1 /*
   2  * Copyright (c) 1997, 2019, 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_OOPS_SYMBOL_HPP
  26 #define SHARE_OOPS_SYMBOL_HPP
  27 
  28 #include "memory/allocation.hpp"
  29 #include "utilities/exceptions.hpp"
  30 #include "utilities/macros.hpp"
  31 #include "utilities/utf8.hpp"
  32 
  33 // A Symbol is a canonicalized string.
  34 // All Symbols reside in global SymbolTable and are reference counted.
  35 
  36 // Reference counting
  37 //
  38 // All Symbols are allocated and added to the SymbolTable.
  39 // When a class is unloaded, the reference counts of the Symbol pointers in
  40 // the ConstantPool and in InstanceKlass (see release_C_heap_structures) are
  41 // decremented.  When the reference count for a Symbol goes to 0, the garbage
  42 // collector can free the Symbol and remove it from the SymbolTable.
  43 //
  44 // 0) Symbols need to be reference counted when a pointer to the Symbol is
  45 // saved in persistent storage.  This does not include the pointer
  46 // in the SymbolTable bucket (the _literal field in HashtableEntry)
  47 // that points to the Symbol.  All other stores of a Symbol*
  48 // to a field of a persistent variable (e.g., the _name filed in
  49 // fieldDescriptor or _ptr in a CPSlot) is reference counted.
  50 //
  51 // 1) The lookup of a "name" in the SymbolTable either creates a Symbol F for
  52 // "name" and returns a pointer to F or finds a pre-existing Symbol F for
  53 // "name" and returns a pointer to it. In both cases the reference count for F
  54 // is incremented under the assumption that a pointer to F will be created from
  55 // the return value. Thus the increment of the reference count is on the lookup
  56 // and not on the assignment to the new Symbol*.  That is
  57 //    Symbol* G = lookup()
  58 //                ^ increment on lookup()
  59 // and not
  60 //    Symbol* G = lookup()
  61 //              ^ increment on assignmnet
  62 // The reference count must be decremented manually when the copy of the
  63 // pointer G is destroyed.
  64 //
  65 // 2) For a local Symbol* A that is a copy of an existing Symbol* B, the
  66 // reference counting is elided when the scope of B is greater than the scope
  67 // of A.  For example, in the code fragment
  68 // below "klass" is passed as a parameter to the method.  Symbol* "kn"
  69 // is a copy of the name in "klass".
  70 //
  71 //   Symbol*  kn = klass->name();
  72 //   unsigned int d_hash = dictionary()->compute_hash(kn, class_loader);
  73 //
  74 // The scope of "klass" is greater than the scope of "kn" so the reference
  75 // counting for "kn" is elided.
  76 //
  77 // Symbol* copied from ConstantPool entries are good candidates for reference
  78 // counting elision.  The ConstantPool entries for a class C exist until C is
  79 // unloaded.  If a Symbol* is copied out of the ConstantPool into Symbol* X,
  80 // the Symbol* in the ConstantPool will in general out live X so the reference
  81 // counting on X can be elided.
  82 //
  83 // For cases where the scope of A is not greater than the scope of B,
  84 // the reference counting is explicitly done.  See ciSymbol,
  85 // ResolutionErrorEntry and ClassVerifier for examples.
  86 //
  87 // 3) When a Symbol K is created for temporary use, generally for substrings of
  88 // an existing symbol or to create a new symbol, assign it to a
  89 // TempNewSymbol. The SymbolTable methods new_symbol(), lookup()
  90 // and probe() all potentially return a pointer to a new Symbol.
  91 // The allocation (or lookup) of K increments the reference count for K
  92 // and the destructor decrements the reference count.
  93 //
  94 // This cannot be inherited from ResourceObj because it cannot have a vtable.
  95 // Since sometimes this is allocated from Metadata, pick a base allocation
  96 // type without virtual functions.
  97 class ClassLoaderData;
  98 
  99 // Set _refcount to PERM_REFCOUNT to prevent the Symbol from being freed.
 100 #ifndef PERM_REFCOUNT
 101 #define PERM_REFCOUNT ((1 << 16) - 1)
 102 #endif
 103 
 104 class Symbol : public MetaspaceObj {
 105   friend class VMStructs;
 106   friend class SymbolTable;
 107 
 108  private:
 109 
 110   // This is an int because it needs atomic operation on the refcount.  Mask length
 111   // in high half word. length is the number of UTF8 characters in the symbol
 112   volatile uint32_t _length_and_refcount;
 113   short _identity_hash;
 114   u1 _body[2];
 115 
 116   enum {
 117     // max_symbol_length must fit into the top 16 bits of _length_and_refcount
 118     max_symbol_length = (1 << 16) -1
 119   };
 120 
 121   static int byte_size(int length) {
 122     // minimum number of natural words needed to hold these bits (no non-heap version)
 123     return (int)(sizeof(Symbol) + (length > 2 ? length - 2 : 0));
 124   }
 125   static int size(int length) {
 126     // minimum number of natural words needed to hold these bits (no non-heap version)
 127     return (int)heap_word_size(byte_size(length));
 128   }
 129 
 130   void byte_at_put(int index, u1 value) {
 131     assert(index >=0 && index < length(), "symbol index overflow");
 132     _body[index] = value;
 133   }
 134 
 135   Symbol(const u1* name, int length, int refcount);
 136   void* operator new(size_t size, int len, TRAPS) throw();
 137   void* operator new(size_t size, int len, Arena* arena, TRAPS) throw();
 138 
 139   void  operator delete(void* p);
 140 
 141   static int extract_length(uint32_t value)   { return value >> 16; }
 142   static int extract_refcount(uint32_t value) { return value & 0xffff; }
 143   static uint32_t pack_length_and_refcount(int length, int refcount);
 144 
 145   int length() const   { return extract_length(_length_and_refcount); }
 146 
 147  public:
 148   // Low-level access (used with care, since not GC-safe)
 149   const u1* base() const { return &_body[0]; }
 150 
 151   int size()                { return size(utf8_length()); }
 152   int byte_size()           { return byte_size(utf8_length()); }
 153 
 154   // Symbols should be stored in the read-only region of CDS archive.
 155   static bool is_read_only_by_default() { return true; }
 156 
 157   // Returns the largest size symbol we can safely hold.
 158   static int max_length() { return max_symbol_length; }
 159   unsigned identity_hash() const {
 160     unsigned addr_bits = (unsigned)((uintptr_t)this >> (LogMinObjAlignmentInBytes + 3));
 161     return ((unsigned)_identity_hash & 0xffff) |
 162            ((addr_bits ^ (length() << 8) ^ (( _body[0] << 8) | _body[1])) << 16);
 163   }
 164 
 165   // Reference counting.  See comments above this class for when to use.
 166   int refcount() const { return extract_refcount(_length_and_refcount); }
 167   bool try_increment_refcount();
 168   void increment_refcount();
 169   void decrement_refcount();
 170   bool is_permanent() {
 171     return (refcount() == PERM_REFCOUNT);
 172   }
 173 
 174   // Function char_at() returns the Symbol's selected u1 byte as a char type.
 175   //
 176   // Note that all multi-byte chars have the sign bit set on all their bytes.
 177   // No single byte chars have their sign bit set.
 178   char char_at(int index) const {
 179     assert(index >=0 && index < length(), "symbol index overflow");
 180     return (char)base()[index];
 181   }
 182 
 183   const u1* bytes() const { return base(); }
 184 
 185   int utf8_length() const { return length(); }
 186 
 187   // Compares the symbol with a string.
 188   bool equals(const char* str, int len) const {
 189     int l = utf8_length();
 190     if (l != len) return false;
 191     while (l-- > 0) {
 192       if (str[l] != char_at(l))
 193         return false;
 194     }
 195     assert(l == -1, "we should be at the beginning");
 196     return true;
 197   }
 198   bool equals(const char* str) const { return equals(str, (int) strlen(str)); }
 199 
 200   // Tests if the symbol starts with the given prefix.
 201   bool starts_with(const char* prefix, int len) const;
 202   bool starts_with(const char* prefix) const {
 203     return starts_with(prefix, (int) strlen(prefix));
 204   }
 205   bool is_Q_signature() const;
 206   bool is_Q_array_signature() const;
 207   Symbol* fundamental_name(TRAPS);
 208   bool is_same_fundamental_type(Symbol*) const;
 209 
 210   // Tests if the symbol starts with the given prefix.
 211   int index_of_at(int i, const char* str, int len) const;
 212 
 213   // Three-way compare for sorting; returns -1/0/1 if receiver is </==/> than arg
 214   // note that the ordering is not alfabetical
 215   inline int fast_compare(const Symbol* other) const;
 216 
 217   // Returns receiver converted to null-terminated UTF-8 string; string is
 218   // allocated in resource area, or in the char buffer provided by caller.
 219   char* as_C_string() const;
 220   char* as_C_string(char* buf, int size) const;
 221 
 222   // Returns an escaped form of a Java string.
 223   char* as_quoted_ascii() const;
 224 
 225   // Returns a null terminated utf8 string in a resource array
 226   char* as_utf8() const { return as_C_string(); }
 227 
 228   jchar* as_unicode(int& length) const;
 229 
 230   // Treating this symbol as a class name, returns the Java name for the class.
 231   // String is allocated in resource area if buffer is not provided.
 232   // See Klass::external_name()
 233   const char* as_klass_external_name() const;
 234   const char* as_klass_external_name(char* buf, int size) const;
 235 
 236   void metaspace_pointers_do(MetaspaceClosure* it);
 237   MetaspaceObj::Type type() const { return SymbolType; }
 238 
 239   // Printing
 240   void print_symbol_on(outputStream* st = NULL) const;
 241   void print_utf8_on(outputStream* st) const;
 242   void print_on(outputStream* st) const;         // First level print
 243   void print_value_on(outputStream* st) const;   // Second level print.
 244   void print_Qvalue_on(outputStream* st) const;  // Second level print for Q-types.
 245 
 246   // printing on default output stream
 247   void print()         { print_on(tty);       }
 248   void print_value()   { print_value_on(tty); }
 249 
 250   static bool is_valid(Symbol* s);
 251 
 252 #ifndef PRODUCT
 253   // Empty constructor to create a dummy symbol object on stack
 254   // only for getting its vtable pointer.
 255   Symbol() { }
 256 
 257   static size_t _total_count;
 258 #endif
 259 };
 260 
 261 // Note: this comparison is used for vtable sorting only; it doesn't matter
 262 // what order it defines, as long as it is a total, time-invariant order
 263 // Since Symbol*s are in C_HEAP, their relative order in memory never changes,
 264 // so use address comparison for speed
 265 int Symbol::fast_compare(const Symbol* other) const {
 266  return (((uintptr_t)this < (uintptr_t)other) ? -1
 267    : ((uintptr_t)this == (uintptr_t) other) ? 0 : 1);
 268 }
 269 #endif // SHARE_OOPS_SYMBOL_HPP