--- a/src/share/vm/classfile/stringTable.cpp	Thu May 11 14:13:59 2017 -0700
+++ b/src/share/vm/classfile/stringTable.cpp	Mon May 15 12:20:15 2017 +0200
@@ -314,7 +314,11 @@
 }
 
 void StringTable::unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int* processed, int* removed) {
-  buckets_unlink_or_oops_do(is_alive, f, 0, the_table()->table_size(), processed, removed);
+  BucketUnlinkContext context;
+  buckets_unlink_or_oops_do(is_alive, f, 0, the_table()->table_size(), &context);
+  _the_table->bulk_free_entries(&context);
+  *processed = context._num_processed;
+  *removed = context._num_removed;
 }
 
 void StringTable::possibly_parallel_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int* processed, int* removed) {
@@ -323,6 +327,7 @@
   assert(SafepointSynchronize::is_at_safepoint(), "must be at safepoint");
   const int limit = the_table()->table_size();
 
+  BucketUnlinkContext context;
   for (;;) {
     // Grab next set of buckets to scan
     int start_idx = Atomic::add(ClaimChunkSize, &_parallel_claimed_idx) - ClaimChunkSize;
@@ -332,8 +337,11 @@
     }
 
     int end_idx = MIN2(limit, start_idx + ClaimChunkSize);
-    buckets_unlink_or_oops_do(is_alive, f, start_idx, end_idx, processed, removed);
+    buckets_unlink_or_oops_do(is_alive, f, start_idx, end_idx, &context);
   }
+  _the_table->bulk_free_entries(&context);
+  *processed = context._num_processed;
+  *removed = context._num_removed;
 }
 
 void StringTable::buckets_oops_do(OopClosure* f, int start_idx, int end_idx) {
@@ -359,7 +367,7 @@
   }
 }
 
-void StringTable::buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, int* processed, int* removed) {
+void StringTable::buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, BucketUnlinkContext* context) {
   const int limit = the_table()->table_size();
 
   assert(0 <= start_idx && start_idx <= limit,
@@ -383,10 +391,9 @@
         p = entry->next_addr();
       } else {
         *p = entry->next();
-        the_table()->free_entry(entry);
-        (*removed)++;
+        context->free_entry(entry);
       }
-      (*processed)++;
+      context->_num_processed++;
       entry = *p;
     }
   }

--- a/src/share/vm/classfile/stringTable.hpp	Thu May 11 14:13:59 2017 -0700
+++ b/src/share/vm/classfile/stringTable.hpp	Mon May 15 12:20:15 2017 +0200
@@ -61,9 +61,13 @@
   // Apply the give oop closure to the entries to the buckets
   // in the range [start_idx, end_idx).
   static void buckets_oops_do(OopClosure* f, int start_idx, int end_idx);
+
+  typedef StringTable::BucketUnlinkContext BucketUnlinkContext;
   // Unlink or apply the give oop closure to the entries to the buckets
-  // in the range [start_idx, end_idx).
-  static void buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, int* processed, int* removed);
+  // in the range [start_idx, end_idx). Unlinked bucket entries are collected in the given
+  // context to be freed later.
+  // This allows multiple threads to work on the table at once.
+  static void buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, BucketUnlinkContext* context);
 
   // Hashing algorithm, used as the hash value used by the
   //     StringTable for bucket selection and comparison (stored in the

--- a/src/share/vm/classfile/symbolTable.cpp	Thu May 11 14:13:59 2017 -0700
+++ b/src/share/vm/classfile/symbolTable.cpp	Mon May 15 12:20:15 2017 +0200
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2016, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2017, 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
@@ -98,7 +98,7 @@
 int SymbolTable::_symbols_counted = 0;
 volatile int SymbolTable::_parallel_claimed_idx = 0;
 
-void SymbolTable::buckets_unlink(int start_idx, int end_idx, int* processed, int* removed) {
+void SymbolTable::buckets_unlink(int start_idx, int end_idx, BucketUnlinkContext* context) {
   for (int i = start_idx; i < end_idx; ++i) {
     HashtableEntry<Symbol*, mtSymbol>** p = the_table()->bucket_addr(i);
     HashtableEntry<Symbol*, mtSymbol>* entry = the_table()->bucket(i);
@@ -111,15 +111,14 @@
         break;
       }
       Symbol* s = entry->literal();
-      (*processed)++;
+      context->_num_processed++;
       assert(s != NULL, "just checking");
       // If reference count is zero, remove.
       if (s->refcount() == 0) {
         assert(!entry->is_shared(), "shared entries should be kept live");
         delete s;
-        (*removed)++;
         *p = entry->next();
-        the_table()->free_entry(entry);
+        context->free_entry(entry);
       } else {
         p = entry->next_addr();
       }
@@ -132,17 +131,20 @@
 // Remove unreferenced symbols from the symbol table
 // This is done late during GC.
 void SymbolTable::unlink(int* processed, int* removed) {
-  size_t memory_total = 0;
-  buckets_unlink(0, the_table()->table_size(), processed, removed);
-  _symbols_removed += *removed;
-  _symbols_counted += *processed;
+  BucketUnlinkContext context;
+  buckets_unlink(0, the_table()->table_size(), &context);
+  _the_table->bulk_free_entries(&context);
+  *processed = context._num_processed;
+  *removed = context._num_removed;
+
+  _symbols_removed = context._num_removed;
+  _symbols_counted = context._num_processed;
 }
 
 void SymbolTable::possibly_parallel_unlink(int* processed, int* removed) {
   const int limit = the_table()->table_size();
 
-  size_t memory_total = 0;
-
+  BucketUnlinkContext context;
   for (;;) {
     // Grab next set of buckets to scan
     int start_idx = Atomic::add(ClaimChunkSize, &_parallel_claimed_idx) - ClaimChunkSize;
@@ -152,10 +154,15 @@
     }
 
     int end_idx = MIN2(limit, start_idx + ClaimChunkSize);
-    buckets_unlink(start_idx, end_idx, processed, removed);
+    buckets_unlink(start_idx, end_idx, &context);
   }
-  Atomic::add(*processed, &_symbols_counted);
-  Atomic::add(*removed, &_symbols_removed);
+
+  _the_table->bulk_free_entries(&context);
+  *processed = context._num_processed;
+  *removed = context._num_removed;
+
+  Atomic::add(context._num_processed, &_symbols_counted);
+  Atomic::add(context._num_removed, &_symbols_removed);
 }
 
 // Create a new table and using alternate hash code, populate the new table

--- a/src/share/vm/classfile/symbolTable.hpp	Thu May 11 14:13:59 2017 -0700
+++ b/src/share/vm/classfile/symbolTable.hpp	Mon May 15 12:20:15 2017 +0200
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2016, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2017, 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
@@ -154,8 +154,11 @@
 
   static volatile int _parallel_claimed_idx;
 
-  // Release any dead symbols
-  static void buckets_unlink(int start_idx, int end_idx, int* processed, int* removed);
+  typedef SymbolTable::BucketUnlinkContext BucketUnlinkContext;
+  // Release any dead symbols. Unlinked bucket entries are collected in the given
+  // context to be freed later.
+  // This allows multiple threads to work on the table at once.
+  static void buckets_unlink(int start_idx, int end_idx, BucketUnlinkContext* context);
 public:
   enum {
     symbol_alloc_batch_size = 8,

--- a/src/share/vm/runtime/vmStructs.cpp	Thu May 11 14:13:59 2017 -0700
+++ b/src/share/vm/runtime/vmStructs.cpp	Mon May 15 12:20:15 2017 +0200
@@ -673,7 +673,7 @@
                                                                                                                                      \
   nonstatic_field(BasicHashtable<mtInternal>,  _table_size,                                   int)                                   \
   nonstatic_field(BasicHashtable<mtInternal>,  _buckets,                                      HashtableBucket<mtInternal>*)          \
-  nonstatic_field(BasicHashtable<mtInternal>,  _free_list,                                    BasicHashtableEntry<mtInternal>*)      \
+  volatile_nonstatic_field(BasicHashtable<mtInternal>,  _free_list,                           BasicHashtableEntry<mtInternal>*)      \
   nonstatic_field(BasicHashtable<mtInternal>,  _first_free_entry,                             char*)                                 \
   nonstatic_field(BasicHashtable<mtInternal>,  _end_block,                                    char*)                                 \
   nonstatic_field(BasicHashtable<mtInternal>,  _entry_size,                                   int)                                   \

--- a/src/share/vm/utilities/hashtable.cpp	Thu May 11 14:13:59 2017 -0700
+++ b/src/share/vm/utilities/hashtable.cpp	Mon May 15 12:20:15 2017 +0200
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2016, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2017, 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
@@ -173,6 +173,35 @@
   }
 }
 
+template <MEMFLAGS F> void BasicHashtable<F>::BucketUnlinkContext::free_entry(BasicHashtableEntry<F>* entry) {
+  entry->set_next(_removed_head);
+  _removed_head = entry;
+  if (_removed_tail == NULL) {
+    _removed_tail = entry;
+  }
+  _num_removed++;
+}
+
+template <MEMFLAGS F> void BasicHashtable<F>::bulk_free_entries(BucketUnlinkContext* context) {
+  if (context->_num_removed == 0) {
+    assert(context->_removed_head == NULL && context->_removed_tail == NULL,
+           "Zero entries in the unlink context, but elements linked from " PTR_FORMAT " to " PTR_FORMAT,
+           p2i(context->_removed_head), p2i(context->_removed_tail));
+    return;
+  }
+
+  // MT-safe add of the list of BasicHashTableEntrys from the context to the free list.
+  BasicHashtableEntry<F>* current = _free_list;
+  while (true) {
+    context->_removed_tail->set_next(current);
+    BasicHashtableEntry<F>* old = (BasicHashtableEntry<F>*)Atomic::cmpxchg_ptr(context->_removed_head, &_free_list, current);
+    if (old == current) {
+      break;
+    }
+    current = old;
+  }
+  Atomic::add(-context->_num_removed, &_number_of_entries);
+}
 
 // Copy the table to the shared space.
 

--- a/src/share/vm/utilities/hashtable.hpp	Thu May 11 14:13:59 2017 -0700
+++ b/src/share/vm/utilities/hashtable.hpp	Mon May 15 12:20:15 2017 +0200
@@ -173,11 +173,11 @@
   // Instance variables
   int               _table_size;
   HashtableBucket<F>*     _buckets;
-  BasicHashtableEntry<F>* _free_list;
+  BasicHashtableEntry<F>* volatile _free_list;
   char*             _first_free_entry;
   char*             _end_block;
   int               _entry_size;
-  int               _number_of_entries;
+  volatile int      _number_of_entries;
 
 protected:
 
@@ -225,6 +225,24 @@
   // Free the buckets in this hashtable
   void free_buckets();
 
+  // Helper data structure containing context for the bucket entry unlink process,
+  // storing the unlinked buckets in a linked list.
+  // Also avoids the need to pass around these four members as parameters everywhere.
+  struct BucketUnlinkContext {
+    int _num_processed;
+    int _num_removed;
+    // Head and tail pointers for the linked list of removed entries.
+    BasicHashtableEntry<F>* _removed_head;
+    BasicHashtableEntry<F>* _removed_tail;
+
+    BucketUnlinkContext() : _num_processed(0), _num_removed(0), _removed_head(NULL), _removed_tail(NULL) {
+    }
+
+    void free_entry(BasicHashtableEntry<F>* entry);
+  };
+  // Add of bucket entries linked together in the given context to the global free list. This method
+  // is mt-safe wrt. to other calls of this method.
+  void bulk_free_entries(BucketUnlinkContext* context);
 public:
   int table_size() { return _table_size; }
   void set_entry(int index, BasicHashtableEntry<F>* entry);