1 /*

     2  * Copyright (c) 1997, 2009, 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 #ifndef SHARE_VM_OOPS_SYMBOL_HPP

    26 #define SHARE_VM_OOPS_SYMBOL_HPP

    27 

    28 #include "utilities/utf8.hpp"

    29 #include "memory/allocation.hpp"

    30 

    31 // A Symbol is a canonicalized string.

    32 // All Symbols reside in global SymbolTable and are reference counted.

    33 

    34 // Reference counting

    35 //

    36 // All Symbols are allocated and added to the SymbolTable.

    37 // When a class is unloaded, the reference counts of the Symbol pointers in

    38 // the ConstantPool and in instanceKlass (see release_C_heap_structures) are

    39 // decremented.  When the reference count for a Symbol goes to 0, the garbage

    40 // collector can free the Symbol and remove it from the SymbolTable.

    41 //

    42 // 0) Symbols need to be reference counted when a pointer to the Symbol is

    43 // saved in persistent storage.  This does not include the pointer

    44 // in the SymbolTable bucket (the _literal field in HashtableEntry)

    45 // that points to the Symbol.  All other stores of a Symbol*

    46 // to a field of a persistent variable (e.g., the _name filed in

    47 // FieldAccessInfo or _ptr in a CPSlot) is reference counted.

    48 //

    49 // 1) The lookup of a "name" in the SymbolTable either creates a Symbol F for

    50 // "name" and returns a pointer to F or finds a pre-existing Symbol F for

    51 // "name" and returns a pointer to it. In both cases the reference count for F

    52 // is incremented under the assumption that a pointer to F will be created from

    53 // the return value. Thus the increment of the reference count is on the lookup

    54 // and not on the assignment to the new Symbol*.  That is

    55 //    Symbol* G = lookup()

    56 //                ^ increment on lookup()

    57 // and not

    58 //    Symbol* G = lookup()

    59 //              ^ increment on assignmnet

    60 // The reference count must be decremented manually when the copy of the

    61 // pointer G is destroyed.

    62 //

    63 // 2) For a local Symbol* A that is a copy of an existing Symbol* B, the

    64 // reference counting is elided when the scope of B is greater than the scope

    65 // of A.  For example, in the code fragment

    66 // below "klass" is passed as a parameter to the method.  Symbol* "kn"

    67 // is a copy of the name in "klass".

    68 //

    69 //   Symbol*  kn = klass->name();

    70 //   unsigned int d_hash = dictionary()->compute_hash(kn, class_loader);

    71 //

    72 // The scope of "klass" is greater than the scope of "kn" so the reference

    73 // counting for "kn" is elided.

    74 //

    75 // Symbol* copied from ConstantPool entries are good candidates for reference

    76 // counting elision.  The ConstantPool entries for a class C exist until C is

    77 // unloaded.  If a Symbol* is copied out of the ConstantPool into Symbol* X,

    78 // the Symbol* in the ConstantPool will in general out live X so the reference

    79 // counting on X can be elided.

    80 //

    81 // For cases where the scope of A is not greater than the scope of B,

    82 // the reference counting is explicitly done.  See ciSymbol,

    83 // ResolutionErrorEntry and ClassVerifier for examples.

    84 //

    85 // 3) When a Symbol K is created for temporary use, generally for substrings of

    86 // an existing symbol or to create a new symbol, assign it to a

    87 // TempNewSymbol. The SymbolTable methods new_symbol(), lookup()

    88 // and probe() all potentially return a pointer to a new Symbol.

    89 // The allocation (or lookup) of K increments the reference count for K

    90 // and the destructor decrements the reference count.

    91 //

    92 // Another example of TempNewSymbol usage is parsed_name used in

    93 // ClassFileParser::parseClassFile() where parsed_name is used in the cleanup

    94 // after a failed attempt to load a class.  Here parsed_name is a

    95 // TempNewSymbol (passed in as a parameter) so the reference count on its symbol

    96 // will be decremented when it goes out of scope.

    97 

    98 class Symbol : public CHeapObj {

    99   friend class VMStructs;

   100   friend class SymbolTable;

   101   friend class MoveSymbols;

   102  private:

   103   volatile int   _refcount;

   104   int            _identity_hash;

   105   unsigned short _length; // number of UTF8 characters in the symbol

   106   jbyte _body[1];

   107 

   108   enum {

   109     // max_symbol_length is constrained by type of _length

   110     max_symbol_length = (1 << 16) -1

   111   };

   112 

   113   static int object_size(int length) {

   114     size_t size = heap_word_size(sizeof(Symbol) + length);

   115     return align_object_size(size);

   116   }

   117 

   118   void byte_at_put(int index, int value) {

   119     assert(index >=0 && index < _length, "symbol index overflow");

   120     _body[index] = value;

   121   }

   122 

   123   Symbol(const u1* name, int length);

   124   void* operator new(size_t size, int len);

   125 

   126  public:

   127   // Low-level access (used with care, since not GC-safe)

   128   const jbyte* base() const { return &_body[0]; }

   129 

   130   int object_size() { return object_size(utf8_length()); }

   131 

   132   // Returns the largest size symbol we can safely hold.

   133   static int max_length() {

   134     return max_symbol_length;

   135   }

   136 

   137   int identity_hash() {

   138     return _identity_hash;

   139   }

   140 

   141   // Reference counting.  See comments above this class for when to use.

   142   int refcount() const { return _refcount; }

   143   void increment_refcount();

   144   void decrement_refcount();

   145 

   146   int byte_at(int index) const {

   147     assert(index >=0 && index < _length, "symbol index overflow");

   148     return base()[index];

   149   }

   150 

   151   const jbyte* bytes() const { return base(); }

   152 

   153   int utf8_length() const { return _length; }

   154 

   155   // Compares the symbol with a string.

   156   bool equals(const char* str, int len) const;

   157   bool equals(const char* str) const { return equals(str, (int) strlen(str)); }

   158 

   159   // Tests if the symbol starts with the given prefix.

   160   bool starts_with(const char* prefix, int len) const;

   161   bool starts_with(const char* prefix) const {

   162     return starts_with(prefix, (int) strlen(prefix));

   163   }

   164 

   165   // Tests if the symbol starts with the given prefix.

   166   int index_of_at(int i, const char* str, int len) const;

   167   int index_of_at(int i, const char* str) const {

   168     return index_of_at(i, str, (int) strlen(str));

   169   }

   170 

   171   // Three-way compare for sorting; returns -1/0/1 if receiver is </==/> than arg

   172   // note that the ordering is not alfabetical

   173   inline int fast_compare(Symbol* other) const;

   174 

   175   // Returns receiver converted to null-terminated UTF-8 string; string is

   176   // allocated in resource area, or in the char buffer provided by caller.

   177   char* as_C_string() const;

   178   char* as_C_string(char* buf, int size) const;

   179   // Use buf if needed buffer length is <= size.

   180   char* as_C_string_flexible_buffer(Thread* t, char* buf, int size) const;

   181 

   182 

   183   // Returns a null terminated utf8 string in a resource array

   184   char* as_utf8() const { return as_C_string(); }

   185   char* as_utf8_flexible_buffer(Thread* t, char* buf, int size) const {

   186     return as_C_string_flexible_buffer(t, buf, size);

   187   }

   188 

   189   jchar* as_unicode(int& length) const;

   190 

   191   // Treating this symbol as a class name, returns the Java name for the class.

   192   // String is allocated in resource area if buffer is not provided.

   193   // See Klass::external_name()

   194   const char* as_klass_external_name() const;

   195   const char* as_klass_external_name(char* buf, int size) const;

   196 

   197   // Printing

   198   void print_symbol_on(outputStream* st = NULL) const;

   199   void print_on(outputStream* st) const;         // First level print

   200   void print_value_on(outputStream* st) const;   // Second level print.

   201 

   202   // printing on default output stream

   203   void print()         { print_on(tty);       }

   204   void print_value()   { print_value_on(tty); }

   205 

   206 #ifndef PRODUCT

   207   // Empty constructor to create a dummy symbol object on stack

   208   // only for getting its vtable pointer.

   209   Symbol() { }

   210 

   211   static int _total_count;

   212 #endif

   213 };

   214 

   215 // Note: this comparison is used for vtable sorting only; it doesn't matter

   216 // what order it defines, as long as it is a total, time-invariant order

   217 // Since Symbol*s are in C_HEAP, their relative order in memory never changes,

   218 // so use address comparison for speed

   219 int Symbol::fast_compare(Symbol* other) const {

   220  return (((uintptr_t)this < (uintptr_t)other) ? -1

   221    : ((uintptr_t)this == (uintptr_t) other) ? 0 : 1);

   222 }

   223 #endif // SHARE_VM_OOPS_SYMBOL_HPP