk_mr.start() > used.start()) {
-    // first_block is the block possibly spanning the chunk start
-    HeapWord* first_block = sp->block_start(chunk_mr.start());
-    // Does the block span the start of the chunk and is it
-    // an object?
-    if (first_block < chunk_mr.start() &&
-        sp->block_is_obj(first_block)) {
-      jbyte* first_dirty_card = NULL;
-      jbyte* last_card_of_first_obj =
-          byte_for(first_block + sp->block_size(first_block) - 1);
-      jbyte* first_card_of_cur_chunk = byte_for(chunk_mr.start());
-      jbyte* last_card_of_cur_chunk = byte_for(chunk_mr.last());
-      jbyte* last_card_to_check =
-        (jbyte*) MIN2((intptr_t) last_card_of_cur_chunk,
-                      (intptr_t) last_card_of_first_obj);
-      for (jbyte* cur = first_card_of_cur_chunk;
-           cur <= last_card_to_check; cur++) {
-        if (card_will_be_scanned(*cur)) {
-          first_dirty_card = cur; break;
-        }
-      }
-      if (first_dirty_card != NULL) {
-        assert(0 <= cur_chunk_index &&
-                 cur_chunk_index < lowest_non_clean_chunk_size,
-               "Bounds error.");
-        lowest_non_clean[cur_chunk_index] = first_dirty_card;
-      }
-    }
-  }
-}
+#undef NOISY
 
 void
 CardTableModRefBS::
@@ -283,8 +413,8 @@
   // LNC array for the covered region.  Any later expansion can't affect
   // the used_at_save_marks region.
   // (I observed a bug in which the first thread to execute this would
-  // resize, and then it would cause "expand_and_allocates" that would
-  // Increase the number of chunks in the covered region.  Then a second
+  // resize, and then it would cause "expand_and_allocate" that would
+  // increase the number of chunks in the covered region.  Then a second
   // thread would come and execute this, see that the size didn't match,
   // and free and allocate again.  So the first thread would be using a
   // freed "_lowest_non_clean" array.)
--- a/src/share/vm/gc_implementation/parNew/parOopClosures.inline.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/parNew/parOopClosures.inline.hpp	Thu May 12 19:39:58 2011 -0700
@@ -77,7 +77,23 @@
   if (!oopDesc::is_null(heap_oop)) {
     oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
     if ((HeapWord*)obj < _boundary) {
-      assert(!_g->to()->is_in_reserved(obj), "Scanning field twice?");
+#ifndef PRODUCT
+      if (_g->to()->is_in_reserved(obj)) {
+        tty->print_cr("Scanning field (" PTR_FORMAT ") twice?", p);
+        GenCollectedHeap* gch =  (GenCollectedHeap*)Universe::heap();
+        Space* sp = gch->space_containing(p);
+        oop obj = oop(sp->block_start(p));
+        assert((HeapWord*)obj < (HeapWord*)p, "Error");
+        tty->print_cr("Object: " PTR_FORMAT, obj);
+        tty->print_cr("-------");
+        obj->print();
+        tty->print_cr("-----");
+        tty->print_cr("Heap:");
+        tty->print_cr("-----");
+        gch->print();
+        ShouldNotReachHere();
+      }
+#endif
       // OK, we need to ensure that it is copied.
       // We read the klass and mark in this order, so that we can reliably
       // get the size of the object: if the mark we read is not a
--- a/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp	Thu May 12 19:39:58 2011 -0700
@@ -173,7 +173,7 @@
     TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
     TraceTime t1(gc_cause_str, PrintGC, !PrintGCDetails, gclog_or_tty);
     TraceCollectorStats tcs(counters());
-    TraceMemoryManagerStats tms(true /* Full GC */);
+    TraceMemoryManagerStats tms(true /* Full GC */,gc_cause);
 
     if (TraceGen1Time) accumulated_time()->start();
 
--- a/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp	Thu May 12 19:39:58 2011 -0700
@@ -2057,7 +2057,7 @@
     TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
     TraceTime t1(gc_cause_str, PrintGC, !PrintGCDetails, gclog_or_tty);
     TraceCollectorStats tcs(counters());
-    TraceMemoryManagerStats tms(true /* Full GC */);
+    TraceMemoryManagerStats tms(true /* Full GC */,gc_cause);
 
     if (TraceGen1Time) accumulated_time()->start();
 
--- a/src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp	Thu May 12 19:39:58 2011 -0700
@@ -322,7 +322,7 @@
     TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
     TraceTime t1("GC", PrintGC, !PrintGCDetails, gclog_or_tty);
     TraceCollectorStats tcs(counters());
-    TraceMemoryManagerStats tms(false /* not full GC */);
+    TraceMemoryManagerStats tms(false /* not full GC */,gc_cause);
 
     if (TraceGen0Time) accumulated_time()->start();
 
--- a/src/share/vm/memory/blockOffsetTable.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/memory/blockOffsetTable.cpp	Thu May 12 19:39:58 2011 -0700
@@ -541,20 +541,33 @@
     // to go back by.
     size_t n_cards_back = entry_to_cards_back(offset);
     q -= (N_words * n_cards_back);
-    assert(q >= _sp->bottom(), "Went below bottom!");
+    assert(q >= _sp->bottom(),
+           err_msg("q = " PTR_FORMAT " crossed below bottom = " PTR_FORMAT,
+                   q, _sp->bottom()));
+    assert(q < _sp->end(),
+           err_msg("q = " PTR_FORMAT " crossed above end = " PTR_FORMAT,
+                   q, _sp->end()));
     index -= n_cards_back;
     offset = _array->offset_array(index);
   }
   assert(offset < N_words, "offset too large");
   index--;
   q -= offset;
+  assert(q >= _sp->bottom(),
+         err_msg("q = " PTR_FORMAT " crossed below bottom = " PTR_FORMAT,
+                 q, _sp->bottom()));
+  assert(q < _sp->end(),
+         err_msg("q = " PTR_FORMAT " crossed above end = " PTR_FORMAT,
+                 q, _sp->end()));
   HeapWord* n = q;
 
   while (n <= addr) {
     debug_only(HeapWord* last = q);   // for debugging
     q = n;
     n += _sp->block_size(n);
-    assert(n > q, err_msg("Looping at: " INTPTR_FORMAT, n));
+    assert(n > q,
+           err_msg("Looping at n = " PTR_FORMAT " with last = " PTR_FORMAT " _sp = [" PTR_FORMAT "," PTR_FORMAT ")",
+                   n, last, _sp->bottom(), _sp->end()));
   }
   assert(q <= addr, err_msg("wrong order for current (" INTPTR_FORMAT ") <= arg (" INTPTR_FORMAT ")", q, addr));
   assert(addr <= n, err_msg("wrong order for arg (" INTPTR_FORMAT ") <= next (" INTPTR_FORMAT ")", addr, n));
--- a/src/share/vm/memory/cardTableModRefBS.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/memory/cardTableModRefBS.cpp	Thu May 12 19:39:58 2011 -0700
@@ -455,25 +455,29 @@
   return true;
 }
 
-
 void CardTableModRefBS::non_clean_card_iterate_possibly_parallel(Space* sp,
                                                                  MemRegion mr,
-                                                                 DirtyCardToOopClosure* dcto_cl,
-                                                                 ClearNoncleanCardWrapper* cl) {
+                                                                 OopsInGenClosure* cl,
+                                                                 CardTableRS* ct) {
   if (!mr.is_empty()) {
     int n_threads = SharedHeap::heap()->n_par_threads();
     if (n_threads > 0) {
 #ifndef SERIALGC
-      non_clean_card_iterate_parallel_work(sp, mr, dcto_cl, cl, n_threads);
+      non_clean_card_iterate_parallel_work(sp, mr, cl, ct, n_threads);
 #else  // SERIALGC
       fatal("Parallel gc not supported here.");
 #endif // SERIALGC
     } else {
       // We do not call the non_clean_card_iterate_serial() version below because
       // we want to clear the cards (which non_clean_card_iterate_serial() does not
-      // do for us), and the ClearNoncleanCardWrapper closure itself does the work
-      // of finding contiguous dirty ranges of cards to process (and clear).
-      cl->do_MemRegion(mr);
+      // do for us): clear_cl here does the work of finding contiguous dirty ranges
+      // of cards to process and clear.
+
+      DirtyCardToOopClosure* dcto_cl = sp->new_dcto_cl(cl, precision(),
+                                                       cl->gen_boundary());
+      ClearNoncleanCardWrapper clear_cl(dcto_cl, ct);
+
+      clear_cl.do_MemRegion(mr);
     }
   }
 }
--- a/src/share/vm/memory/cardTableModRefBS.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/memory/cardTableModRefBS.hpp	Thu May 12 19:39:58 2011 -0700
@@ -150,7 +150,9 @@
   // Mapping from address to card marking array entry
   jbyte* byte_for(const void* p) const {
     assert(_whole_heap.contains(p),
-           "out of bounds access to card marking array");
+           err_msg("Attempt to access p = "PTR_FORMAT" out of bounds of "
+                   " card marking array's _whole_heap = ["PTR_FORMAT","PTR_FORMAT")",
+                   p, _whole_heap.start(), _whole_heap.end()));
     jbyte* result = &byte_map_base[uintptr_t(p) >> card_shift];
     assert(result >= _byte_map && result < _byte_map + _byte_map_size,
            "out of bounds accessor for card marking array");
@@ -173,18 +175,17 @@
   // A variant of the above that will operate in a parallel mode if
   // worker threads are available, and clear the dirty cards as it
   // processes them.
-  // ClearNoncleanCardWrapper cl must wrap the DirtyCardToOopClosure dcto_cl,
-  // which may itself be modified by the method.
+  // XXX ??? MemRegionClosure above vs OopsInGenClosure below XXX
+  // XXX some new_dcto_cl's take OopClosure's, plus as above there are
+  // some MemRegionClosures. Clean this up everywhere. XXX
   void non_clean_card_iterate_possibly_parallel(Space* sp, MemRegion mr,
-                                                DirtyCardToOopClosure* dcto_cl,
-                                                ClearNoncleanCardWrapper* cl);
+                                                OopsInGenClosure* cl, CardTableRS* ct);
 
  private:
   // Work method used to implement non_clean_card_iterate_possibly_parallel()
   // above in the parallel case.
   void non_clean_card_iterate_parallel_work(Space* sp, MemRegion mr,
-                                            DirtyCardToOopClosure* dcto_cl,
-                                            ClearNoncleanCardWrapper* cl,
+                                            OopsInGenClosure* cl, CardTableRS* ct,
                                             int n_threads);
 
  protected:
@@ -198,11 +199,6 @@
 
   // *** Support for parallel card scanning.
 
-  enum SomeConstantsForParallelism {
-    StridesPerThread    = 2,
-    CardsPerStrideChunk = 256
-  };
-
   // This is an array, one element per covered region of the card table.
   // Each entry is itself an array, with one element per chunk in the
   // covered region.  Each entry of these arrays is the lowest non-clean
@@ -235,7 +231,7 @@
   // covers the given address.
   uintptr_t addr_to_chunk_index(const void* addr) {
     uintptr_t card = (uintptr_t) byte_for(addr);
-    return card / CardsPerStrideChunk;
+    return card / ParGCCardsPerStrideChunk;
   }
 
   // Apply cl, which must either itself apply dcto_cl or be dcto_cl,
@@ -243,8 +239,8 @@
   void process_stride(Space* sp,
                       MemRegion used,
                       jint stride, int n_strides,
-                      DirtyCardToOopClosure* dcto_cl,
-                      ClearNoncleanCardWrapper* cl,
+                      OopsInGenClosure* cl,
+                      CardTableRS* ct,
                       jbyte** lowest_non_clean,
                       uintptr_t lowest_non_clean_base_chunk_index,
                       size_t lowest_non_clean_chunk_size);
@@ -457,14 +453,18 @@
     size_t delta = pointer_delta(p, byte_map_base, sizeof(jbyte));
     HeapWord* result = (HeapWord*) (delta << card_shift);
     assert(_whole_heap.contains(result),
-           "out of bounds accessor from card marking array");
+           err_msg("Returning result = "PTR_FORMAT" out of bounds of "
+                   " card marking array's _whole_heap = ["PTR_FORMAT","PTR_FORMAT")",
+                   result, _whole_heap.start(), _whole_heap.end()));
     return result;
   }
 
   // Mapping from address to card marking array index.
   size_t index_for(void* p) {
     assert(_whole_heap.contains(p),
-           "out of bounds access to card marking array");
+           err_msg("Attempt to access p = "PTR_FORMAT" out of bounds of "
+                   " card marking array's _whole_heap = ["PTR_FORMAT","PTR_FORMAT")",
+                   p, _whole_heap.start(), _whole_heap.end()));
     return byte_for(p) - _byte_map;
   }
 
@@ -482,7 +482,7 @@
   void verify_dirty_region(MemRegion mr) PRODUCT_RETURN;
 
   static size_t par_chunk_heapword_alignment() {
-    return CardsPerStrideChunk * card_size_in_words;
+    return ParGCCardsPerStrideChunk * card_size_in_words;
   }
 
 };
--- a/src/share/vm/memory/cardTableRS.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/memory/cardTableRS.cpp	Thu May 12 19:39:58 2011 -0700
@@ -162,7 +162,7 @@
 }
 
 ClearNoncleanCardWrapper::ClearNoncleanCardWrapper(
-  MemRegionClosure* dirty_card_closure, CardTableRS* ct) :
+  DirtyCardToOopClosure* dirty_card_closure, CardTableRS* ct) :
     _dirty_card_closure(dirty_card_closure), _ct(ct) {
     _is_par = (SharedHeap::heap()->n_par_threads() > 0);
 }
@@ -246,10 +246,6 @@
 
 void CardTableRS::younger_refs_in_space_iterate(Space* sp,
                                                 OopsInGenClosure* cl) {
-  DirtyCardToOopClosure* dcto_cl = sp->new_dcto_cl(cl, _ct_bs->precision(),
-                                                   cl->gen_boundary());
-  ClearNoncleanCardWrapper clear_cl(dcto_cl, this);
-
   const MemRegion urasm = sp->used_region_at_save_marks();
 #ifdef ASSERT
   // Convert the assertion check to a warning if we are running
@@ -275,10 +271,10 @@
     if (!urasm.equals(urasm2)) {
       warning("CMS+ParNew: Flickering used_region_at_save_marks()!!");
     }
+    ShouldNotReachHere();
   }
 #endif
-  _ct_bs->non_clean_card_iterate_possibly_parallel(sp, urasm,
-                                                   dcto_cl, &clear_cl);
+  _ct_bs->non_clean_card_iterate_possibly_parallel(sp, urasm, cl, this);
 }
 
 void CardTableRS::clear_into_younger(Generation* gen, bool clear_perm) {
--- a/src/share/vm/memory/cardTableRS.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/memory/cardTableRS.hpp	Thu May 12 19:39:58 2011 -0700
@@ -31,7 +31,6 @@
 
 class Space;
 class OopsInGenClosure;
-class DirtyCardToOopClosure;
 
 // This kind of "GenRemSet" uses a card table both as shared data structure
 // for a mod ref barrier set and for the rem set information.
@@ -167,7 +166,7 @@
 };
 
 class ClearNoncleanCardWrapper: public MemRegionClosure {
-  MemRegionClosure* _dirty_card_closure;
+  DirtyCardToOopClosure* _dirty_card_closure;
   CardTableRS* _ct;
   bool _is_par;
 private:
@@ -179,7 +178,7 @@
   inline bool clear_card_parallel(jbyte* entry);
 
 public:
-  ClearNoncleanCardWrapper(MemRegionClosure* dirty_card_closure, CardTableRS* ct);
+  ClearNoncleanCardWrapper(DirtyCardToOopClosure* dirty_card_closure, CardTableRS* ct);
   void do_MemRegion(MemRegion mr);
 };
 
--- a/src/share/vm/memory/genCollectedHeap.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/memory/genCollectedHeap.cpp	Thu May 12 19:39:58 2011 -0700
@@ -537,7 +537,7 @@
         // Timer for individual generations. Last argument is false: no CR
         TraceTime t1(_gens[i]->short_name(), PrintGCDetails, false, gclog_or_tty);
         TraceCollectorStats tcs(_gens[i]->counters());
-        TraceMemoryManagerStats tmms(_gens[i]->kind());
+        TraceMemoryManagerStats tmms(_gens[i]->kind(),gc_cause());
 
         size_t prev_used = _gens[i]->used();
         _gens[i]->stat_record()->invocations++;
--- a/src/share/vm/memory/space.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/memory/space.cpp	Thu May 12 19:39:58 2011 -0700
@@ -97,6 +97,14 @@
   }
 }
 
+// We get called with "mr" representing the dirty region
+// that we want to process. Because of imprecise marking,
+// we may need to extend the incoming "mr" to the right,
+// and scan more. However, because we may already have
+// scanned some of that extended region, we may need to
+// trim its right-end back some so we do not scan what
+// we (or another worker thread) may already have scanned
+// or planning to scan.
 void DirtyCardToOopClosure::do_MemRegion(MemRegion mr) {
 
   // Some collectors need to do special things whenever their dirty
@@ -148,7 +156,7 @@
   // e.g. the dirty card region is entirely in a now free object
   // -- something that could happen with a concurrent sweeper.
   bottom = MIN2(bottom, top);
-  mr     = MemRegion(bottom, top);
+  MemRegion extended_mr = MemRegion(bottom, top);
   assert(bottom <= top &&
          (_precision != CardTableModRefBS::ObjHeadPreciseArray ||
           _min_done == NULL ||
@@ -156,8 +164,8 @@
          "overlap!");
 
   // Walk the region if it is not empty; otherwise there is nothing to do.
-  if (!mr.is_empty()) {
-    walk_mem_region(mr, bottom_obj, top);
+  if (!extended_mr.is_empty()) {
+    walk_mem_region(extended_mr, bottom_obj, top);
   }
 
   // An idempotent closure might be applied in any order, so we don't
--- a/src/share/vm/runtime/advancedThresholdPolicy.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/runtime/advancedThresholdPolicy.cpp	Thu May 12 19:39:58 2011 -0700
@@ -1,7 +1,26 @@
 /*
-* Copyright (c) 2010, 2011 Oracle and/or its affiliates. All rights reserved.
-* ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
-*/
+ * Copyright (c) 2010, 2011, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
 
 #include "precompiled.hpp"
 #include "runtime/advancedThresholdPolicy.hpp"
--- a/src/share/vm/runtime/advancedThresholdPolicy.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/runtime/advancedThresholdPolicy.hpp	Thu May 12 19:39:58 2011 -0700
@@ -1,7 +1,26 @@
 /*
-* Copyright (c) 2010, 2011 Oracle and/or its affiliates. All rights reserved.
-* ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
-*/
+ * Copyright (c) 2010, 2011, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
 
 #ifndef SHARE_VM_RUNTIME_ADVANCEDTHRESHOLDPOLICY_HPP
 #define SHARE_VM_RUNTIME_ADVANCEDTHRESHOLDPOLICY_HPP
--- a/src/share/vm/runtime/globals.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/runtime/globals.hpp	Thu May 12 19:39:58 2011 -0700
@@ -1460,8 +1460,10 @@
   product(intx, ParallelGCBufferWastePct, 10,                               \
           "wasted fraction of parallel allocation buffer.")                 \
                                                                             \
-  product(bool, ParallelGCRetainPLAB, true,                                 \
-          "Retain parallel allocation buffers across scavenges.")           \
+  diagnostic(bool, ParallelGCRetainPLAB, false,                             \
+             "Retain parallel allocation buffers across scavenges; "        \
+             " -- disabled because this currently conflicts with "          \
+             " parallel card scanning under certain conditions ")           \
                                                                             \
   product(intx, TargetPLABWastePct, 10,                                     \
           "target wasted space in last buffer as pct of overall allocation")\
@@ -1495,7 +1497,15 @@
   product(uintx, ParGCDesiredObjsFromOverflowList, 20,                      \
           "The desired number of objects to claim from the overflow list")  \
                                                                             \
-  product(uintx, CMSParPromoteBlocksToClaim, 16,                             \
+  diagnostic(intx, ParGCStridesPerThread, 2,                                \
+          "The number of strides per worker thread that we divide up the "  \
+          "card table scanning work into")                                  \
+                                                                            \
+  diagnostic(intx, ParGCCardsPerStrideChunk, 256,                           \
+          "The number of cards in each chunk of the parallel chunks used "  \
+          "during card table scanning")                                     \
+                                                                            \
+  product(uintx, CMSParPromoteBlocksToClaim, 16,                            \
           "Number of blocks to attempt to claim when refilling CMS LAB for "\
           "parallel GC.")                                                   \
                                                                             \
--- a/src/share/vm/runtime/serviceThread.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/runtime/serviceThread.cpp	Thu May 12 19:39:58 2011 -0700
@@ -28,6 +28,7 @@
 #include "runtime/serviceThread.hpp"
 #include "runtime/mutexLocker.hpp"
 #include "prims/jvmtiImpl.hpp"
+#include "services/gcNotifier.hpp"
 
 ServiceThread* ServiceThread::_instance = NULL;
 
@@ -81,6 +82,7 @@
   while (true) {
     bool sensors_changed = false;
     bool has_jvmti_events = false;
+    bool has_gc_notification_event = false;
     JvmtiDeferredEvent jvmti_event;
     {
       // Need state transition ThreadBlockInVM so that this thread
@@ -95,9 +97,10 @@
 
       MutexLockerEx ml(Service_lock, Mutex::_no_safepoint_check_flag);
       while (!(sensors_changed = LowMemoryDetector::has_pending_requests()) &&
-             !(has_jvmti_events = JvmtiDeferredEventQueue::has_events())) {
+             !(has_jvmti_events = JvmtiDeferredEventQueue::has_events()) &&
+              !(has_gc_notification_event = GCNotifier::has_event())) {
         // wait until one of the sensors has pending requests, or there is a
-        // pending JVMTI event to post
+        // pending JVMTI event or JMX GC notification to post
         Service_lock->wait(Mutex::_no_safepoint_check_flag);
       }
 
@@ -113,6 +116,10 @@
     if (sensors_changed) {
       LowMemoryDetector::process_sensor_changes(jt);
     }
+
+    if(has_gc_notification_event) {
+        GCNotifier::sendNotification(CHECK);
+    }
   }
 }
 
--- a/src/share/vm/runtime/simpleThresholdPolicy.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/runtime/simpleThresholdPolicy.cpp	Thu May 12 19:39:58 2011 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2011, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2010, 2011, Oracle and/or its affiliates. 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
--- a/src/share/vm/runtime/simpleThresholdPolicy.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/runtime/simpleThresholdPolicy.hpp	Thu May 12 19:39:58 2011 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2011, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2010, 2011, Oracle and/or its affiliates. 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
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/share/vm/services/gcNotifier.cpp	Thu May 12 19:39:58 2011 -0700
@@ -0,0 +1,216 @@
+/*
+ * Copyright (c) 2011, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "classfile/systemDictionary.hpp"
+#include "classfile/vmSymbols.hpp"
+#include "oops/oop.inline.hpp"
+#include "runtime/interfaceSupport.hpp"
+#include "runtime/java.hpp"
+#include "runtime/javaCalls.hpp"
+#include "runtime/mutex.hpp"
+#include "runtime/mutexLocker.hpp"
+#include "services/gcNotifier.hpp"
+#include "services/management.hpp"
+#include "services/memoryService.hpp"
+#include "memoryManager.hpp"
+#include "memory/oopFactory.hpp"
+
+GCNotificationRequest *GCNotifier::first_request = NULL;
+GCNotificationRequest *GCNotifier::last_request = NULL;
+
+void GCNotifier::pushNotification(GCMemoryManager *mgr, const char *action, const char *cause) {
+  // Make a copy of the last GC statistics
+  // GC may occur between now and the creation of the notification
+  int num_pools = MemoryService::num_memory_pools();
+  GCStatInfo* stat = new GCStatInfo(num_pools);
+  mgr->get_last_gc_stat(stat);
+  GCNotificationRequest *request = new GCNotificationRequest(os::javaTimeMillis(),mgr,action,cause,stat);
+  addRequest(request);
+ }
+
+void GCNotifier::addRequest(GCNotificationRequest *request) {
+  MutexLockerEx ml(Service_lock, Mutex::_no_safepoint_check_flag);
+  if(first_request == NULL) {
+    first_request = request;
+  } else {
+    last_request->next = request;
+  }
+  last_request = request;
+  Service_lock->notify_all();
+}
+
+GCNotificationRequest *GCNotifier::getRequest() {
+  MutexLockerEx ml(Service_lock, Mutex::_no_safepoint_check_flag);
+  GCNotificationRequest *request = first_request;
+  if(first_request != NULL) {
+    first_request = first_request->next;
+  }
+  return request;
+}
+
+bool GCNotifier::has_event() {
+  return first_request != NULL;
+}
+
+static Handle getGcInfoBuilder(GCMemoryManager *gcManager,TRAPS) {
+
+  klassOop k = Management::sun_management_GarbageCollectorImpl_klass(CHECK_NH);
+  instanceKlassHandle gcMBeanKlass (THREAD, k);
+
+  instanceOop i = gcManager->get_memory_manager_instance(THREAD);
+  instanceHandle ih(THREAD, i);
+
+  JavaValue result(T_OBJECT);
+  JavaCallArguments args(ih);
+
+  JavaCalls::call_virtual(&result,
+                          gcMBeanKlass,
+                          vmSymbols::getGcInfoBuilder_name(),
+                          vmSymbols::getGcInfoBuilder_signature(),
+                          &args,
+                          CHECK_NH);
+  return Handle(THREAD,(oop)result.get_jobject());
+
+}
+
+static Handle createGcInfo(GCMemoryManager *gcManager, GCStatInfo *gcStatInfo,TRAPS) {
+
+  // Fill the arrays of MemoryUsage objects with before and after GC
+  // per pool memory usage
+
+  klassOop muKlass = Management::java_lang_management_MemoryUsage_klass(CHECK_NH);   objArrayOop bu = oopFactory::new_objArray( muKlass,MemoryService::num_memory_pools(), CHECK_NH);
+  objArrayHandle usage_before_gc_ah(THREAD, bu);
+  objArrayOop au = oopFactory::new_objArray(muKlass,MemoryService::num_memory_pools(), CHECK_NH);
+  objArrayHandle usage_after_gc_ah(THREAD, au);
+
+  for (int i = 0; i < MemoryService::num_memory_pools(); i++) {
+    Handle before_usage = MemoryService::create_MemoryUsage_obj(gcStatInfo->before_gc_usage_for_pool(i), CHECK_NH);
+    Handle after_usage;
+
+    MemoryUsage u = gcStatInfo->after_gc_usage_for_pool(i);
+    if (u.max_size() == 0 && u.used() > 0) {
+      // If max size == 0, this pool is a survivor space.
+      // Set max size = -1 since the pools will be swapped after GC.
+      MemoryUsage usage(u.init_size(), u.used(), u.committed(), (size_t)-1);
+      after_usage = MemoryService::create_MemoryUsage_obj(usage, CHECK_NH);
+    } else {
+        after_usage = MemoryService::create_MemoryUsage_obj(u, CHECK_NH);
+    }
+    usage_before_gc_ah->obj_at_put(i, before_usage());
+    usage_after_gc_ah->obj_at_put(i, after_usage());
+  }
+
+  // Current implementation only has 1 attribute (number of GC threads)
+  // The type is 'I'
+  objArrayOop extra_args_array = oopFactory::new_objArray(SystemDictionary::Integer_klass(), 1, CHECK_NH);
+  objArrayHandle extra_array (THREAD, extra_args_array);
+  klassOop itKlass= SystemDictionary::Integer_klass();
+  instanceKlassHandle intK(THREAD, itKlass);
+
+  instanceHandle extra_arg_val = intK->allocate_instance_handle(CHECK_NH);
+
+  {
+    JavaValue res(T_VOID);
+    JavaCallArguments argsInt;
+    argsInt.push_oop(extra_arg_val);
+    argsInt.push_int(gcManager->num_gc_threads());
+
+    JavaCalls::call_special(&res,
+                            intK,
+                            vmSymbols::object_initializer_name(),
+                            vmSymbols::int_void_signature(),
+                            &argsInt,
+                            CHECK_NH);
+  }
+  extra_array->obj_at_put(0,extra_arg_val());
+
+  klassOop gcInfoklass = Management::com_sun_management_GcInfo_klass(CHECK_NH);
+  instanceKlassHandle ik (THREAD,gcInfoklass);
+
+  Handle gcInfo_instance = ik->allocate_instance_handle(CHECK_NH);
+
+  JavaValue constructor_result(T_VOID);
+  JavaCallArguments constructor_args(16);
+  constructor_args.push_oop(gcInfo_instance);
+  constructor_args.push_oop(getGcInfoBuilder(gcManager,THREAD));
+  constructor_args.push_long(gcStatInfo->gc_index());
+  constructor_args.push_long(gcStatInfo->start_time());
+  constructor_args.push_long(gcStatInfo->end_time());
+  constructor_args.push_oop(usage_before_gc_ah);
+  constructor_args.push_oop(usage_after_gc_ah);
+  constructor_args.push_oop(extra_array);
+
+  JavaCalls::call_special(&constructor_result,
+                          ik,
+                          vmSymbols::object_initializer_name(),
+                          vmSymbols::com_sun_management_GcInfo_constructor_signature(),
+                          &constructor_args,
+                          CHECK_NH);
+
+  return Handle(gcInfo_instance());
+}
+
+void GCNotifier::sendNotification(TRAPS) {
+  ResourceMark rm(THREAD);
+  GCNotificationRequest *request = getRequest();
+  if(request != NULL) {
+    Handle objGcInfo = createGcInfo(request->gcManager,request->gcStatInfo,THREAD);
+
+    Handle objName = java_lang_String::create_from_platform_dependent_str(request->gcManager->name(), CHECK);
+    Handle objAction = java_lang_String::create_from_platform_dependent_str(request->gcAction, CHECK);
+    Handle objCause = java_lang_String::create_from_platform_dependent_str(request->gcCause, CHECK);
+
+    klassOop k = Management::sun_management_GarbageCollectorImpl_klass(CHECK);
+    instanceKlassHandle gc_mbean_klass (THREAD, k);
+
+    instanceOop gc_mbean = request->gcManager->get_memory_manager_instance(THREAD);
+    instanceHandle gc_mbean_h(THREAD, gc_mbean);
+    if (!gc_mbean_h->is_a(k)) {
+      THROW_MSG(vmSymbols::java_lang_IllegalArgumentException(),
+                "This GCMemoryManager doesn't have a GarbageCollectorMXBean");
+    }
+
+    JavaValue result(T_VOID);
+    JavaCallArguments args(gc_mbean_h);
+    args.push_long(request->timestamp);
+    args.push_oop(objName);
+    args.push_oop(objAction);
+    args.push_oop(objCause);
+    args.push_oop(objGcInfo);
+
+    JavaCalls::call_virtual(&result,
+                            gc_mbean_klass,
+                            vmSymbols::createGCNotification_name(),
+                            vmSymbols::createGCNotification_signature(),
+                            &args,
+                            CHECK);
+    if (HAS_PENDING_EXCEPTION) {
+      CLEAR_PENDING_EXCEPTION;
+    }
+
+    delete request;
+  }
+}
+
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/share/vm/services/gcNotifier.hpp	Thu May 12 19:39:58 2011 -0700
@@ -0,0 +1,69 @@
+/*
+ * Copyright (c) 2011, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_SERVICES_GCNOTIFIER_HPP
+#define SHARE_VM_SERVICES_GCNOTIFIER_HPP
+
+#include "memory/allocation.hpp"
+#include "services/memoryPool.hpp"
+#include "services/memoryService.hpp"
+#include "services/memoryManager.hpp"
+
+class GCNotificationRequest : public CHeapObj {
+  friend class GCNotifier;
+  GCNotificationRequest *next;
+  jlong timestamp;
+  GCMemoryManager *gcManager;
+  const char *gcAction;
+  const char *gcCause;
+  GCStatInfo *gcStatInfo;
+public:
+  GCNotificationRequest(jlong ts, GCMemoryManager *manager, const char*action, const char *cause,GCStatInfo *info) {
+    next = NULL;
+    timestamp = ts;
+    gcManager = manager;
+    gcAction = action;
+    gcCause = cause;
+    gcStatInfo = info;
+  }
+
+  ~GCNotificationRequest() {
+    delete gcStatInfo;
+  }
+};
+
+class GCNotifier : public AllStatic {
+  friend class ServiceThread;
+private:
+  static GCNotificationRequest *first_request;
+  static GCNotificationRequest *last_request;
+  static void addRequest(GCNotificationRequest *request);
+  static GCNotificationRequest *getRequest();
+public:
+  static void pushNotification(GCMemoryManager *manager, const char *action, const char *cause);
+  static bool has_event();
+  static void sendNotification(TRAPS);
+};
+
+#endif // SHARE_VM_SERVICES_GCNOTIFIER_HPP
--- a/src/share/vm/services/jmm.h	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/services/jmm.h	Thu May 12 19:39:58 2011 -0700
@@ -48,7 +48,7 @@
   JMM_VERSION_1_0 = 0x20010000,
   JMM_VERSION_1_1 = 0x20010100, // JDK 6
   JMM_VERSION_1_2 = 0x20010200, // JDK 7
-  JMM_VERSION     = 0x20010200
+  JMM_VERSION     = 0x20010201
 };
 
 typedef struct {
@@ -293,6 +293,9 @@
                                                   jlongArray ids,
                                                   jboolean lockedMonitors,
                                                   jboolean lockedSynchronizers);
+  void         (JNICALL *SetGCNotificationEnabled) (JNIEnv *env,
+                                                    jobject mgr,
+                                                    jboolean enabled);
 } JmmInterface;
 
 #ifdef __cplusplus
--- a/src/share/vm/services/management.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/services/management.cpp	Thu May 12 19:39:58 2011 -0700
@@ -42,6 +42,7 @@
 #include "services/classLoadingService.hpp"
 #include "services/heapDumper.hpp"
 #include "services/lowMemoryDetector.hpp"
+#include "services/gcNotifier.hpp"
 #include "services/management.hpp"
 #include "services/memoryManager.hpp"
 #include "services/memoryPool.hpp"
@@ -60,6 +61,8 @@
 klassOop Management::_memoryManagerMXBean_klass = NULL;
 klassOop Management::_garbageCollectorMXBean_klass = NULL;
 klassOop Management::_managementFactory_klass = NULL;
+klassOop Management::_garbageCollectorImpl_klass = NULL;
+klassOop Management::_gcInfo_klass = NULL;
 
 jmmOptionalSupport Management::_optional_support = {0};
 TimeStamp Management::_stamp;
@@ -179,6 +182,8 @@
   f->do_oop((oop*) &_memoryManagerMXBean_klass);
   f->do_oop((oop*) &_garbageCollectorMXBean_klass);
   f->do_oop((oop*) &_managementFactory_klass);
+  f->do_oop((oop*) &_garbageCollectorImpl_klass);
+  f->do_oop((oop*) &_gcInfo_klass);
 }
 
 klassOop Management::java_lang_management_ThreadInfo_klass(TRAPS) {
@@ -230,6 +235,20 @@
   return _managementFactory_klass;
 }
 
+klassOop Management::sun_management_GarbageCollectorImpl_klass(TRAPS) {
+  if (_garbageCollectorImpl_klass == NULL) {
+    _garbageCollectorImpl_klass = load_and_initialize_klass(vmSymbols::sun_management_GarbageCollectorImpl(), CHECK_NULL);
+  }
+  return _garbageCollectorImpl_klass;
+}
+
+klassOop Management::com_sun_management_GcInfo_klass(TRAPS) {
+  if (_gcInfo_klass == NULL) {
+    _gcInfo_klass = load_and_initialize_klass(vmSymbols::com_sun_management_GcInfo(), CHECK_NULL);
+  }
+  return _gcInfo_klass;
+}
+
 static void initialize_ThreadInfo_constructor_arguments(JavaCallArguments* args, ThreadSnapshot* snapshot, TRAPS) {
   Handle snapshot_thread(THREAD, snapshot->threadObj());
 
@@ -2056,6 +2075,13 @@
   }
 JVM_END
 
+JVM_ENTRY(void, jmm_SetGCNotificationEnabled(JNIEnv *env, jobject obj, jboolean enabled))
+  ResourceMark rm(THREAD);
+  // Get the GCMemoryManager
+  GCMemoryManager* mgr = get_gc_memory_manager_from_jobject(obj, CHECK);
+  mgr->set_notification_enabled(enabled?true:false);
+JVM_END
+
 // Dump heap - Returns 0 if succeeds.
 JVM_ENTRY(jint, jmm_DumpHeap0(JNIEnv *env, jstring outputfile, jboolean live))
 #ifndef SERVICES_KERNEL
@@ -2122,7 +2148,8 @@
   jmm_FindDeadlockedThreads,
   jmm_SetVMGlobal,
   NULL,
-  jmm_DumpThreads
+  jmm_DumpThreads,
+  jmm_SetGCNotificationEnabled
 };
 
 void* Management::get_jmm_interface(int version) {
--- a/src/share/vm/services/management.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/services/management.hpp	Thu May 12 19:39:58 2011 -0700
@@ -49,6 +49,8 @@
   static klassOop           _memoryManagerMXBean_klass;
   static klassOop           _garbageCollectorMXBean_klass;
   static klassOop           _managementFactory_klass;
+  static klassOop           _garbageCollectorImpl_klass;
+  static klassOop           _gcInfo_klass;
 
   static klassOop load_and_initialize_klass(Symbol* sh, TRAPS);
 
@@ -86,6 +88,8 @@
   static klassOop java_lang_management_GarbageCollectorMXBean_klass(TRAPS);
   static klassOop sun_management_Sensor_klass(TRAPS);
   static klassOop sun_management_ManagementFactory_klass(TRAPS);
+  static klassOop sun_management_GarbageCollectorImpl_klass(TRAPS);
+  static klassOop com_sun_management_GcInfo_klass(TRAPS);
 
   static instanceOop create_thread_info_instance(ThreadSnapshot* snapshot, TRAPS);
   static instanceOop create_thread_info_instance(ThreadSnapshot* snapshot, objArrayHandle monitors_array, typeArrayHandle depths_array, objArrayHandle synchronizers_array, TRAPS);
--- a/src/share/vm/services/memoryManager.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/services/memoryManager.cpp	Thu May 12 19:39:58 2011 -0700
@@ -33,6 +33,7 @@
 #include "services/memoryManager.hpp"
 #include "services/memoryPool.hpp"
 #include "services/memoryService.hpp"
+#include "services/gcNotifier.hpp"
 #include "utilities/dtrace.hpp"
 
 HS_DTRACE_PROBE_DECL8(hotspot, mem__pool__gc__begin, char*, int, char*, int,
@@ -202,6 +203,7 @@
   _last_gc_lock = new Mutex(Mutex::leaf, "_last_gc_lock", true);
   _current_gc_stat = NULL;
   _num_gc_threads = 1;
+  _notification_enabled = false;
 }
 
 GCMemoryManager::~GCMemoryManager() {
@@ -250,7 +252,8 @@
 // to ensure the current gc stat is placed in _last_gc_stat.
 void GCMemoryManager::gc_end(bool recordPostGCUsage,
                              bool recordAccumulatedGCTime,
-                             bool recordGCEndTime, bool countCollection) {
+                             bool recordGCEndTime, bool countCollection,
+                             GCCause::Cause cause) {
   if (recordAccumulatedGCTime) {
     _accumulated_timer.stop();
   }
@@ -283,6 +286,11 @@
       pool->set_last_collection_usage(usage);
       LowMemoryDetector::detect_after_gc_memory(pool);
     }
+    if(is_notification_enabled()) {
+      bool isMajorGC = this == MemoryService::get_major_gc_manager();
+      GCNotifier::pushNotification(this, isMajorGC ? "end of major GC" : "end of minor GC",
+                                   GCCause::to_string(cause));
+    }
   }
   if (countCollection) {
     _num_collections++;
--- a/src/share/vm/services/memoryManager.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/services/memoryManager.hpp	Thu May 12 19:39:58 2011 -0700
@@ -166,6 +166,7 @@
   Mutex*       _last_gc_lock;
   GCStatInfo*  _current_gc_stat;
   int          _num_gc_threads;
+  volatile bool _notification_enabled;
 public:
   GCMemoryManager();
   ~GCMemoryManager();
@@ -181,7 +182,7 @@
   void   gc_begin(bool recordGCBeginTime, bool recordPreGCUsage,
                   bool recordAccumulatedGCTime);
   void   gc_end(bool recordPostGCUsage, bool recordAccumulatedGCTime,
-                bool recordGCEndTime, bool countCollection);
+                bool recordGCEndTime, bool countCollection, GCCause::Cause cause);
 
   void        reset_gc_stat()   { _num_collections = 0; _accumulated_timer.reset(); }
 
@@ -189,6 +190,8 @@
   // the collection count. Zero signifies no gc has taken place.
   size_t get_last_gc_stat(GCStatInfo* dest);
 
+  void set_notification_enabled(bool enabled) { _notification_enabled = enabled; }
+  bool is_notification_enabled() { return _notification_enabled; }
   virtual MemoryManager::Name kind() = 0;
 };
 
--- a/src/share/vm/services/memoryService.cpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/services/memoryService.cpp	Thu May 12 19:39:58 2011 -0700
@@ -565,7 +565,8 @@
 
 void MemoryService::gc_end(bool fullGC, bool recordPostGCUsage,
                            bool recordAccumulatedGCTime,
-                           bool recordGCEndTime, bool countCollection) {
+                           bool recordGCEndTime, bool countCollection,
+                           GCCause::Cause cause) {
 
   GCMemoryManager* mgr;
   if (fullGC) {
@@ -577,7 +578,7 @@
 
   // register the GC end statistics and memory usage
   mgr->gc_end(recordPostGCUsage, recordAccumulatedGCTime, recordGCEndTime,
-              countCollection);
+              countCollection, cause);
 }
 
 void MemoryService::oops_do(OopClosure* f) {
@@ -633,7 +634,7 @@
 // gc manager (so _fullGC is set to false ) and for other generation kinds
 // doing mark-sweep-compact uses major gc manager (so _fullGC is set
 // to true).
-TraceMemoryManagerStats::TraceMemoryManagerStats(Generation::Name kind) {
+TraceMemoryManagerStats::TraceMemoryManagerStats(Generation::Name kind, GCCause::Cause cause) {
   switch (kind) {
     case Generation::DefNew:
 #ifndef SERIALGC
@@ -654,9 +655,10 @@
   }
   // this has to be called in a stop the world pause and represent
   // an entire gc pause, start to finish:
-  initialize(_fullGC, true, true, true, true, true, true, true);
+  initialize(_fullGC, cause,true, true, true, true, true, true, true);
 }
 TraceMemoryManagerStats::TraceMemoryManagerStats(bool fullGC,
+                                                 GCCause::Cause cause,
                                                  bool recordGCBeginTime,
                                                  bool recordPreGCUsage,
                                                  bool recordPeakUsage,
@@ -664,7 +666,7 @@
                                                  bool recordAccumulatedGCTime,
                                                  bool recordGCEndTime,
                                                  bool countCollection) {
-  initialize(fullGC, recordGCBeginTime, recordPreGCUsage, recordPeakUsage,
+    initialize(fullGC, cause, recordGCBeginTime, recordPreGCUsage, recordPeakUsage,
              recordPostGCUsage, recordAccumulatedGCTime, recordGCEndTime,
              countCollection);
 }
@@ -672,6 +674,7 @@
 // for a subclass to create then initialize an instance before invoking
 // the MemoryService
 void TraceMemoryManagerStats::initialize(bool fullGC,
+                                         GCCause::Cause cause,
                                          bool recordGCBeginTime,
                                          bool recordPreGCUsage,
                                          bool recordPeakUsage,
@@ -687,6 +690,7 @@
   _recordAccumulatedGCTime = recordAccumulatedGCTime;
   _recordGCEndTime = recordGCEndTime;
   _countCollection = countCollection;
+  _cause = cause;
 
   MemoryService::gc_begin(_fullGC, _recordGCBeginTime, _recordAccumulatedGCTime,
                           _recordPreGCUsage, _recordPeakUsage);
@@ -694,6 +698,6 @@
 
 TraceMemoryManagerStats::~TraceMemoryManagerStats() {
   MemoryService::gc_end(_fullGC, _recordPostGCUsage, _recordAccumulatedGCTime,
-                        _recordGCEndTime, _countCollection);
+                        _recordGCEndTime, _countCollection, _cause);
 }
 
--- a/src/share/vm/services/memoryService.hpp	Thu May 12 16:24:44 2011 -0700
+++ b/src/share/vm/services/memoryService.hpp	Thu May 12 19:39:58 2011 -0700
@@ -29,6 +29,7 @@
 #include "memory/generation.hpp"
 #include "runtime/handles.hpp"
 #include "services/memoryUsage.hpp"
+#include "gc_interface/gcCause.hpp"
 
 // Forward declaration
 class MemoryPool;
@@ -162,7 +163,8 @@
                        bool recordPreGCUsage, bool recordPeakUsage);
   static void gc_end(bool fullGC, bool recordPostGCUsage,
                      bool recordAccumulatedGCTime,
-                     bool recordGCEndTime, bool countCollection);
+                     bool recordGCEndTime, bool countCollection,
+                     GCCause::Cause cause);
 
 
   static void oops_do(OopClosure* f);
@@ -172,6 +174,14 @@
 
   // Create an instance of java/lang/management/MemoryUsage
   static Handle create_MemoryUsage_obj(MemoryUsage usage, TRAPS);
+
+  static const GCMemoryManager* get_minor_gc_manager() {
+      return _minor_gc_manager;
+  }
+
+  static const GCMemoryManager* get_major_gc_manager() {
+      return _major_gc_manager;
+  }
 };
 
 class TraceMemoryManagerStats : public StackObj {
@@ -184,10 +194,11 @@
   bool         _recordAccumulatedGCTime;
   bool         _recordGCEndTime;
   bool         _countCollection;
-
+  GCCause::Cause _cause;
 public:
   TraceMemoryManagerStats() {}
   TraceMemoryManagerStats(bool fullGC,
+                          GCCause::Cause cause,
                           bool recordGCBeginTime = true,
                           bool recordPreGCUsage = true,
                           bool recordPeakUsage = true,
@@ -197,6 +208,7 @@
                           bool countCollection = true);
 
   void initialize(bool fullGC,
+                  GCCause::Cause cause,
                   bool recordGCBeginTime,
                   bool recordPreGCUsage,
                   bool recordPeakUsage,
@@ -205,7 +217,7 @@
                   bool recordGCEndTime,
                   bool countCollection);
 
-  TraceMemoryManagerStats(Generation::Name kind);
+  TraceMemoryManagerStats(Generation::Name kind, GCCause::Cause cause);
   ~TraceMemoryManagerStats();
 };