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 #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 // FieldAccessInfo 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 // We separate the fields in SymbolBase from Symbol::_body so that
106 // Symbol::size(int) can correctly calculate the space needed.
107 class SymbolBase : public MetaspaceObj {
108 friend class VMStructs;
109 friend class SymbolTable;
110 friend class MoveSymbols;
111 public:
112 ATOMIC_SHORT_PAIR(
113 volatile short _refcount, // needs atomic operation
114 unsigned short _length // number of UTF8 characters in the symbol (does not need atomic op)
115 );
116 int _identity_hash;
117 };
118
119 class Symbol : private SymbolBase {
120 friend class VMStructs;
121 friend class SymbolTable;
122 friend class MoveSymbols;
123 private:
124 jbyte _body[1];
125
126 enum {
127 // max_symbol_length is constrained by type of _length
128 max_symbol_length = (1 << 16) -1
129 };
130
131 static int size(int length) {
132 size_t sz = heap_word_size(sizeof(SymbolBase) + (length > 0 ? length : 0));
133 return align_object_size(sz);
134 }
135
136 void byte_at_put(int index, int value) {
137 assert(index >=0 && index < _length, "symbol index overflow");
138 _body[index] = value;
139 }
140
141 Symbol(const u1* name, int length, int refcount);
142 void* operator new(size_t size, int len, TRAPS);
143 void* operator new(size_t size, int len, Arena* arena, TRAPS);
144 void* operator new(size_t size, int len, ClassLoaderData* loader_data, TRAPS);
145
146 void operator delete(void* p);
147
148 public:
149 // Low-level access (used with care, since not GC-safe)
150 const jbyte* base() const { return &_body[0]; }
151
152 int size() { return size(utf8_length()); }
153
154 // Returns the largest size symbol we can safely hold.
155 static int max_length() { return max_symbol_length; }
156
157 int identity_hash() { return _identity_hash; }
158
159 // For symbol table alternate hashing
160 unsigned int new_hash(jint seed);
161
162 // Reference counting. See comments above this class for when to use.
163 int refcount() const { return _refcount; }
164 void increment_refcount();
165 void decrement_refcount();
166
167 int byte_at(int index) const {
168 assert(index >=0 && index < _length, "symbol index overflow");
169 return base()[index];
170 }
171
172 const jbyte* bytes() const { return base(); }
173
174 int utf8_length() const { return _length; }
175
176 // Compares the symbol with a string.
177 bool equals(const char* str, int len) const;
178 bool equals(const char* str) const { return equals(str, (int) strlen(str)); }
179
180 // Tests if the symbol starts with the given prefix.
181 bool starts_with(const char* prefix, int len) const;
182 bool starts_with(const char* prefix) const {
183 return starts_with(prefix, (int) strlen(prefix));
184 }
185
186 // Tests if the symbol starts with the given prefix.
187 int index_of_at(int i, const char* str, int len) const;
188 int index_of_at(int i, const char* str) const {
189 return index_of_at(i, str, (int) strlen(str));
190 }
191
192 // Three-way compare for sorting; returns -1/0/1 if receiver is </==/> than arg
193 // note that the ordering is not alfabetical
194 inline int fast_compare(Symbol* other) const;
195
196 // Returns receiver converted to null-terminated UTF-8 string; string is
197 // allocated in resource area, or in the char buffer provided by caller.
198 char* as_C_string() const;
199 char* as_C_string(char* buf, int size) const;
200 // Use buf if needed buffer length is <= size.
201 char* as_C_string_flexible_buffer(Thread* t, char* buf, int size) const;
202
203 // Returns an escaped form of a Java string.
204 char* as_quoted_ascii() const;
205
206 // Returns a null terminated utf8 string in a resource array
207 char* as_utf8() const { return as_C_string(); }
208 char* as_utf8_flexible_buffer(Thread* t, char* buf, int size) const {
209 return as_C_string_flexible_buffer(t, buf, size);
210 }
211
212 jchar* as_unicode(int& length) const;
213
214 // Treating this symbol as a class name, returns the Java name for the class.
215 // String is allocated in resource area if buffer is not provided.
216 // See Klass::external_name()
217 const char* as_klass_external_name() const;
218 const char* as_klass_external_name(char* buf, int size) const;
219
220 // Printing
221 void print_symbol_on(outputStream* st = NULL) const;
222 void print_on(outputStream* st) const; // First level print
223 void print_value_on(outputStream* st) const; // Second level print.
224
225 // printing on default output stream
226 void print() { print_on(tty); }
227 void print_value() { print_value_on(tty); }
228
229 #ifndef PRODUCT
230 // Empty constructor to create a dummy symbol object on stack
231 // only for getting its vtable pointer.
232 Symbol() { }
233
234 static int _total_count;
235 #endif
236 };
237
238 // Note: this comparison is used for vtable sorting only; it doesn't matter
239 // what order it defines, as long as it is a total, time-invariant order
240 // Since Symbol*s are in C_HEAP, their relative order in memory never changes,
241 // so use address comparison for speed
242 int Symbol::fast_compare(Symbol* other) const {
243 return (((uintptr_t)this < (uintptr_t)other) ? -1
244 : ((uintptr_t)this == (uintptr_t) other) ? 0 : 1);
245 }
246 #endif // SHARE_VM_OOPS_SYMBOL_HPP