1 /*
2 * Copyright (c) 1997, 2016, 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 "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 GC'ed.
100 #ifndef PERM_REFCOUNT
101 #define PERM_REFCOUNT -1
102 #endif
103
104 class Symbol : public MetaspaceObj {
105 friend class VMStructs;
106 friend class SymbolTable;
107 friend class MoveSymbols;
108
109 private:
110 ATOMIC_SHORT_PAIR(
111 volatile short _refcount, // needs atomic operation
112 unsigned short _length // number of UTF8 characters in the symbol (does not need atomic op)
113 );
114 short _identity_hash;
115 jbyte _body[2];
116
117 enum {
118 // max_symbol_length is constrained by type of _length
119 max_symbol_length = (1 << 16) -1
120 };
121
122 static int size(int length) {
123 // minimum number of natural words needed to hold these bits (no non-heap version)
124 return (int)heap_word_size(sizeof(Symbol) + (length > 2 ? length - 2 : 0));
125 }
126
127 void byte_at_put(int index, int value) {
128 assert(index >=0 && index < _length, "symbol index overflow");
129 _body[index] = value;
130 }
131
132 Symbol(const u1* name, int length, int refcount);
133 void* operator new(size_t size, int len, TRAPS) throw();
134 void* operator new(size_t size, int len, Arena* arena, TRAPS) throw();
135 void* operator new(size_t size, int len, ClassLoaderData* loader_data, TRAPS) throw();
136
137 void operator delete(void* p);
138
139 public:
140 // Low-level access (used with care, since not GC-safe)
141 const jbyte* base() const { return &_body[0]; }
142
143 int size() { return size(utf8_length()); }
144
145 // Returns the largest size symbol we can safely hold.
146 static int max_length() { return max_symbol_length; }
147 unsigned identity_hash() const {
148 unsigned addr_bits = (unsigned)((uintptr_t)this >> (LogMinObjAlignmentInBytes + 3));
149 return ((unsigned)_identity_hash & 0xffff) |
150 ((addr_bits ^ (_length << 8) ^ (( _body[0] << 8) | _body[1])) << 16);
151 }
152
153 // For symbol table alternate hashing
154 unsigned int new_hash(juint seed);
155
156 // Reference counting. See comments above this class for when to use.
157 int refcount() const { return _refcount; }
158 void increment_refcount();
159 void decrement_refcount();
160 // Set _refcount non zero to avoid being reclaimed by GC.
161 void set_permanent() {
162 assert(LogTouchedMethods, "Should not be called with LogTouchedMethods off");
163 if (_refcount != PERM_REFCOUNT) {
164 _refcount = PERM_REFCOUNT;
165 }
166 }
167
168 int byte_at(int index) const {
169 assert(index >=0 && index < _length, "symbol index overflow");
170 return base()[index];
171 }
172
173 const jbyte* bytes() const { return base(); }
174
175 int utf8_length() const { return _length; }
176
177 // Compares the symbol with a string.
178 bool equals(const char* str, int len) const;
179 bool equals(const char* str) const { return equals(str, (int) strlen(str)); }
180
181 // Tests if the symbol starts with the given prefix.
182 bool starts_with(const char* prefix, int len) const;
183 bool starts_with(const char* prefix) const {
184 return starts_with(prefix, (int) strlen(prefix));
185 }
186
187 // Tests if the symbol ends with the given suffix.
188 bool ends_with(const char* suffix, int len) const;
189 bool ends_with(const char* suffix) const {
190 return ends_with(suffix, (int) strlen(suffix));
191 }
192
193 // Returns index of the given substring in the symbol, starting
194 // at index 'i'.
195 // If the symbol doesn't contain the given substring, return -1.
196 int index_of_at(int i, const char* str, int len) const;
197 int index_of_at(int i, const char* str) const {
198 return index_of_at(i, str, (int) strlen(str));
199 }
200
201 // Three-way compare for sorting; returns -1/0/1 if receiver is </==/> than arg
202 // note that the ordering is not alfabetical
203 inline int fast_compare(const Symbol* other) const;
204
205 // Returns receiver converted to null-terminated UTF-8 string; string is
206 // allocated in resource area, or in the char buffer provided by caller.
207 char* as_C_string() const;
208 char* as_C_string(char* buf, int size) const;
209 // Use buf if needed buffer length is <= size.
210 char* as_C_string_flexible_buffer(Thread* t, char* buf, int size) const;
211
212 // Returns an escaped form of a Java string.
213 char* as_quoted_ascii() const;
214
215 // Returns a null terminated utf8 string in a resource array
216 char* as_utf8() const { return as_C_string(); }
217 char* as_utf8_flexible_buffer(Thread* t, char* buf, int size) const {
218 return as_C_string_flexible_buffer(t, buf, size);
219 }
220
221 jchar* as_unicode(int& length) const;
222
223 // Treating this symbol as a class name, returns the Java name for the class.
224 // String is allocated in resource area if buffer is not provided.
225 // See Klass::external_name()
226 const char* as_klass_external_name() const;
227 const char* as_klass_external_name(char* buf, int size) const;
228
229 // Printing
230 void print_symbol_on(outputStream* st = NULL) const;
231 void print_utf8_on(outputStream* st) const;
232 void print_on(outputStream* st) const; // First level print
233 void print_value_on(outputStream* st) const; // Second level print.
234
235 // printing on default output stream
236 void print() { print_on(tty); }
237 void print_value() { print_value_on(tty); }
238
239 #ifndef PRODUCT
240 // Empty constructor to create a dummy symbol object on stack
241 // only for getting its vtable pointer.
242 Symbol() { }
243
244 static int _total_count;
245 #endif
246 };
247
248 // Note: this comparison is used for vtable sorting only; it doesn't matter
249 // what order it defines, as long as it is a total, time-invariant order
250 // Since Symbol*s are in C_HEAP, their relative order in memory never changes,
251 // so use address comparison for speed
252 int Symbol::fast_compare(const Symbol* other) const {
253 return (((uintptr_t)this < (uintptr_t)other) ? -1
254 : ((uintptr_t)this == (uintptr_t) other) ? 0 : 1);
255 }
256 #endif // SHARE_VM_OOPS_SYMBOL_HPP