1 /*
   2  * Copyright (c) 1997, 2013, 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 
  99 // This cannot be inherited from ResourceObj because it cannot have a vtable.
 100 // Since sometimes this is allocated from Metadata, pick a base allocation
 101 // type without virtual functions.
 102 class ClassLoaderData;
 103 
 104 class Symbol : public MetaspaceObj {
 105   friend class VMStructs;
 106   friend class SymbolTable;
 107   friend class MoveSymbols;
 108  private:
 109   volatile int   _refcount;
 110   int            _identity_hash;
 111   unsigned short _length; // number of UTF8 characters in the symbol
 112   jbyte _body[1];
 113 
 114   enum {
 115     // max_symbol_length is constrained by type of _length
 116     max_symbol_length = (1 << 16) -1
 117   };
 118 
 119   static int size(int length) {
 120     size_t sz = heap_word_size(sizeof(Symbol) + (length > 0 ? length - 1 : 0));
 121     return align_object_size(sz);
 122   }
 123 
 124   void byte_at_put(int index, int value) {
 125     assert(index >=0 && index < _length, "symbol index overflow");
 126     _body[index] = value;
 127   }
 128 
 129   Symbol(const u1* name, int length, int refcount);
 130   void* operator new(size_t size, int len, TRAPS);
 131   void* operator new(size_t size, int len, Arena* arena, TRAPS);
 132   void* operator new(size_t size, int len, ClassLoaderData* loader_data, TRAPS);
 133 
 134   void  operator delete(void* p);
 135 
 136  public:
 137   // Low-level access (used with care, since not GC-safe)
 138   const jbyte* base() const { return &_body[0]; }
 139 
 140   int size()                { return size(utf8_length()); }
 141 
 142   // Returns the largest size symbol we can safely hold.
 143   static int max_length()   { return max_symbol_length; }
 144 
 145   int identity_hash()       { return _identity_hash; }
 146 
 147   // For symbol table alternate hashing
 148   unsigned int new_hash(jint seed);
 149 
 150   // Reference counting.  See comments above this class for when to use.
 151   int refcount() const      { return _refcount; }
 152   void increment_refcount();
 153   void decrement_refcount();
 154 
 155   int byte_at(int index) const {
 156     assert(index >=0 && index < _length, "symbol index overflow");
 157     return base()[index];
 158   }
 159 
 160   const jbyte* bytes() const { return base(); }
 161 
 162   int utf8_length() const { return _length; }
 163 
 164   // Compares the symbol with a string.
 165   bool equals(const char* str, int len) const;
 166   bool equals(const char* str) const { return equals(str, (int) strlen(str)); }
 167 
 168   // Tests if the symbol starts with the given prefix.
 169   bool starts_with(const char* prefix, int len) const;
 170   bool starts_with(const char* prefix) const {
 171     return starts_with(prefix, (int) strlen(prefix));
 172   }
 173 
 174   // Tests if the symbol starts with the given prefix.
 175   int index_of_at(int i, const char* str, int len) const;
 176   int index_of_at(int i, const char* str) const {
 177     return index_of_at(i, str, (int) strlen(str));
 178   }
 179 
 180   // Three-way compare for sorting; returns -1/0/1 if receiver is </==/> than arg
 181   // note that the ordering is not alfabetical
 182   inline int fast_compare(Symbol* other) const;
 183 
 184   // Returns receiver converted to null-terminated UTF-8 string; string is
 185   // allocated in resource area, or in the char buffer provided by caller.
 186   char* as_C_string() const;
 187   char* as_C_string(char* buf, int size) const;
 188   // Use buf if needed buffer length is <= size.
 189   char* as_C_string_flexible_buffer(Thread* t, char* buf, int size) const;
 190 
 191   // Returns an escaped form of a Java string.
 192   char* as_quoted_ascii() const;
 193 
 194   // Returns a null terminated utf8 string in a resource array
 195   char* as_utf8() const { return as_C_string(); }
 196   char* as_utf8_flexible_buffer(Thread* t, char* buf, int size) const {
 197     return as_C_string_flexible_buffer(t, buf, size);
 198   }
 199 
 200   jchar* as_unicode(int& length) const;
 201 
 202   // Treating this symbol as a class name, returns the Java name for the class.
 203   // String is allocated in resource area if buffer is not provided.
 204   // See Klass::external_name()
 205   const char* as_klass_external_name() const;
 206   const char* as_klass_external_name(char* buf, int size) const;
 207 
 208   // Printing
 209   void print_symbol_on(outputStream* st = NULL) const;
 210   void print_on(outputStream* st) const;         // First level print
 211   void print_value_on(outputStream* st) const;   // Second level print.
 212 
 213   // printing on default output stream
 214   void print()         { print_on(tty);       }
 215   void print_value()   { print_value_on(tty); }
 216 
 217 #ifndef PRODUCT
 218   // Empty constructor to create a dummy symbol object on stack
 219   // only for getting its vtable pointer.
 220   Symbol() { }
 221 
 222   static int _total_count;
 223 #endif
 224 };
 225 
 226 // Note: this comparison is used for vtable sorting only; it doesn't matter
 227 // what order it defines, as long as it is a total, time-invariant order
 228 // Since Symbol*s are in C_HEAP, their relative order in memory never changes,
 229 // so use address comparison for speed
 230 int Symbol::fast_compare(Symbol* other) const {
 231  return (((uintptr_t)this < (uintptr_t)other) ? -1
 232    : ((uintptr_t)this == (uintptr_t) other) ? 0 : 1);
 233 }
 234 #endif // SHARE_VM_OOPS_SYMBOL_HPP