1 /*

        2  * Copyright 1997-2009 Sun Microsystems, Inc.  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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,

       20  * CA 95054 USA or visit www.sun.com if you need additional information or

       21  * have any questions.

       22  *

       23  */

       24 

       25 # include "incls/_precompiled.incl"

       26 # include "incls/_symbolTable.cpp.incl"

       27 

       28 // --------------------------------------------------------------------------

       29 

       30 SymbolTable* SymbolTable::_the_table = NULL;

       31 

       32 // Lookup a symbol in a bucket.

       33 

       34 symbolOop SymbolTable::lookup(int index, const char* name,

       35                               int len, unsigned int hash) {

       36   for (HashtableEntry* e = bucket(index); e != NULL; e = e->next()) {

       37     if (e->hash() == hash) {

       38       symbolOop sym = symbolOop(e->literal());

       39       if (sym->equals(name, len)) {

       40         return sym;

       41       }

       42     }

       43   }

       44   return NULL;

       45 }

       46 

       47 

       48 // We take care not to be blocking while holding the

       49 // SymbolTable_lock. Otherwise, the system might deadlock, since the

       50 // symboltable is used during compilation (VM_thread) The lock free

       51 // synchronization is simplified by the fact that we do not delete

       52 // entries in the symbol table during normal execution (only during

       53 // safepoints).

       54 

       55 symbolOop SymbolTable::lookup(const char* name, int len, TRAPS) {

       56   unsigned int hashValue = hash_symbol(name, len);

       57   int index = the_table()->hash_to_index(hashValue);

       58 

       59   symbolOop s = the_table()->lookup(index, name, len, hashValue);

       60 

       61   // Found

       62   if (s != NULL) return s;

       63 

       64   // Otherwise, add to symbol to table

       65   return the_table()->basic_add(index, (u1*)name, len, hashValue, CHECK_NULL);

       66 }

       67 

       68 symbolOop SymbolTable::lookup(symbolHandle sym, int begin, int end, TRAPS) {

       69   char* buffer;

       70   int index, len;

       71   unsigned int hashValue;

       72   char* name;

       73   {

       74     debug_only(No_Safepoint_Verifier nsv;)

       75 

       76     name = (char*)sym->base() + begin;

       77     len = end - begin;

       78     hashValue = hash_symbol(name, len);

       79     index = the_table()->hash_to_index(hashValue);

       80     symbolOop s = the_table()->lookup(index, name, len, hashValue);

       81 

       82     // Found

       83     if (s != NULL) return s;

       84   }

       85 

       86   // Otherwise, add to symbol to table. Copy to a C string first.

       87   char stack_buf[128];

       88   ResourceMark rm(THREAD);

       89   if (len <= 128) {

       90     buffer = stack_buf;

       91   } else {

       92     buffer = NEW_RESOURCE_ARRAY_IN_THREAD(THREAD, char, len);

       93   }

       94   for (int i=0; i<len; i++) {

       95     buffer[i] = name[i];

       96   }

       97   // Make sure there is no safepoint in the code above since name can't move.

       98   // We can't include the code in No_Safepoint_Verifier because of the

       99   // ResourceMark.

      100 

      101   return the_table()->basic_add(index, (u1*)buffer, len, hashValue, CHECK_NULL);

      102 }

      103 

      104 symbolOop SymbolTable::lookup_only(const char* name, int len,

      105                                    unsigned int& hash) {

      106   hash = hash_symbol(name, len);

      107   int index = the_table()->hash_to_index(hash);

      108 

      109   return the_table()->lookup(index, name, len, hash);

      110 }

      111 

      112 // Suggestion: Push unicode-based lookup all the way into the hashing

      113 // and probing logic, so there is no need for convert_to_utf8 until

      114 // an actual new symbolOop is created.

      115 symbolOop SymbolTable::lookup_unicode(const jchar* name, int utf16_length, TRAPS) {

      116   int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);

      117   char stack_buf[128];

      118   if (utf8_length < (int) sizeof(stack_buf)) {

      119     char* chars = stack_buf;

      120     UNICODE::convert_to_utf8(name, utf16_length, chars);

      121     return lookup(chars, utf8_length, THREAD);

      122   } else {

      123     ResourceMark rm(THREAD);

      124     char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;

      125     UNICODE::convert_to_utf8(name, utf16_length, chars);

      126     return lookup(chars, utf8_length, THREAD);

      127   }

      128 }

      129 

      130 symbolOop SymbolTable::lookup_only_unicode(const jchar* name, int utf16_length,

      131                                            unsigned int& hash) {

      132   int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);

      133   char stack_buf[128];

      134   if (utf8_length < (int) sizeof(stack_buf)) {

      135     char* chars = stack_buf;

      136     UNICODE::convert_to_utf8(name, utf16_length, chars);

      137     return lookup_only(chars, utf8_length, hash);

      138   } else {

      139     ResourceMark rm;

      140     char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;

      141     UNICODE::convert_to_utf8(name, utf16_length, chars);

      142     return lookup_only(chars, utf8_length, hash);

      143   }

      144 }

      145 

      146 void SymbolTable::add(constantPoolHandle cp, int names_count,

      147                       const char** names, int* lengths, int* cp_indices,

      148                       unsigned int* hashValues, TRAPS) {

      149   SymbolTable* table = the_table();

      150   bool added = table->basic_add(cp, names_count, names, lengths,

      151                                 cp_indices, hashValues, CHECK);

      152   if (!added) {

      153     // do it the hard way

      154     for (int i=0; i<names_count; i++) {

      155       int index = table->hash_to_index(hashValues[i]);

      156       symbolOop sym = table->basic_add(index, (u1*)names[i], lengths[i],

      157                                        hashValues[i], CHECK);

      158       cp->symbol_at_put(cp_indices[i], sym);

      159     }

      160   }

      161 }

      162 

      163 symbolOop SymbolTable::basic_add(int index, u1 *name, int len,

      164                                  unsigned int hashValue, TRAPS) {

      165   assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(),

      166          "proposed name of symbol must be stable");

      167 

      168   // We assume that lookup() has been called already, that it failed,

      169   // and symbol was not found.  We create the symbol here.

      170   symbolKlass* sk  = (symbolKlass*) Universe::symbolKlassObj()->klass_part();

      171   symbolOop s_oop = sk->allocate_symbol(name, len, CHECK_NULL);

      172   symbolHandle sym (THREAD, s_oop);

      173 

      174   // Allocation must be done before grapping the SymbolTable_lock lock

      175   MutexLocker ml(SymbolTable_lock, THREAD);

      176 

      177   assert(sym->equals((char*)name, len), "symbol must be properly initialized");

      178 

      179   // Since look-up was done lock-free, we need to check if another

      180   // thread beat us in the race to insert the symbol.

      181 

      182   symbolOop test = lookup(index, (char*)name, len, hashValue);

      183   if (test != NULL) {

      184     // A race occurred and another thread introduced the symbol, this one

      185     // will be dropped and collected.

      186     return test;

      187   }

      188 

      189   HashtableEntry* entry = new_entry(hashValue, sym());

      190   add_entry(index, entry);

      191   return sym();

      192 }

      193 

      194 bool SymbolTable::basic_add(constantPoolHandle cp, int names_count,

      195                             const char** names, int* lengths,

      196                             int* cp_indices, unsigned int* hashValues,

      197                             TRAPS) {

      198   symbolKlass* sk  = (symbolKlass*) Universe::symbolKlassObj()->klass_part();

      199   symbolOop sym_oops[symbol_alloc_batch_size];

      200   bool allocated = sk->allocate_symbols(names_count, names, lengths,

      201                                         sym_oops, CHECK_false);

      202   if (!allocated) {

      203     return false;

      204   }

      205   symbolHandle syms[symbol_alloc_batch_size];

      206   int i;

      207   for (i=0; i<names_count; i++) {

      208     syms[i] = symbolHandle(THREAD, sym_oops[i]);

      209   }

      210 

      211   // Allocation must be done before grabbing the SymbolTable_lock lock

      212   MutexLocker ml(SymbolTable_lock, THREAD);

      213 

      214   for (i=0; i<names_count; i++) {

      215     assert(syms[i]->equals(names[i], lengths[i]), "symbol must be properly initialized");

      216     // Since look-up was done lock-free, we need to check if another

      217     // thread beat us in the race to insert the symbol.

      218     int index = hash_to_index(hashValues[i]);

      219     symbolOop test = lookup(index, names[i], lengths[i], hashValues[i]);

      220     if (test != NULL) {

      221       // A race occurred and another thread introduced the symbol, this one

      222       // will be dropped and collected. Use test instead.

      223       cp->symbol_at_put(cp_indices[i], test);

      224     } else {

      225       symbolOop sym = syms[i]();

      226       HashtableEntry* entry = new_entry(hashValues[i], sym);

      227       add_entry(index, entry);

      228       cp->symbol_at_put(cp_indices[i], sym);

      229     }

      230   }

      231 

      232   return true;

      233 }

      234 

      235 

      236 void SymbolTable::verify() {

      237   for (int i = 0; i < the_table()->table_size(); ++i) {

      238     HashtableEntry* p = the_table()->bucket(i);

      239     for ( ; p != NULL; p = p->next()) {

      240       symbolOop s = symbolOop(p->literal());

      241       guarantee(s != NULL, "symbol is NULL");

      242       s->verify();

      243       guarantee(s->is_perm(), "symbol not in permspace");

      244       unsigned int h = hash_symbol((char*)s->bytes(), s->utf8_length());

      245       guarantee(p->hash() == h, "broken hash in symbol table entry");

      246       guarantee(the_table()->hash_to_index(h) == i,

      247                 "wrong index in symbol table");

      248     }

      249   }

      250 }

      251 

      252 

      253 //---------------------------------------------------------------------------

      254 // Non-product code

      255 

      256 #ifndef PRODUCT

      257 

      258 void SymbolTable::print_histogram() {

      259   MutexLocker ml(SymbolTable_lock);

      260   const int results_length = 100;

      261   int results[results_length];

      262   int i,j;

      263 

      264   // initialize results to zero

      265   for (j = 0; j < results_length; j++) {

      266     results[j] = 0;

      267   }

      268 

      269   int total = 0;

      270   int max_symbols = 0;

      271   int out_of_range = 0;

      272   for (i = 0; i < the_table()->table_size(); i++) {

      273     HashtableEntry* p = the_table()->bucket(i);

      274     for ( ; p != NULL; p = p->next()) {

      275       int counter = symbolOop(p->literal())->utf8_length();

      276       total += counter;

      277       if (counter < results_length) {

      278         results[counter]++;

      279       } else {

      280         out_of_range++;

      281       }

      282       max_symbols = MAX2(max_symbols, counter);

      283     }

      284   }

      285   tty->print_cr("Symbol Table:");

      286   tty->print_cr("%8s %5d", "Total  ", total);

      287   tty->print_cr("%8s %5d", "Maximum", max_symbols);

      288   tty->print_cr("%8s %3.2f", "Average",

      289           ((float) total / (float) the_table()->table_size()));

      290   tty->print_cr("%s", "Histogram:");

      291   tty->print_cr(" %s %29s", "Length", "Number chains that length");

      292   for (i = 0; i < results_length; i++) {

      293     if (results[i] > 0) {

      294       tty->print_cr("%6d %10d", i, results[i]);

      295     }

      296   }

      297   int line_length = 70;

      298   tty->print_cr("%s %30s", " Length", "Number chains that length");

      299   for (i = 0; i < results_length; i++) {

      300     if (results[i] > 0) {

      301       tty->print("%4d", i);

      302       for (j = 0; (j < results[i]) && (j < line_length);  j++) {

      303         tty->print("%1s", "*");

      304       }

      305       if (j == line_length) {

      306         tty->print("%1s", "+");

      307       }

      308       tty->cr();

      309     }

      310   }

      311   tty->print_cr(" %s %d: %d\n", "Number chains longer than",

      312                     results_length, out_of_range);

      313 }

      314 

      315 #endif // PRODUCT

      316 

      317 // --------------------------------------------------------------------------

      318 

      319 #ifdef ASSERT

      320 class StableMemoryChecker : public StackObj {

      321   enum { _bufsize = wordSize*4 };

      322 

      323   address _region;

      324   jint    _size;

      325   u1      _save_buf[_bufsize];

      326 

      327   int sample(u1* save_buf) {

      328     if (_size <= _bufsize) {

      329       memcpy(save_buf, _region, _size);

      330       return _size;

      331     } else {

      332       // copy head and tail

      333       memcpy(&save_buf[0],          _region,                      _bufsize/2);

      334       memcpy(&save_buf[_bufsize/2], _region + _size - _bufsize/2, _bufsize/2);

      335       return (_bufsize/2)*2;

      336     }

      337   }

      338 

      339  public:

      340   StableMemoryChecker(const void* region, jint size) {

      341     _region = (address) region;

      342     _size   = size;

      343     sample(_save_buf);

      344   }

      345 

      346   bool verify() {

      347     u1 check_buf[sizeof(_save_buf)];

      348     int check_size = sample(check_buf);

      349     return (0 == memcmp(_save_buf, check_buf, check_size));

      350   }

      351 

      352   void set_region(const void* region) { _region = (address) region; }

      353 };

      354 #endif

      355 

      356 

      357 // --------------------------------------------------------------------------

      358 

      359 

      360 // Compute the hash value for a java.lang.String object which would

      361 // contain the characters passed in. This hash value is used for at

      362 // least two purposes.

      363 //

      364 // (a) As the hash value used by the StringTable for bucket selection

      365 //     and comparison (stored in the HashtableEntry structures).  This

      366 //     is used in the String.intern() method.

      367 //

      368 // (b) As the hash value used by the String object itself, in

      369 //     String.hashCode().  This value is normally calculate in Java code

      370 //     in the String.hashCode method(), but is precomputed for String

      371 //     objects in the shared archive file.

      372 //

      373 //     For this reason, THIS ALGORITHM MUST MATCH String.hashCode().

      374 

      375 int StringTable::hash_string(jchar* s, int len) {

      376   unsigned h = 0;

      377   while (len-- > 0) {

      378     h = 31*h + (unsigned) *s;

      379     s++;

      380   }

      381   return h;

      382 }

      383 

      384 

      385 StringTable* StringTable::_the_table = NULL;

      386 

      387 oop StringTable::lookup(int index, jchar* name,

      388                         int len, unsigned int hash) {

      389   for (HashtableEntry* l = bucket(index); l != NULL; l = l->next()) {

      390     if (l->hash() == hash) {

      391       if (java_lang_String::equals(l->literal(), name, len)) {

      392         return l->literal();

      393       }

      394     }

      395   }

      396   return NULL;

      397 }

      398 

      399 

      400 oop StringTable::basic_add(int index, Handle string_or_null, jchar* name,

      401                            int len, unsigned int hashValue, TRAPS) {

      402   debug_only(StableMemoryChecker smc(name, len * sizeof(name[0])));

      403   assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(),

      404          "proposed name of symbol must be stable");

      405 

      406   Handle string;

      407   // try to reuse the string if possible

      408   if (!string_or_null.is_null() && string_or_null()->is_perm()) {

      409     string = string_or_null;

      410   } else {

      411     string = java_lang_String::create_tenured_from_unicode(name, len, CHECK_NULL);

      412   }

      413 

      414   // Allocation must be done before grapping the SymbolTable_lock lock

      415   MutexLocker ml(StringTable_lock, THREAD);

      416 

      417   assert(java_lang_String::equals(string(), name, len),

      418          "string must be properly initialized");

      419 

      420   // Since look-up was done lock-free, we need to check if another

      421   // thread beat us in the race to insert the symbol.

      422 

      423   oop test = lookup(index, name, len, hashValue); // calls lookup(u1*, int)

      424   if (test != NULL) {

      425     // Entry already added

      426     return test;

      427   }

      428 

      429   HashtableEntry* entry = new_entry(hashValue, string());

      430   add_entry(index, entry);

      431   return string();

      432 }

      433 

      434 

      435 oop StringTable::lookup(symbolOop symbol) {

      436   ResourceMark rm;

      437   int length;

      438   jchar* chars = symbol->as_unicode(length);

      439   unsigned int hashValue = hash_string(chars, length);

      440   int index = the_table()->hash_to_index(hashValue);

      441   return the_table()->lookup(index, chars, length, hashValue);

      442 }

      443 

      444 

      445 oop StringTable::intern(Handle string_or_null, jchar* name,

      446                         int len, TRAPS) {

      447   unsigned int hashValue = hash_string(name, len);

      448   int index = the_table()->hash_to_index(hashValue);

      449   oop string = the_table()->lookup(index, name, len, hashValue);

      450 

      451   // Found

      452   if (string != NULL) return string;

      453 

      454   // Otherwise, add to symbol to table

      455   return the_table()->basic_add(index, string_or_null, name, len,

      456                                 hashValue, CHECK_NULL);

      457 }

      458 

      459 oop StringTable::intern(symbolOop symbol, TRAPS) {

      460   if (symbol == NULL) return NULL;

      461   ResourceMark rm(THREAD);

      462   int length;

      463   jchar* chars = symbol->as_unicode(length);

      464   Handle string;

      465   oop result = intern(string, chars, length, CHECK_NULL);

      466   return result;

      467 }

      468 

      469 

      470 oop StringTable::intern(oop string, TRAPS)

      471 {

      472   if (string == NULL) return NULL;

      473   ResourceMark rm(THREAD);

      474   int length;

      475   Handle h_string (THREAD, string);

      476   jchar* chars = java_lang_String::as_unicode_string(string, length);

      477   oop result = intern(h_string, chars, length, CHECK_NULL);

      478   return result;

      479 }

      480 

      481 

      482 oop StringTable::intern(const char* utf8_string, TRAPS) {

      483   if (utf8_string == NULL) return NULL;

      484   ResourceMark rm(THREAD);

      485   int length = UTF8::unicode_length(utf8_string);

      486   jchar* chars = NEW_RESOURCE_ARRAY(jchar, length);

      487   UTF8::convert_to_unicode(utf8_string, chars, length);

      488   Handle string;

      489   oop result = intern(string, chars, length, CHECK_NULL);

      490   return result;

      491 }

      492 

      493 void StringTable::verify() {

      494   for (int i = 0; i < the_table()->table_size(); ++i) {

      495     HashtableEntry* p = the_table()->bucket(i);

      496     for ( ; p != NULL; p = p->next()) {

      497       oop s = p->literal();

      498       guarantee(s != NULL, "interned string is NULL");

      499       guarantee(s->is_perm(), "interned string not in permspace");

      500 

      501       int length;

      502       jchar* chars = java_lang_String::as_unicode_string(s, length);

      503       unsigned int h = hash_string(chars, length);

      504       guarantee(p->hash() == h, "broken hash in string table entry");

      505       guarantee(the_table()->hash_to_index(h) == i,

      506                 "wrong index in string table");

      507     }

      508   }

      509 }