1 /*
   2  * Copyright (c) 2003, 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/dictionary.hpp"
  28 #include "classfile/javaClasses.inline.hpp"
  29 #include "classfile/moduleEntry.hpp"
  30 #include "classfile/packageEntry.hpp"
  31 #include "classfile/placeholders.hpp"
  32 #include "classfile/protectionDomainCache.hpp"
  33 #include "classfile/stringTable.hpp"
  34 #include "memory/allocation.inline.hpp"
  35 #include "memory/filemap.hpp"
  36 #include "memory/resourceArea.hpp"
  37 #include "oops/oop.inline.hpp"
  38 #include "runtime/safepoint.hpp"
  39 #include "utilities/dtrace.hpp"
  40 #include "utilities/hashtable.hpp"
  41 #include "utilities/hashtable.inline.hpp"
  42 #include "utilities/numberSeq.hpp"
  43 
  44 
  45 // This hashtable is implemented as an open hash table with a fixed number of buckets.
  46 
  47 template <MEMFLAGS F> BasicHashtableEntry<F>* BasicHashtable<F>::new_entry_free_list() {
  48   BasicHashtableEntry<F>* entry = NULL;
  49   if (_free_list != NULL) {
  50     entry = _free_list;
  51     _free_list = _free_list->next();
  52   }
  53   return entry;
  54 }
  55 
  56 // HashtableEntrys are allocated in blocks to reduce the space overhead.
  57 template <MEMFLAGS F> BasicHashtableEntry<F>* BasicHashtable<F>::new_entry(unsigned int hashValue) {
  58   BasicHashtableEntry<F>* entry = new_entry_free_list();
  59 
  60   if (entry == NULL) {
  61     if (_first_free_entry + _entry_size >= _end_block) {
  62       int block_size = MIN2(512, MAX2((int)_table_size / 2, (int)_number_of_entries));
  63       int len = _entry_size * block_size;
  64       len = 1 << log2_intptr(len); // round down to power of 2
  65       assert(len >= _entry_size, "");
  66       _first_free_entry = NEW_C_HEAP_ARRAY2(char, len, F, CURRENT_PC);
  67       _end_block = _first_free_entry + len;
  68     }
  69     entry = (BasicHashtableEntry<F>*)_first_free_entry;
  70     _first_free_entry += _entry_size;
  71   }
  72 
  73   assert(_entry_size % HeapWordSize == 0, "");
  74   entry->set_hash(hashValue);
  75   return entry;
  76 }
  77 
  78 
  79 template <class T, MEMFLAGS F> HashtableEntry<T, F>* Hashtable<T, F>::new_entry(unsigned int hashValue, T obj) {
  80   HashtableEntry<T, F>* entry;
  81 
  82   entry = (HashtableEntry<T, F>*)BasicHashtable<F>::new_entry(hashValue);
  83   entry->set_literal(obj);
  84   return entry;
  85 }
  86 
  87 // Version of hashtable entry allocation that allocates in the C heap directly.
  88 // The allocator in blocks is preferable but doesn't have free semantics.
  89 template <class T, MEMFLAGS F> HashtableEntry<T, F>* Hashtable<T, F>::allocate_new_entry(unsigned int hashValue, T obj) {
  90   HashtableEntry<T, F>* entry = (HashtableEntry<T, F>*) NEW_C_HEAP_ARRAY(char, this->entry_size(), F);
  91 
  92   entry->set_hash(hashValue);
  93   entry->set_literal(obj);
  94   entry->set_next(NULL);
  95   return entry;
  96 }
  97 
  98 // Check to see if the hashtable is unbalanced.  The caller set a flag to
  99 // rehash at the next safepoint.  If this bucket is 60 times greater than the
 100 // expected average bucket length, it's an unbalanced hashtable.
 101 // This is somewhat an arbitrary heuristic but if one bucket gets to
 102 // rehash_count which is currently 100, there's probably something wrong.
 103 
 104 template <class T, MEMFLAGS F> bool RehashableHashtable<T, F>::check_rehash_table(int count) {
 105   assert(this->table_size() != 0, "underflow");
 106   if (count > (((double)this->number_of_entries()/(double)this->table_size())*rehash_multiple)) {
 107     // Set a flag for the next safepoint, which should be at some guaranteed
 108     // safepoint interval.
 109     return true;
 110   }
 111   return false;
 112 }
 113 
 114 // Create a new table and using alternate hash code, populate the new table
 115 // with the existing elements.   This can be used to change the hash code
 116 // and could in the future change the size of the table.
 117 
 118 template <class T, MEMFLAGS F> void RehashableHashtable<T, F>::move_to(RehashableHashtable<T, F>* new_table) {
 119 
 120   // Initialize the global seed for hashing.
 121   _seed = AltHashing::compute_seed();
 122   assert(seed() != 0, "shouldn't be zero");
 123 
 124   int saved_entry_count = this->number_of_entries();
 125 
 126   // Iterate through the table and create a new entry for the new table
 127   for (int i = 0; i < new_table->table_size(); ++i) {
 128     for (HashtableEntry<T, F>* p = this->bucket(i); p != NULL; ) {
 129       HashtableEntry<T, F>* next = p->next();
 130       T string = p->literal();
 131       // Use alternate hashing algorithm on the symbol in the first table
 132       unsigned int hashValue = string->new_hash(seed());
 133       // Get a new index relative to the new table (can also change size)
 134       int index = new_table->hash_to_index(hashValue);
 135       p->set_hash(hashValue);
 136       // Keep the shared bit in the Hashtable entry to indicate that this entry
 137       // can't be deleted.   The shared bit is the LSB in the _next field so
 138       // walking the hashtable past these entries requires
 139       // BasicHashtableEntry::make_ptr() call.
 140       bool keep_shared = p->is_shared();
 141       this->unlink_entry(p);
 142       new_table->add_entry(index, p);
 143       if (keep_shared) {
 144         p->set_shared();
 145       }
 146       p = next;
 147     }
 148   }
 149   // give the new table the free list as well
 150   new_table->copy_freelist(this);
 151   assert(new_table->number_of_entries() == saved_entry_count, "lost entry on dictionary copy?");
 152 
 153   // Destroy memory used by the buckets in the hashtable.  The memory
 154   // for the elements has been used in a new table and is not
 155   // destroyed.  The memory reuse will benefit resizing the SystemDictionary
 156   // to avoid a memory allocation spike at safepoint.
 157   BasicHashtable<F>::free_buckets();
 158 }
 159 
 160 template <MEMFLAGS F> void BasicHashtable<F>::free_buckets() {
 161   if (NULL != _buckets) {
 162     // Don't delete the buckets in the shared space.  They aren't
 163     // allocated by os::malloc
 164     if (!UseSharedSpaces ||
 165         !FileMapInfo::current_info()->is_in_shared_space(_buckets)) {
 166        FREE_C_HEAP_ARRAY(HashtableBucket, _buckets);
 167     }
 168     _buckets = NULL;
 169   }
 170 }
 171 
 172 template <MEMFLAGS F> void BasicHashtable<F>::BucketUnlinkContext::free_entry(BasicHashtableEntry<F>* entry) {
 173   entry->set_next(_removed_head);
 174   _removed_head = entry;
 175   if (_removed_tail == NULL) {
 176     _removed_tail = entry;
 177   }
 178   _num_removed++;
 179 }
 180 
 181 template <MEMFLAGS F> void BasicHashtable<F>::bulk_free_entries(BucketUnlinkContext* context) {
 182   if (context->_num_removed == 0) {
 183     assert(context->_removed_head == NULL && context->_removed_tail == NULL,
 184            "Zero entries in the unlink context, but elements linked from " PTR_FORMAT " to " PTR_FORMAT,
 185            p2i(context->_removed_head), p2i(context->_removed_tail));
 186     return;
 187   }
 188 
 189   // MT-safe add of the list of BasicHashTableEntrys from the context to the free list.
 190   BasicHashtableEntry<F>* current = _free_list;
 191   while (true) {
 192     context->_removed_tail->set_next(current);
 193     BasicHashtableEntry<F>* old = Atomic::cmpxchg(context->_removed_head, &_free_list, current);
 194     if (old == current) {
 195       break;
 196     }
 197     current = old;
 198   }
 199   Atomic::add(-context->_num_removed, &_number_of_entries);
 200 }
 201 // Copy the table to the shared space.
 202 template <MEMFLAGS F> size_t BasicHashtable<F>::count_bytes_for_table() {
 203   size_t bytes = 0;
 204   bytes += sizeof(intptr_t); // len
 205 
 206   for (int i = 0; i < _table_size; ++i) {
 207     for (BasicHashtableEntry<F>** p = _buckets[i].entry_addr();
 208          *p != NULL;
 209          p = (*p)->next_addr()) {
 210       bytes += entry_size();
 211     }
 212   }
 213 
 214   return bytes;
 215 }
 216 
 217 // Dump the hash table entries (into CDS archive)
 218 template <MEMFLAGS F> void BasicHashtable<F>::copy_table(char* top, char* end) {
 219   assert(is_aligned(top, sizeof(intptr_t)), "bad alignment");
 220   intptr_t *plen = (intptr_t*)(top);
 221   top += sizeof(*plen);
 222 
 223   int i;
 224   for (i = 0; i < _table_size; ++i) {
 225     for (BasicHashtableEntry<F>** p = _buckets[i].entry_addr();
 226          *p != NULL;
 227          p = (*p)->next_addr()) {
 228       *p = (BasicHashtableEntry<F>*)memcpy(top, (void*)*p, entry_size());
 229       top += entry_size();
 230     }
 231   }
 232   *plen = (char*)(top) - (char*)plen - sizeof(*plen);
 233   assert(top == end, "count_bytes_for_table is wrong");
 234   // Set the shared bit.
 235 
 236   for (i = 0; i < _table_size; ++i) {
 237     for (BasicHashtableEntry<F>* p = bucket(i); p != NULL; p = p->next()) {
 238       p->set_shared();
 239     }
 240   }
 241 }
 242 
 243 // For oops and Strings the size of the literal is interesting. For other types, nobody cares.
 244 static int literal_size(ConstantPool*) { return 0; }
 245 static int literal_size(Klass*)        { return 0; }
 246 #if INCLUDE_ALL_GCS
 247 static int literal_size(nmethod*)      { return 0; }
 248 #endif
 249 
 250 static int literal_size(Symbol *symbol) {
 251   return symbol->size() * HeapWordSize;
 252 }
 253 
 254 static int literal_size(oop obj) {
 255   // NOTE: this would over-count if (pre-JDK8) java_lang_Class::has_offset_field() is true,
 256   // and the String.value array is shared by several Strings. However, starting from JDK8,
 257   // the String.value array is not shared anymore.
 258   if (obj == NULL) {
 259     return 0;
 260   } else if (obj->klass() == SystemDictionary::String_klass()) {
 261     return (obj->size() + java_lang_String::value(obj)->size()) * HeapWordSize;
 262   } else {
 263     return obj->size();
 264   }
 265 }
 266 
 267 template <MEMFLAGS F> bool BasicHashtable<F>::resize(int new_size) {
 268   assert(SafepointSynchronize::is_at_safepoint(), "must be at safepoint");
 269 
 270   // Allocate new buckets
 271   HashtableBucket<F>* buckets_new = NEW_C_HEAP_ARRAY2_RETURN_NULL(HashtableBucket<F>, new_size, F, CURRENT_PC);
 272   if (buckets_new == NULL) {
 273     return false;
 274   }
 275 
 276   // Clear the new buckets
 277   for (int i = 0; i < new_size; i++) {
 278     buckets_new[i].clear();
 279   }
 280 
 281   int table_size_old = _table_size;
 282   // hash_to_index() uses _table_size, so switch the sizes now
 283   _table_size = new_size;
 284 
 285   // Move entries from the old table to a new table
 286   for (int index_old = 0; index_old < table_size_old; index_old++) {
 287     for (BasicHashtableEntry<F>* p = _buckets[index_old].get_entry(); p != NULL; ) {
 288       BasicHashtableEntry<F>* next = p->next();
 289       bool keep_shared = p->is_shared();
 290       int index_new = hash_to_index(p->hash());
 291 
 292       p->set_next(buckets_new[index_new].get_entry());
 293       buckets_new[index_new].set_entry(p);
 294 
 295       if (keep_shared) {
 296         p->set_shared();
 297       }
 298       p = next;
 299     }
 300   }
 301 
 302   // The old backets now can be released
 303   BasicHashtable<F>::free_buckets();
 304 
 305   // Switch to the new storage
 306   _buckets = buckets_new;
 307 
 308   return true;
 309 }
 310 
 311 // Dump footprint and bucket length statistics
 312 //
 313 // Note: if you create a new subclass of Hashtable<MyNewType, F>, you will need to
 314 // add a new function static int literal_size(MyNewType lit)
 315 // because I can't get template <class T> int literal_size(T) to pick the specializations for Symbol and oop.
 316 //
 317 // The StringTable and SymbolTable dumping print how much footprint is used by the String and Symbol
 318 // literals.
 319 
 320 template <class T, MEMFLAGS F> void Hashtable<T, F>::print_table_statistics(outputStream* st,
 321                                                                             const char *table_name) {
 322   NumberSeq summary;
 323   int literal_bytes = 0;
 324   for (int i = 0; i < this->table_size(); ++i) {
 325     int count = 0;
 326     for (HashtableEntry<T, F>* e = this->bucket(i);
 327          e != NULL; e = e->next()) {
 328       count++;
 329       literal_bytes += literal_size(e->literal());
 330     }
 331     summary.add((double)count);
 332   }
 333   double num_buckets = summary.num();
 334   double num_entries = summary.sum();
 335 
 336   int bucket_bytes = (int)num_buckets * sizeof(HashtableBucket<F>);
 337   int entry_bytes  = (int)num_entries * sizeof(HashtableEntry<T, F>);
 338   int total_bytes = literal_bytes +  bucket_bytes + entry_bytes;
 339 
 340   int bucket_size  = (num_buckets <= 0) ? 0 : (bucket_bytes  / num_buckets);
 341   int entry_size   = (num_entries <= 0) ? 0 : (entry_bytes   / num_entries);
 342 
 343   st->print_cr("%s statistics:", table_name);
 344   st->print_cr("Number of buckets       : %9d = %9d bytes, each %d", (int)num_buckets, bucket_bytes,  bucket_size);
 345   st->print_cr("Number of entries       : %9d = %9d bytes, each %d", (int)num_entries, entry_bytes,   entry_size);
 346   if (literal_bytes != 0) {
 347     double literal_avg = (num_entries <= 0) ? 0 : (literal_bytes / num_entries);
 348     st->print_cr("Number of literals      : %9d = %9d bytes, avg %7.3f", (int)num_entries, literal_bytes, literal_avg);
 349   }
 350   st->print_cr("Total footprint         : %9s = %9d bytes", "", total_bytes);
 351   st->print_cr("Average bucket size     : %9.3f", summary.avg());
 352   st->print_cr("Variance of bucket size : %9.3f", summary.variance());
 353   st->print_cr("Std. dev. of bucket size: %9.3f", summary.sd());
 354   st->print_cr("Maximum bucket size     : %9d", (int)summary.maximum());
 355 }
 356 
 357 
 358 // Dump the hash table buckets.
 359 
 360 template <MEMFLAGS F> size_t BasicHashtable<F>::count_bytes_for_buckets() {
 361   size_t bytes = 0;
 362   bytes += sizeof(intptr_t); // len
 363   bytes += sizeof(intptr_t); // _number_of_entries
 364   bytes += _table_size * sizeof(HashtableBucket<F>); // the buckets
 365 
 366   return bytes;
 367 }
 368 
 369 // Dump the buckets (into CDS archive)
 370 template <MEMFLAGS F> void BasicHashtable<F>::copy_buckets(char* top, char* end) {
 371   assert(is_aligned(top, sizeof(intptr_t)), "bad alignment");
 372   intptr_t len = _table_size * sizeof(HashtableBucket<F>);
 373   *(intptr_t*)(top) = len;
 374   top += sizeof(intptr_t);
 375 
 376   *(intptr_t*)(top) = _number_of_entries;
 377   top += sizeof(intptr_t);
 378 
 379   _buckets = (HashtableBucket<F>*)memcpy(top, (void*)_buckets, len);
 380   top += len;
 381 
 382   assert(top == end, "count_bytes_for_buckets is wrong");
 383 }
 384 
 385 #ifndef PRODUCT
 386 
 387 template <class T, MEMFLAGS F> void Hashtable<T, F>::print() {
 388   ResourceMark rm;
 389 
 390   for (int i = 0; i < BasicHashtable<F>::table_size(); i++) {
 391     HashtableEntry<T, F>* entry = bucket(i);
 392     while(entry != NULL) {
 393       tty->print("%d : ", i);
 394       entry->literal()->print();
 395       tty->cr();
 396       entry = entry->next();
 397     }
 398   }
 399 }
 400 
 401 template <MEMFLAGS F>
 402 template <class T> void BasicHashtable<F>::verify_table(const char* table_name) {
 403   int element_count = 0;
 404   int max_bucket_count = 0;
 405   int max_bucket_number = 0;
 406   for (int index = 0; index < table_size(); index++) {
 407     int bucket_count = 0;
 408     for (T* probe = (T*)bucket(index); probe != NULL; probe = probe->next()) {
 409       probe->verify();
 410       bucket_count++;
 411     }
 412     element_count += bucket_count;
 413     if (bucket_count > max_bucket_count) {
 414       max_bucket_count = bucket_count;
 415       max_bucket_number = index;
 416     }
 417   }
 418   guarantee(number_of_entries() == element_count,
 419             "Verify of %s failed", table_name);
 420 
 421   // Log some statistics about the hashtable
 422   log_info(hashtables)("%s max bucket size %d bucket %d element count %d table size %d", table_name,
 423                        max_bucket_count, max_bucket_number, _number_of_entries, _table_size);
 424   if (_number_of_entries > 0 && log_is_enabled(Debug, hashtables)) {
 425     for (int index = 0; index < table_size(); index++) {
 426       int bucket_count = 0;
 427       for (T* probe = (T*)bucket(index); probe != NULL; probe = probe->next()) {
 428         log_debug(hashtables)("bucket %d hash " INTPTR_FORMAT, index, (intptr_t)probe->hash());
 429         bucket_count++;
 430       }
 431       if (bucket_count > 0) {
 432         log_debug(hashtables)("bucket %d count %d", index, bucket_count);
 433       }
 434     }
 435   }
 436 }
 437 #endif // PRODUCT
 438 
 439 // Explicitly instantiate these types
 440 #if INCLUDE_ALL_GCS
 441 template class Hashtable<nmethod*, mtGC>;
 442 template class HashtableEntry<nmethod*, mtGC>;
 443 template class BasicHashtable<mtGC>;
 444 #endif
 445 template class Hashtable<ConstantPool*, mtClass>;
 446 template class RehashableHashtable<Symbol*, mtSymbol>;
 447 template class RehashableHashtable<oopDesc*, mtSymbol>;
 448 template class Hashtable<Symbol*, mtSymbol>;
 449 template class Hashtable<Klass*, mtClass>;
 450 template class Hashtable<InstanceKlass*, mtClass>;
 451 template class Hashtable<oop, mtClass>;
 452 template class Hashtable<Symbol*, mtModule>;
 453 #if defined(SOLARIS) || defined(CHECK_UNHANDLED_OOPS)
 454 template class Hashtable<oop, mtSymbol>;
 455 template class RehashableHashtable<oop, mtSymbol>;
 456 #endif // SOLARIS || CHECK_UNHANDLED_OOPS
 457 template class Hashtable<oopDesc*, mtSymbol>;
 458 template class Hashtable<Symbol*, mtClass>;
 459 template class HashtableEntry<Symbol*, mtSymbol>;
 460 template class HashtableEntry<Symbol*, mtClass>;
 461 template class HashtableEntry<oop, mtSymbol>;
 462 template class HashtableBucket<mtClass>;
 463 template class BasicHashtableEntry<mtSymbol>;
 464 template class BasicHashtableEntry<mtCode>;
 465 template class BasicHashtable<mtClass>;
 466 template class BasicHashtable<mtClassShared>;
 467 template class BasicHashtable<mtSymbol>;
 468 template class BasicHashtable<mtCode>;
 469 template class BasicHashtable<mtInternal>;
 470 template class BasicHashtable<mtModule>;
 471 #if INCLUDE_TRACE
 472 template class Hashtable<Symbol*, mtTracing>;
 473 template class HashtableEntry<Symbol*, mtTracing>;
 474 template class BasicHashtable<mtTracing>;
 475 #endif
 476 template class BasicHashtable<mtCompiler>;
 477 
 478 template void BasicHashtable<mtClass>::verify_table<DictionaryEntry>(char const*);
 479 template void BasicHashtable<mtModule>::verify_table<ModuleEntry>(char const*);
 480 template void BasicHashtable<mtModule>::verify_table<PackageEntry>(char const*);
 481 template void BasicHashtable<mtClass>::verify_table<ProtectionDomainCacheEntry>(char const*);
 482 template void BasicHashtable<mtClass>::verify_table<PlaceholderEntry>(char const*);