1 /*
2 * Copyright (c) 1997, 2014, 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 #include "runtime/atomic.hpp"
31
32 // A Symbol is a canonicalized string.
33 // All Symbols reside in global SymbolTable and are reference counted.
34
35 // Reference counting
36 //
37 // All Symbols are allocated and added to the SymbolTable.
38 // When a class is unloaded, the reference counts of the Symbol pointers in
39 // the ConstantPool and in InstanceKlass (see release_C_heap_structures) are
40 // decremented. When the reference count for a Symbol goes to 0, the garbage
41 // collector can free the Symbol and remove it from the SymbolTable.
42 //
43 // 0) Symbols need to be reference counted when a pointer to the Symbol is
44 // saved in persistent storage. This does not include the pointer
45 // in the SymbolTable bucket (the _literal field in HashtableEntry)
46 // that points to the Symbol. All other stores of a Symbol*
47 // to a field of a persistent variable (e.g., the _name filed in
48 // fieldDescriptor or _ptr in a CPSlot) is reference counted.
49 //
50 // 1) The lookup of a "name" in the SymbolTable either creates a Symbol F for
51 // "name" and returns a pointer to F or finds a pre-existing Symbol F for
52 // "name" and returns a pointer to it. In both cases the reference count for F
53 // is incremented under the assumption that a pointer to F will be created from
54 // the return value. Thus the increment of the reference count is on the lookup
55 // and not on the assignment to the new Symbol*. That is
56 // Symbol* G = lookup()
57 // ^ increment on lookup()
58 // and not
59 // Symbol* G = lookup()
60 // ^ increment on assignmnet
61 // The reference count must be decremented manually when the copy of the
62 // pointer G is destroyed.
63 //
64 // 2) For a local Symbol* A that is a copy of an existing Symbol* B, the
65 // reference counting is elided when the scope of B is greater than the scope
66 // of A. For example, in the code fragment
67 // below "klass" is passed as a parameter to the method. Symbol* "kn"
68 // is a copy of the name in "klass".
69 //
70 // Symbol* kn = klass->name();
71 // unsigned int d_hash = dictionary()->compute_hash(kn, class_loader);
72 //
73 // The scope of "klass" is greater than the scope of "kn" so the reference
74 // counting for "kn" is elided.
75 //
76 // Symbol* copied from ConstantPool entries are good candidates for reference
77 // counting elision. The ConstantPool entries for a class C exist until C is
78 // unloaded. If a Symbol* is copied out of the ConstantPool into Symbol* X,
79 // the Symbol* in the ConstantPool will in general out live X so the reference
80 // counting on X can be elided.
81 //
82 // For cases where the scope of A is not greater than the scope of B,
83 // the reference counting is explicitly done. See ciSymbol,
84 // ResolutionErrorEntry and ClassVerifier for examples.
85 //
86 // 3) When a Symbol K is created for temporary use, generally for substrings of
87 // an existing symbol or to create a new symbol, assign it to a
88 // TempNewSymbol. The SymbolTable methods new_symbol(), lookup()
89 // and probe() all potentially return a pointer to a new Symbol.
90 // The allocation (or lookup) of K increments the reference count for K
91 // and the destructor decrements the reference count.
92 //
93 // Another example of TempNewSymbol usage is parsed_name used in
94 // ClassFileParser::parseClassFile() where parsed_name is used in the cleanup
95 // after a failed attempt to load a class. Here parsed_name is a
96 // TempNewSymbol (passed in as a parameter) so the reference count on its symbol
97 // will be decremented when it goes out of scope.
98
99
100 // This cannot be inherited from ResourceObj because it cannot have a vtable.
101 // Since sometimes this is allocated from Metadata, pick a base allocation
102 // type without virtual functions.
103 class ClassLoaderData;
104
105 class Symbol : public MetaspaceObj {
106 friend class VMStructs;
107 friend class SymbolTable;
108 friend class MoveSymbols;
109
110 private:
111 ATOMIC_SHORT_PAIR(
112 volatile short _refcount, // needs atomic operation
113 unsigned short _length // number of UTF8 characters in the symbol (does not need atomic op)
114 );
115 short _identity_hash;
116 jbyte _body[2];
117
118 enum {
119 // max_symbol_length is constrained by type of _length
120 max_symbol_length = (1 << 16) -1
121 };
122
123 static int size(int length) {
124 size_t sz = heap_word_size(sizeof(Symbol) + (length > 2 ? length - 2 : 0));
125 return align_object_size(sz);
126 }
127
128 void byte_at_put(int index, int value) {
129 assert(index >=0 && index < _length, "symbol index overflow");
130 _body[index] = value;
131 }
132
133 Symbol(const u1* name, int length, int refcount);
134 void* operator new(size_t size, int len, TRAPS) throw();
135 void* operator new(size_t size, int len, Arena* arena, TRAPS) throw();
136 void* operator new(size_t size, int len, ClassLoaderData* loader_data, TRAPS) throw();
137
138 void operator delete(void* p);
139
140 public:
141 // Low-level access (used with care, since not GC-safe)
142 const jbyte* base() const { return &_body[0]; }
143
144 int size() { return size(utf8_length()); }
145
146 // Returns the largest size symbol we can safely hold.
147 static int max_length() { return max_symbol_length; }
148 unsigned identity_hash() {
149 unsigned addr_bits = (unsigned)((uintptr_t)this >> (LogMinObjAlignmentInBytes + 3));
150 return (unsigned)_identity_hash |
151 ((addr_bits ^ (_length << 8) ^ (( _body[0] << 8) | _body[1])) << 16);
152 }
153
154 // For symbol table alternate hashing
155 unsigned int new_hash(juint seed);
156
157 // Reference counting. See comments above this class for when to use.
158 int refcount() const { return _refcount; }
159 void increment_refcount();
160 void decrement_refcount();
161
162 int byte_at(int index) const {
163 assert(index >=0 && index < _length, "symbol index overflow");
164 return base()[index];
165 }
166
167 const jbyte* bytes() const { return base(); }
168
169 int utf8_length() const { return _length; }
170
171 // Compares the symbol with a string.
172 bool equals(const char* str, int len) const;
173 bool equals(const char* str) const { return equals(str, (int) strlen(str)); }
174
175 // Tests if the symbol starts with the given prefix.
176 bool starts_with(const char* prefix, int len) const;
177 bool starts_with(const char* prefix) const {
178 return starts_with(prefix, (int) strlen(prefix));
179 }
180
181 // Tests if the symbol starts with the given prefix.
182 int index_of_at(int i, const char* str, int len) const;
183 int index_of_at(int i, const char* str) const {
184 return index_of_at(i, str, (int) strlen(str));
185 }
186
187 // Three-way compare for sorting; returns -1/0/1 if receiver is </==/> than arg
188 // note that the ordering is not alfabetical
189 inline int fast_compare(Symbol* other) const;
190
191 // Returns receiver converted to null-terminated UTF-8 string; string is
192 // allocated in resource area, or in the char buffer provided by caller.
193 char* as_C_string() const;
194 char* as_C_string(char* buf, int size) const;
195 // Use buf if needed buffer length is <= size.
196 char* as_C_string_flexible_buffer(Thread* t, char* buf, int size) const;
197
198 // Returns an escaped form of a Java string.
199 char* as_quoted_ascii() const;
200
201 // Returns a null terminated utf8 string in a resource array
202 char* as_utf8() const { return as_C_string(); }
203 char* as_utf8_flexible_buffer(Thread* t, char* buf, int size) const {
204 return as_C_string_flexible_buffer(t, buf, size);
205 }
206
207 jchar* as_unicode(int& length) const;
208
209 // Treating this symbol as a class name, returns the Java name for the class.
210 // String is allocated in resource area if buffer is not provided.
211 // See Klass::external_name()
212 const char* as_klass_external_name() const;
213 const char* as_klass_external_name(char* buf, int size) const;
214
215 // Printing
216 void print_symbol_on(outputStream* st = NULL) const;
217 void print_on(outputStream* st) const; // First level print
218 void print_value_on(outputStream* st) const; // Second level print.
219
220 // printing on default output stream
221 void print() { print_on(tty); }
222 void print_value() { print_value_on(tty); }
223
224 #ifndef PRODUCT
225 // Empty constructor to create a dummy symbol object on stack
226 // only for getting its vtable pointer.
227 Symbol() { }
228
229 static int _total_count;
230 #endif
231 };
232
233 // Note: this comparison is used for vtable sorting only; it doesn't matter
234 // what order it defines, as long as it is a total, time-invariant order
235 // Since Symbol*s are in C_HEAP, their relative order in memory never changes,
236 // so use address comparison for speed
237 int Symbol::fast_compare(Symbol* other) const {
238 return (((uintptr_t)this < (uintptr_t)other) ? -1
239 : ((uintptr_t)this == (uintptr_t) other) ? 0 : 1);
240 }
241 #endif // SHARE_VM_OOPS_SYMBOL_HPP