1 /*
   2  * Copyright (c) 1997, 2017, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #include "precompiled.hpp"
  26 #include "classfile/altHashing.hpp"
  27 #include "classfile/compactHashtable.inline.hpp"
  28 #include "classfile/javaClasses.hpp"
  29 #include "classfile/symbolTable.hpp"
  30 #include "classfile/systemDictionary.hpp"
  31 #include "gc/shared/collectedHeap.inline.hpp"
  32 #include "memory/allocation.inline.hpp"
  33 #include "memory/filemap.hpp"
  34 #include "memory/metaspaceClosure.hpp"
  35 #include "memory/resourceArea.hpp"
  36 #include "oops/oop.inline.hpp"
  37 #include "runtime/atomic.hpp"
  38 #include "runtime/mutexLocker.hpp"
  39 #include "runtime/safepointVerifiers.hpp"
  40 #include "services/diagnosticCommand.hpp"
  41 #include "utilities/hashtable.inline.hpp"
  42 
  43 // --------------------------------------------------------------------------
  44 // the number of buckets a thread claims
  45 const int ClaimChunkSize = 32;
  46 
  47 SymbolTable* SymbolTable::_the_table = NULL;
  48 // Static arena for symbols that are not deallocated
  49 Arena* SymbolTable::_arena = NULL;
  50 bool SymbolTable::_needs_rehashing = false;
  51 bool SymbolTable::_lookup_shared_first = false;
  52 
  53 CompactHashtable<Symbol*, char> SymbolTable::_shared_table;
  54 
  55 Symbol* SymbolTable::allocate_symbol(const u1* name, int len, bool c_heap, TRAPS) {
  56   assert (len <= Symbol::max_length(), "should be checked by caller");
  57 
  58   Symbol* sym;
  59 
  60   if (DumpSharedSpaces) {
  61     c_heap = false;
  62   }
  63   if (c_heap) {
  64     // refcount starts as 1
  65     sym = new (len, THREAD) Symbol(name, len, 1);
  66     assert(sym != NULL, "new should call vm_exit_out_of_memory if C_HEAP is exhausted");
  67   } else {
  68     // Allocate to global arena
  69     sym = new (len, arena(), THREAD) Symbol(name, len, PERM_REFCOUNT);
  70   }
  71   return sym;
  72 }
  73 
  74 void SymbolTable::initialize_symbols(int arena_alloc_size) {
  75   // Initialize the arena for global symbols, size passed in depends on CDS.
  76   if (arena_alloc_size == 0) {
  77     _arena = new (mtSymbol) Arena(mtSymbol);
  78   } else {
  79     _arena = new (mtSymbol) Arena(mtSymbol, arena_alloc_size);
  80   }
  81 }
  82 
  83 // Call function for all symbols in the symbol table.
  84 void SymbolTable::symbols_do(SymbolClosure *cl) {
  85   // all symbols from shared table
  86   _shared_table.symbols_do(cl);
  87 
  88   // all symbols from the dynamic table
  89   const int n = the_table()->table_size();
  90   for (int i = 0; i < n; i++) {
  91     for (HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
  92          p != NULL;
  93          p = p->next()) {
  94       cl->do_symbol(p->literal_addr());
  95     }
  96   }
  97 }
  98 
  99 void SymbolTable::metaspace_pointers_do(MetaspaceClosure* it) {
 100   assert(DumpSharedSpaces, "called only during dump time");
 101   const int n = the_table()->table_size();
 102   for (int i = 0; i < n; i++) {
 103     for (HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 104          p != NULL;
 105          p = p->next()) {
 106       it->push(p->literal_addr());
 107     }
 108   }
 109 }
 110 
 111 int SymbolTable::_symbols_removed = 0;
 112 int SymbolTable::_symbols_counted = 0;
 113 volatile int SymbolTable::_parallel_claimed_idx = 0;
 114 
 115 void SymbolTable::buckets_unlink(int start_idx, int end_idx, BucketUnlinkContext* context) {
 116   for (int i = start_idx; i < end_idx; ++i) {
 117     HashtableEntry<Symbol*, mtSymbol>** p = the_table()->bucket_addr(i);
 118     HashtableEntry<Symbol*, mtSymbol>* entry = the_table()->bucket(i);
 119     while (entry != NULL) {
 120       // Shared entries are normally at the end of the bucket and if we run into
 121       // a shared entry, then there is nothing more to remove. However, if we
 122       // have rehashed the table, then the shared entries are no longer at the
 123       // end of the bucket.
 124       if (entry->is_shared() && !use_alternate_hashcode()) {
 125         break;
 126       }
 127       Symbol* s = entry->literal();
 128       context->_num_processed++;
 129       assert(s != NULL, "just checking");
 130       // If reference count is zero, remove.
 131       if (s->refcount() == 0) {
 132         assert(!entry->is_shared(), "shared entries should be kept live");
 133         delete s;
 134         *p = entry->next();
 135         context->free_entry(entry);
 136       } else {
 137         p = entry->next_addr();
 138       }
 139       // get next entry
 140       entry = (HashtableEntry<Symbol*, mtSymbol>*)HashtableEntry<Symbol*, mtSymbol>::make_ptr(*p);
 141     }
 142   }
 143 }
 144 
 145 // Remove unreferenced symbols from the symbol table
 146 // This is done late during GC.
 147 void SymbolTable::unlink(int* processed, int* removed) {
 148   BucketUnlinkContext context;
 149   buckets_unlink(0, the_table()->table_size(), &context);
 150   _the_table->bulk_free_entries(&context);
 151   *processed = context._num_processed;
 152   *removed = context._num_removed;
 153 
 154   _symbols_removed = context._num_removed;
 155   _symbols_counted = context._num_processed;
 156 }
 157 
 158 void SymbolTable::possibly_parallel_unlink(int* processed, int* removed) {
 159   const int limit = the_table()->table_size();
 160 
 161   BucketUnlinkContext context;
 162   for (;;) {
 163     // Grab next set of buckets to scan
 164     int start_idx = Atomic::add(ClaimChunkSize, &_parallel_claimed_idx) - ClaimChunkSize;
 165     if (start_idx >= limit) {
 166       // End of table
 167       break;
 168     }
 169 
 170     int end_idx = MIN2(limit, start_idx + ClaimChunkSize);
 171     buckets_unlink(start_idx, end_idx, &context);
 172   }
 173 
 174   _the_table->bulk_free_entries(&context);
 175   *processed = context._num_processed;
 176   *removed = context._num_removed;
 177 
 178   Atomic::add(context._num_processed, &_symbols_counted);
 179   Atomic::add(context._num_removed, &_symbols_removed);
 180 }
 181 
 182 // Create a new table and using alternate hash code, populate the new table
 183 // with the existing strings.   Set flag to use the alternate hash code afterwards.
 184 void SymbolTable::rehash_table() {
 185   if (DumpSharedSpaces) {
 186     tty->print_cr("Warning: rehash_table should not be called while dumping archive");
 187     return;
 188   }
 189 
 190   assert(SafepointSynchronize::is_at_safepoint(), "must be at safepoint");
 191   // This should never happen with -Xshare:dump but it might in testing mode.
 192   if (DumpSharedSpaces) return;
 193   // Create a new symbol table
 194   SymbolTable* new_table = new SymbolTable();
 195 
 196   the_table()->move_to(new_table);
 197 
 198   // Delete the table and buckets (entries are reused in new table).
 199   delete _the_table;
 200   // Don't check if we need rehashing until the table gets unbalanced again.
 201   // Then rehash with a new global seed.
 202   _needs_rehashing = false;
 203   _the_table = new_table;
 204 }
 205 
 206 // Lookup a symbol in a bucket.
 207 
 208 Symbol* SymbolTable::lookup_dynamic(int index, const char* name,
 209                                     int len, unsigned int hash) {
 210   int count = 0;
 211   for (HashtableEntry<Symbol*, mtSymbol>* e = bucket(index); e != NULL; e = e->next()) {
 212     count++;  // count all entries in this bucket, not just ones with same hash
 213     if (e->hash() == hash) {
 214       Symbol* sym = e->literal();
 215       if (sym->equals(name, len)) {
 216         // something is referencing this symbol now.
 217         sym->increment_refcount();
 218         return sym;
 219       }
 220     }
 221   }
 222   // If the bucket size is too deep check if this hash code is insufficient.
 223   if (count >= rehash_count && !needs_rehashing()) {
 224     _needs_rehashing = check_rehash_table(count);
 225   }
 226   return NULL;
 227 }
 228 
 229 Symbol* SymbolTable::lookup_shared(const char* name,
 230                                    int len, unsigned int hash) {
 231   if (use_alternate_hashcode()) {
 232     // hash_code parameter may use alternate hashing algorithm but the shared table
 233     // always uses the same original hash code.
 234     hash = hash_shared_symbol(name, len);
 235   }
 236   return _shared_table.lookup(name, hash, len);
 237 }
 238 
 239 Symbol* SymbolTable::lookup(int index, const char* name,
 240                             int len, unsigned int hash) {
 241   Symbol* sym;
 242   if (_lookup_shared_first) {
 243     sym = lookup_shared(name, len, hash);
 244     if (sym != NULL) {
 245       return sym;
 246     }
 247     _lookup_shared_first = false;
 248     return lookup_dynamic(index, name, len, hash);
 249   } else {
 250     sym = lookup_dynamic(index, name, len, hash);
 251     if (sym != NULL) {
 252       return sym;
 253     }
 254     sym = lookup_shared(name, len, hash);
 255     if (sym != NULL) {
 256       _lookup_shared_first = true;
 257     }
 258     return sym;
 259   }
 260 }
 261 
 262 u4 SymbolTable::encode_shared(Symbol* sym) {
 263   assert(DumpSharedSpaces, "called only during dump time");
 264   uintx base_address = uintx(MetaspaceShared::shared_rs()->base());
 265   uintx offset = uintx(sym) - base_address;
 266   assert(offset < 0x7fffffff, "sanity");
 267   return u4(offset);
 268 }
 269 
 270 Symbol* SymbolTable::decode_shared(u4 offset) {
 271   assert(!DumpSharedSpaces, "called only during runtime");
 272   uintx base_address = _shared_table.base_address();
 273   Symbol* sym = (Symbol*)(base_address + offset);
 274 
 275 #ifndef PRODUCT
 276   const char* s = (const char*)sym->bytes();
 277   int len = sym->utf8_length();
 278   unsigned int hash = hash_symbol(s, len);
 279   assert(sym == lookup_shared(s, len, hash), "must be shared symbol");
 280 #endif
 281 
 282   return sym;
 283 }
 284 
 285 // Pick hashing algorithm.
 286 unsigned int SymbolTable::hash_symbol(const char* s, int len) {
 287   return use_alternate_hashcode() ?
 288            AltHashing::murmur3_32(seed(), (const jbyte*)s, len) :
 289            java_lang_String::hash_code((const jbyte*)s, len);
 290 }
 291 
 292 unsigned int SymbolTable::hash_shared_symbol(const char* s, int len) {
 293   return java_lang_String::hash_code((const jbyte*)s, len);
 294 }
 295 
 296 
 297 // We take care not to be blocking while holding the
 298 // SymbolTable_lock. Otherwise, the system might deadlock, since the
 299 // symboltable is used during compilation (VM_thread) The lock free
 300 // synchronization is simplified by the fact that we do not delete
 301 // entries in the symbol table during normal execution (only during
 302 // safepoints).
 303 
 304 Symbol* SymbolTable::lookup(const char* name, int len, TRAPS) {
 305   unsigned int hashValue = hash_symbol(name, len);
 306   int index = the_table()->hash_to_index(hashValue);
 307 
 308   Symbol* s = the_table()->lookup(index, name, len, hashValue);
 309 
 310   // Found
 311   if (s != NULL) return s;
 312 
 313   // Grab SymbolTable_lock first.
 314   MutexLocker ml(SymbolTable_lock, THREAD);
 315 
 316   // Otherwise, add to symbol to table
 317   return the_table()->basic_add(index, (u1*)name, len, hashValue, true, THREAD);
 318 }
 319 
 320 Symbol* SymbolTable::lookup(const Symbol* sym, int begin, int end, TRAPS) {
 321   char* buffer;
 322   int index, len;
 323   unsigned int hashValue;
 324   char* name;
 325   {
 326     debug_only(NoSafepointVerifier nsv;)
 327 
 328     name = (char*)sym->base() + begin;
 329     len = end - begin;
 330     hashValue = hash_symbol(name, len);
 331     index = the_table()->hash_to_index(hashValue);
 332     Symbol* s = the_table()->lookup(index, name, len, hashValue);
 333 
 334     // Found
 335     if (s != NULL) return s;
 336   }
 337 
 338   // Otherwise, add to symbol to table. Copy to a C string first.
 339   char stack_buf[128];
 340   ResourceMark rm(THREAD);
 341   if (len <= 128) {
 342     buffer = stack_buf;
 343   } else {
 344     buffer = NEW_RESOURCE_ARRAY_IN_THREAD(THREAD, char, len);
 345   }
 346   for (int i=0; i<len; i++) {
 347     buffer[i] = name[i];
 348   }
 349   // Make sure there is no safepoint in the code above since name can't move.
 350   // We can't include the code in NoSafepointVerifier because of the
 351   // ResourceMark.
 352 
 353   // Grab SymbolTable_lock first.
 354   MutexLocker ml(SymbolTable_lock, THREAD);
 355 
 356   return the_table()->basic_add(index, (u1*)buffer, len, hashValue, true, THREAD);
 357 }
 358 
 359 Symbol* SymbolTable::lookup_only(const char* name, int len,
 360                                    unsigned int& hash) {
 361   hash = hash_symbol(name, len);
 362   int index = the_table()->hash_to_index(hash);
 363 
 364   Symbol* s = the_table()->lookup(index, name, len, hash);
 365   return s;
 366 }
 367 
 368 // Look up the address of the literal in the SymbolTable for this Symbol*
 369 // Do not create any new symbols
 370 // Do not increment the reference count to keep this alive
 371 Symbol** SymbolTable::lookup_symbol_addr(Symbol* sym){
 372   unsigned int hash = hash_symbol((char*)sym->bytes(), sym->utf8_length());
 373   int index = the_table()->hash_to_index(hash);
 374 
 375   for (HashtableEntry<Symbol*, mtSymbol>* e = the_table()->bucket(index); e != NULL; e = e->next()) {
 376     if (e->hash() == hash) {
 377       Symbol* literal_sym = e->literal();
 378       if (sym == literal_sym) {
 379         return e->literal_addr();
 380       }
 381     }
 382   }
 383   return NULL;
 384 }
 385 
 386 // Suggestion: Push unicode-based lookup all the way into the hashing
 387 // and probing logic, so there is no need for convert_to_utf8 until
 388 // an actual new Symbol* is created.
 389 Symbol* SymbolTable::lookup_unicode(const jchar* name, int utf16_length, TRAPS) {
 390   int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
 391   char stack_buf[128];
 392   if (utf8_length < (int) sizeof(stack_buf)) {
 393     char* chars = stack_buf;
 394     UNICODE::convert_to_utf8(name, utf16_length, chars);
 395     return lookup(chars, utf8_length, THREAD);
 396   } else {
 397     ResourceMark rm(THREAD);
 398     char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
 399     UNICODE::convert_to_utf8(name, utf16_length, chars);
 400     return lookup(chars, utf8_length, THREAD);
 401   }
 402 }
 403 
 404 Symbol* SymbolTable::lookup_only_unicode(const jchar* name, int utf16_length,
 405                                            unsigned int& hash) {
 406   int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
 407   char stack_buf[128];
 408   if (utf8_length < (int) sizeof(stack_buf)) {
 409     char* chars = stack_buf;
 410     UNICODE::convert_to_utf8(name, utf16_length, chars);
 411     return lookup_only(chars, utf8_length, hash);
 412   } else {
 413     ResourceMark rm;
 414     char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
 415     UNICODE::convert_to_utf8(name, utf16_length, chars);
 416     return lookup_only(chars, utf8_length, hash);
 417   }
 418 }
 419 
 420 void SymbolTable::add(ClassLoaderData* loader_data, const constantPoolHandle& cp,
 421                       int names_count,
 422                       const char** names, int* lengths, int* cp_indices,
 423                       unsigned int* hashValues, TRAPS) {
 424   // Grab SymbolTable_lock first.
 425   MutexLocker ml(SymbolTable_lock, THREAD);
 426 
 427   SymbolTable* table = the_table();
 428   bool added = table->basic_add(loader_data, cp, names_count, names, lengths,
 429                                 cp_indices, hashValues, CHECK);
 430   if (!added) {
 431     // do it the hard way
 432     for (int i=0; i<names_count; i++) {
 433       int index = table->hash_to_index(hashValues[i]);
 434       bool c_heap = !loader_data->is_the_null_class_loader_data();
 435       Symbol* sym = table->basic_add(index, (u1*)names[i], lengths[i], hashValues[i], c_heap, CHECK);
 436       cp->symbol_at_put(cp_indices[i], sym);
 437     }
 438   }
 439 }
 440 
 441 Symbol* SymbolTable::new_permanent_symbol(const char* name, TRAPS) {
 442   unsigned int hash;
 443   Symbol* result = SymbolTable::lookup_only((char*)name, (int)strlen(name), hash);
 444   if (result != NULL) {
 445     return result;
 446   }
 447   // Grab SymbolTable_lock first.
 448   MutexLocker ml(SymbolTable_lock, THREAD);
 449 
 450   SymbolTable* table = the_table();
 451   int index = table->hash_to_index(hash);
 452   return table->basic_add(index, (u1*)name, (int)strlen(name), hash, false, THREAD);
 453 }
 454 
 455 Symbol* SymbolTable::basic_add(int index_arg, u1 *name, int len,
 456                                unsigned int hashValue_arg, bool c_heap, TRAPS) {
 457   assert(!Universe::heap()->is_in_reserved(name),
 458          "proposed name of symbol must be stable");
 459 
 460   // Don't allow symbols to be created which cannot fit in a Symbol*.
 461   if (len > Symbol::max_length()) {
 462     THROW_MSG_0(vmSymbols::java_lang_InternalError(),
 463                 "name is too long to represent");
 464   }
 465 
 466   // Cannot hit a safepoint in this function because the "this" pointer can move.
 467   NoSafepointVerifier nsv;
 468 
 469   // Check if the symbol table has been rehashed, if so, need to recalculate
 470   // the hash value and index.
 471   unsigned int hashValue;
 472   int index;
 473   if (use_alternate_hashcode()) {
 474     hashValue = hash_symbol((const char*)name, len);
 475     index = hash_to_index(hashValue);
 476   } else {
 477     hashValue = hashValue_arg;
 478     index = index_arg;
 479   }
 480 
 481   // Since look-up was done lock-free, we need to check if another
 482   // thread beat us in the race to insert the symbol.
 483   Symbol* test = lookup(index, (char*)name, len, hashValue);
 484   if (test != NULL) {
 485     // A race occurred and another thread introduced the symbol.
 486     assert(test->refcount() != 0, "lookup should have incremented the count");
 487     return test;
 488   }
 489 
 490   // Create a new symbol.
 491   Symbol* sym = allocate_symbol(name, len, c_heap, CHECK_NULL);
 492   assert(sym->equals((char*)name, len), "symbol must be properly initialized");
 493 
 494   HashtableEntry<Symbol*, mtSymbol>* entry = new_entry(hashValue, sym);
 495   add_entry(index, entry);
 496   return sym;
 497 }
 498 
 499 // This version of basic_add adds symbols in batch from the constant pool
 500 // parsing.
 501 bool SymbolTable::basic_add(ClassLoaderData* loader_data, const constantPoolHandle& cp,
 502                             int names_count,
 503                             const char** names, int* lengths,
 504                             int* cp_indices, unsigned int* hashValues,
 505                             TRAPS) {
 506 
 507   // Check symbol names are not too long.  If any are too long, don't add any.
 508   for (int i = 0; i< names_count; i++) {
 509     if (lengths[i] > Symbol::max_length()) {
 510       THROW_MSG_0(vmSymbols::java_lang_InternalError(),
 511                   "name is too long to represent");
 512     }
 513   }
 514 
 515   // Cannot hit a safepoint in this function because the "this" pointer can move.
 516   NoSafepointVerifier nsv;
 517 
 518   for (int i=0; i<names_count; i++) {
 519     // Check if the symbol table has been rehashed, if so, need to recalculate
 520     // the hash value.
 521     unsigned int hashValue;
 522     if (use_alternate_hashcode()) {
 523       hashValue = hash_symbol(names[i], lengths[i]);
 524     } else {
 525       hashValue = hashValues[i];
 526     }
 527     // Since look-up was done lock-free, we need to check if another
 528     // thread beat us in the race to insert the symbol.
 529     int index = hash_to_index(hashValue);
 530     Symbol* test = lookup(index, names[i], lengths[i], hashValue);
 531     if (test != NULL) {
 532       // A race occurred and another thread introduced the symbol, this one
 533       // will be dropped and collected. Use test instead.
 534       cp->symbol_at_put(cp_indices[i], test);
 535       assert(test->refcount() != 0, "lookup should have incremented the count");
 536     } else {
 537       // Create a new symbol.  The null class loader is never unloaded so these
 538       // are allocated specially in a permanent arena.
 539       bool c_heap = !loader_data->is_the_null_class_loader_data();
 540       Symbol* sym = allocate_symbol((const u1*)names[i], lengths[i], c_heap, CHECK_(false));
 541       assert(sym->equals(names[i], lengths[i]), "symbol must be properly initialized");  // why wouldn't it be???
 542       HashtableEntry<Symbol*, mtSymbol>* entry = new_entry(hashValue, sym);
 543       add_entry(index, entry);
 544       cp->symbol_at_put(cp_indices[i], sym);
 545     }
 546   }
 547   return true;
 548 }
 549 
 550 
 551 void SymbolTable::verify() {
 552   for (int i = 0; i < the_table()->table_size(); ++i) {
 553     HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 554     for ( ; p != NULL; p = p->next()) {
 555       Symbol* s = (Symbol*)(p->literal());
 556       guarantee(s != NULL, "symbol is NULL");
 557       unsigned int h = hash_symbol((char*)s->bytes(), s->utf8_length());
 558       guarantee(p->hash() == h, "broken hash in symbol table entry");
 559       guarantee(the_table()->hash_to_index(h) == i,
 560                 "wrong index in symbol table");
 561     }
 562   }
 563 }
 564 
 565 void SymbolTable::dump(outputStream* st, bool verbose) {
 566   if (!verbose) {
 567     the_table()->print_table_statistics(st, "SymbolTable");
 568   } else {
 569     st->print_cr("VERSION: 1.0");
 570     for (int i = 0; i < the_table()->table_size(); ++i) {
 571       HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 572       for ( ; p != NULL; p = p->next()) {
 573         Symbol* s = (Symbol*)(p->literal());
 574         const char* utf8_string = (const char*)s->bytes();
 575         int utf8_length = s->utf8_length();
 576         st->print("%d %d: ", utf8_length, s->refcount());
 577         HashtableTextDump::put_utf8(st, utf8_string, utf8_length);
 578         st->cr();
 579       }
 580     }
 581   }
 582 }
 583 
 584 void SymbolTable::write_to_archive() {
 585 #if INCLUDE_CDS
 586     _shared_table.reset();
 587 
 588     int num_buckets = the_table()->number_of_entries() /
 589                             SharedSymbolTableBucketSize;
 590     CompactSymbolTableWriter writer(num_buckets,
 591                                     &MetaspaceShared::stats()->symbol);
 592     for (int i = 0; i < the_table()->table_size(); ++i) {
 593       HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 594       for ( ; p != NULL; p = p->next()) {
 595         Symbol* s = (Symbol*)(p->literal());
 596       unsigned int fixed_hash =  hash_shared_symbol((char*)s->bytes(), s->utf8_length());
 597         assert(fixed_hash == p->hash(), "must not rehash during dumping");
 598         writer.add(fixed_hash, s);
 599       }
 600     }
 601 
 602     writer.dump(&_shared_table);
 603 
 604     // Verify table is correct
 605     Symbol* sym = vmSymbols::java_lang_Object();
 606     const char* name = (const char*)sym->bytes();
 607     int len = sym->utf8_length();
 608     unsigned int hash = hash_symbol(name, len);
 609     assert(sym == _shared_table.lookup(name, hash, len), "sanity");
 610 #endif
 611 }
 612 
 613 void SymbolTable::serialize(SerializeClosure* soc) {
 614 #if INCLUDE_CDS
 615   _shared_table.set_type(CompactHashtable<Symbol*, char>::_symbol_table);
 616   _shared_table.serialize(soc);
 617 
 618   if (soc->writing()) {
 619     // Sanity. Make sure we don't use the shared table at dump time
 620     _shared_table.reset();
 621   }
 622 #endif
 623 }
 624 
 625 //---------------------------------------------------------------------------
 626 // Non-product code
 627 
 628 #ifndef PRODUCT
 629 
 630 void SymbolTable::print_histogram() {
 631   MutexLocker ml(SymbolTable_lock);
 632   const int results_length = 100;
 633   int counts[results_length];
 634   int sizes[results_length];
 635   int i,j;
 636 
 637   // initialize results to zero
 638   for (j = 0; j < results_length; j++) {
 639     counts[j] = 0;
 640     sizes[j] = 0;
 641   }
 642 
 643   int total_size = 0;
 644   int total_count = 0;
 645   int total_length = 0;
 646   int max_length = 0;
 647   int out_of_range_count = 0;
 648   int out_of_range_size = 0;
 649   for (i = 0; i < the_table()->table_size(); i++) {
 650     HashtableEntry<Symbol*, mtSymbol>* p = the_table()->bucket(i);
 651     for ( ; p != NULL; p = p->next()) {
 652       int size = p->literal()->size();
 653       int len = p->literal()->utf8_length();
 654       if (len < results_length) {
 655         counts[len]++;
 656         sizes[len] += size;
 657       } else {
 658         out_of_range_count++;
 659         out_of_range_size += size;
 660       }
 661       total_count++;
 662       total_size += size;
 663       total_length += len;
 664       max_length = MAX2(max_length, len);
 665     }
 666   }
 667   tty->print_cr("Symbol Table Histogram:");
 668   tty->print_cr("  Total number of symbols  %7d", total_count);
 669   tty->print_cr("  Total size in memory     %7dK",
 670           (total_size*wordSize)/1024);
 671   tty->print_cr("  Total counted            %7d", _symbols_counted);
 672   tty->print_cr("  Total removed            %7d", _symbols_removed);
 673   if (_symbols_counted > 0) {
 674     tty->print_cr("  Percent removed          %3.2f",
 675           ((float)_symbols_removed/(float)_symbols_counted)* 100);
 676   }
 677   tty->print_cr("  Reference counts         %7d", Symbol::_total_count);
 678   tty->print_cr("  Symbol arena used        " SIZE_FORMAT_W(7) "K", arena()->used()/1024);
 679   tty->print_cr("  Symbol arena size        " SIZE_FORMAT_W(7) "K", arena()->size_in_bytes()/1024);
 680   tty->print_cr("  Total symbol length      %7d", total_length);
 681   tty->print_cr("  Maximum symbol length    %7d", max_length);
 682   tty->print_cr("  Average symbol length    %7.2f", ((float) total_length / (float) total_count));
 683   tty->print_cr("  Symbol length histogram:");
 684   tty->print_cr("    %6s %10s %10s", "Length", "#Symbols", "Size");
 685   for (i = 0; i < results_length; i++) {
 686     if (counts[i] > 0) {
 687       tty->print_cr("    %6d %10d %10dK", i, counts[i], (sizes[i]*wordSize)/1024);
 688     }
 689   }
 690   tty->print_cr("  >=%6d %10d %10dK\n", results_length,
 691           out_of_range_count, (out_of_range_size*wordSize)/1024);
 692 }
 693 
 694 void SymbolTable::print() {
 695   for (int i = 0; i < the_table()->table_size(); ++i) {
 696     HashtableEntry<Symbol*, mtSymbol>** p = the_table()->bucket_addr(i);
 697     HashtableEntry<Symbol*, mtSymbol>* entry = the_table()->bucket(i);
 698     if (entry != NULL) {
 699       while (entry != NULL) {
 700         tty->print(PTR_FORMAT " ", p2i(entry->literal()));
 701         entry->literal()->print();
 702         tty->print(" %d", entry->literal()->refcount());
 703         p = entry->next_addr();
 704         entry = (HashtableEntry<Symbol*, mtSymbol>*)HashtableEntry<Symbol*, mtSymbol>::make_ptr(*p);
 705       }
 706       tty->cr();
 707     }
 708   }
 709 }
 710 #endif // PRODUCT
 711 
 712 
 713 // Utility for dumping symbols
 714 SymboltableDCmd::SymboltableDCmd(outputStream* output, bool heap) :
 715                                  DCmdWithParser(output, heap),
 716   _verbose("-verbose", "Dump the content of each symbol in the table",
 717            "BOOLEAN", false, "false") {
 718   _dcmdparser.add_dcmd_option(&_verbose);
 719 }
 720 
 721 void SymboltableDCmd::execute(DCmdSource source, TRAPS) {
 722   VM_DumpHashtable dumper(output(), VM_DumpHashtable::DumpSymbols,
 723                          _verbose.value());
 724   VMThread::execute(&dumper);
 725 }
 726 
 727 int SymboltableDCmd::num_arguments() {
 728   ResourceMark rm;
 729   SymboltableDCmd* dcmd = new SymboltableDCmd(NULL, false);
 730   if (dcmd != NULL) {
 731     DCmdMark mark(dcmd);
 732     return dcmd->_dcmdparser.num_arguments();
 733   } else {
 734     return 0;
 735   }
 736 }