1 /*
2 * Copyright (c) 1997, 2010, 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 #include "precompiled.hpp"
26 #include "classfile/javaClasses.hpp"
27 #include "classfile/symbolTable.hpp"
28 #include "classfile/systemDictionary.hpp"
29 #include "gc_interface/collectedHeap.inline.hpp"
30 #include "memory/filemap.hpp"
31 #include "memory/gcLocker.inline.hpp"
32 #include "oops/oop.inline.hpp"
33 #include "oops/oop.inline2.hpp"
34 #include "oops/symbolKlass.hpp"
35 #include "runtime/mutexLocker.hpp"
36 #include "utilities/hashtable.inline.hpp"
37
38 // --------------------------------------------------------------------------
39
40 SymbolTable* SymbolTable::_the_table = NULL;
41
42 // Lookup a symbol in a bucket.
43
44 symbolOop SymbolTable::lookup(int index, const char* name,
45 int len, unsigned int hash) {
46 for (HashtableEntry* e = bucket(index); e != NULL; e = e->next()) {
47 if (e->hash() == hash) {
48 symbolOop sym = symbolOop(e->literal());
49 if (sym->equals(name, len)) {
50 return sym;
51 }
52 }
53 }
54 return NULL;
55 }
56
57
58 // We take care not to be blocking while holding the
59 // SymbolTable_lock. Otherwise, the system might deadlock, since the
60 // symboltable is used during compilation (VM_thread) The lock free
61 // synchronization is simplified by the fact that we do not delete
62 // entries in the symbol table during normal execution (only during
63 // safepoints).
64
65 symbolOop SymbolTable::lookup(const char* name, int len, TRAPS) {
66 unsigned int hashValue = hash_symbol(name, len);
67 int index = the_table()->hash_to_index(hashValue);
68
69 symbolOop s = the_table()->lookup(index, name, len, hashValue);
70
71 // Found
72 if (s != NULL) return s;
73
74 // Otherwise, add to symbol to table
75 return the_table()->basic_add(index, (u1*)name, len, hashValue, CHECK_NULL);
76 }
77
78 symbolOop SymbolTable::lookup(symbolHandle sym, int begin, int end, TRAPS) {
79 char* buffer;
80 int index, len;
81 unsigned int hashValue;
82 char* name;
83 {
84 debug_only(No_Safepoint_Verifier nsv;)
85
86 name = (char*)sym->base() + begin;
87 len = end - begin;
88 hashValue = hash_symbol(name, len);
89 index = the_table()->hash_to_index(hashValue);
90 symbolOop s = the_table()->lookup(index, name, len, hashValue);
91
92 // Found
93 if (s != NULL) return s;
94 }
95
96 // Otherwise, add to symbol to table. Copy to a C string first.
97 char stack_buf[128];
98 ResourceMark rm(THREAD);
99 if (len <= 128) {
100 buffer = stack_buf;
101 } else {
102 buffer = NEW_RESOURCE_ARRAY_IN_THREAD(THREAD, char, len);
103 }
104 for (int i=0; i<len; i++) {
105 buffer[i] = name[i];
106 }
107 // Make sure there is no safepoint in the code above since name can't move.
108 // We can't include the code in No_Safepoint_Verifier because of the
109 // ResourceMark.
110
111 return the_table()->basic_add(index, (u1*)buffer, len, hashValue, CHECK_NULL);
112 }
113
114 symbolOop SymbolTable::lookup_only(const char* name, int len,
115 unsigned int& hash) {
116 hash = hash_symbol(name, len);
117 int index = the_table()->hash_to_index(hash);
118
119 return the_table()->lookup(index, name, len, hash);
120 }
121
122 // Suggestion: Push unicode-based lookup all the way into the hashing
123 // and probing logic, so there is no need for convert_to_utf8 until
124 // an actual new symbolOop is created.
125 symbolOop SymbolTable::lookup_unicode(const jchar* name, int utf16_length, TRAPS) {
126 int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
127 char stack_buf[128];
128 if (utf8_length < (int) sizeof(stack_buf)) {
129 char* chars = stack_buf;
130 UNICODE::convert_to_utf8(name, utf16_length, chars);
131 return lookup(chars, utf8_length, THREAD);
132 } else {
133 ResourceMark rm(THREAD);
134 char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
135 UNICODE::convert_to_utf8(name, utf16_length, chars);
136 return lookup(chars, utf8_length, THREAD);
137 }
138 }
139
140 symbolOop SymbolTable::lookup_only_unicode(const jchar* name, int utf16_length,
141 unsigned int& hash) {
142 int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
143 char stack_buf[128];
144 if (utf8_length < (int) sizeof(stack_buf)) {
145 char* chars = stack_buf;
146 UNICODE::convert_to_utf8(name, utf16_length, chars);
147 return lookup_only(chars, utf8_length, hash);
148 } else {
149 ResourceMark rm;
150 char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
151 UNICODE::convert_to_utf8(name, utf16_length, chars);
152 return lookup_only(chars, utf8_length, hash);
153 }
154 }
155
156 void SymbolTable::add(constantPoolHandle cp, int names_count,
157 const char** names, int* lengths, int* cp_indices,
158 unsigned int* hashValues, TRAPS) {
159 SymbolTable* table = the_table();
160 bool added = table->basic_add(cp, names_count, names, lengths,
161 cp_indices, hashValues, CHECK);
162 if (!added) {
163 // do it the hard way
164 for (int i=0; i<names_count; i++) {
165 int index = table->hash_to_index(hashValues[i]);
166 symbolOop sym = table->basic_add(index, (u1*)names[i], lengths[i],
167 hashValues[i], CHECK);
168 cp->symbol_at_put(cp_indices[i], sym);
169 }
170 }
171 }
172
173 symbolOop SymbolTable::basic_add(int index, u1 *name, int len,
174 unsigned int hashValue, TRAPS) {
175 assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(),
176 "proposed name of symbol must be stable");
177
178 // We assume that lookup() has been called already, that it failed,
179 // and symbol was not found. We create the symbol here.
180 symbolKlass* sk = (symbolKlass*) Universe::symbolKlassObj()->klass_part();
181 symbolOop s_oop = sk->allocate_symbol(name, len, CHECK_NULL);
182 symbolHandle sym (THREAD, s_oop);
183
184 // Allocation must be done before grapping the SymbolTable_lock lock
185 MutexLocker ml(SymbolTable_lock, THREAD);
186
187 assert(sym->equals((char*)name, len), "symbol must be properly initialized");
188
189 // Since look-up was done lock-free, we need to check if another
190 // thread beat us in the race to insert the symbol.
191
192 symbolOop test = lookup(index, (char*)name, len, hashValue);
193 if (test != NULL) {
194 // A race occurred and another thread introduced the symbol, this one
195 // will be dropped and collected.
196 return test;
197 }
198
199 HashtableEntry* entry = new_entry(hashValue, sym());
200 add_entry(index, entry);
201 return sym();
202 }
203
204 bool SymbolTable::basic_add(constantPoolHandle cp, int names_count,
205 const char** names, int* lengths,
206 int* cp_indices, unsigned int* hashValues,
207 TRAPS) {
208 symbolKlass* sk = (symbolKlass*) Universe::symbolKlassObj()->klass_part();
209 symbolOop sym_oops[symbol_alloc_batch_size];
210 bool allocated = sk->allocate_symbols(names_count, names, lengths,
211 sym_oops, CHECK_false);
212 if (!allocated) {
213 return false;
214 }
215 symbolHandle syms[symbol_alloc_batch_size];
216 int i;
217 for (i=0; i<names_count; i++) {
218 syms[i] = symbolHandle(THREAD, sym_oops[i]);
219 }
220
221 // Allocation must be done before grabbing the SymbolTable_lock lock
222 MutexLocker ml(SymbolTable_lock, THREAD);
223
224 for (i=0; i<names_count; i++) {
225 assert(syms[i]->equals(names[i], lengths[i]), "symbol must be properly initialized");
226 // Since look-up was done lock-free, we need to check if another
227 // thread beat us in the race to insert the symbol.
228 int index = hash_to_index(hashValues[i]);
229 symbolOop test = lookup(index, names[i], lengths[i], hashValues[i]);
230 if (test != NULL) {
231 // A race occurred and another thread introduced the symbol, this one
232 // will be dropped and collected. Use test instead.
233 cp->symbol_at_put(cp_indices[i], test);
234 } else {
235 symbolOop sym = syms[i]();
236 HashtableEntry* entry = new_entry(hashValues[i], sym);
237 add_entry(index, entry);
238 cp->symbol_at_put(cp_indices[i], sym);
239 }
240 }
241
242 return true;
243 }
244
245
246 void SymbolTable::verify() {
247 for (int i = 0; i < the_table()->table_size(); ++i) {
248 HashtableEntry* p = the_table()->bucket(i);
249 for ( ; p != NULL; p = p->next()) {
250 symbolOop s = symbolOop(p->literal());
251 guarantee(s != NULL, "symbol is NULL");
252 s->verify();
253 guarantee(s->is_perm(), "symbol not in permspace");
254 unsigned int h = hash_symbol((char*)s->bytes(), s->utf8_length());
255 guarantee(p->hash() == h, "broken hash in symbol table entry");
256 guarantee(the_table()->hash_to_index(h) == i,
257 "wrong index in symbol table");
258 }
259 }
260 }
261
262
263 //---------------------------------------------------------------------------
264 // Non-product code
265
266 #ifndef PRODUCT
267
268 void SymbolTable::print_histogram() {
269 MutexLocker ml(SymbolTable_lock);
270 const int results_length = 100;
271 int results[results_length];
272 int i,j;
273
274 // initialize results to zero
275 for (j = 0; j < results_length; j++) {
276 results[j] = 0;
277 }
278
279 int total = 0;
280 int max_symbols = 0;
281 int out_of_range = 0;
282 for (i = 0; i < the_table()->table_size(); i++) {
283 HashtableEntry* p = the_table()->bucket(i);
284 for ( ; p != NULL; p = p->next()) {
285 int counter = symbolOop(p->literal())->utf8_length();
286 total += counter;
287 if (counter < results_length) {
288 results[counter]++;
289 } else {
290 out_of_range++;
291 }
292 max_symbols = MAX2(max_symbols, counter);
293 }
294 }
295 tty->print_cr("Symbol Table:");
296 tty->print_cr("%8s %5d", "Total ", total);
297 tty->print_cr("%8s %5d", "Maximum", max_symbols);
298 tty->print_cr("%8s %3.2f", "Average",
299 ((float) total / (float) the_table()->table_size()));
300 tty->print_cr("%s", "Histogram:");
301 tty->print_cr(" %s %29s", "Length", "Number chains that length");
302 for (i = 0; i < results_length; i++) {
303 if (results[i] > 0) {
304 tty->print_cr("%6d %10d", i, results[i]);
305 }
306 }
307 int line_length = 70;
308 tty->print_cr("%s %30s", " Length", "Number chains that length");
309 for (i = 0; i < results_length; i++) {
310 if (results[i] > 0) {
311 tty->print("%4d", i);
312 for (j = 0; (j < results[i]) && (j < line_length); j++) {
313 tty->print("%1s", "*");
314 }
315 if (j == line_length) {
316 tty->print("%1s", "+");
317 }
318 tty->cr();
319 }
320 }
321 tty->print_cr(" %s %d: %d\n", "Number chains longer than",
322 results_length, out_of_range);
323 }
324
325 #endif // PRODUCT
326
327 // --------------------------------------------------------------------------
328
329 #ifdef ASSERT
330 class StableMemoryChecker : public StackObj {
331 enum { _bufsize = wordSize*4 };
332
333 address _region;
334 jint _size;
335 u1 _save_buf[_bufsize];
336
337 int sample(u1* save_buf) {
338 if (_size <= _bufsize) {
339 memcpy(save_buf, _region, _size);
340 return _size;
341 } else {
342 // copy head and tail
343 memcpy(&save_buf[0], _region, _bufsize/2);
344 memcpy(&save_buf[_bufsize/2], _region + _size - _bufsize/2, _bufsize/2);
345 return (_bufsize/2)*2;
346 }
347 }
348
349 public:
350 StableMemoryChecker(const void* region, jint size) {
351 _region = (address) region;
352 _size = size;
353 sample(_save_buf);
354 }
355
356 bool verify() {
357 u1 check_buf[sizeof(_save_buf)];
358 int check_size = sample(check_buf);
359 return (0 == memcmp(_save_buf, check_buf, check_size));
360 }
361
362 void set_region(const void* region) { _region = (address) region; }
363 };
364 #endif
365
366
367 // --------------------------------------------------------------------------
368
369
370 // Compute the hash value for a java.lang.String object which would
371 // contain the characters passed in. This hash value is used for at
372 // least two purposes.
373 //
374 // (a) As the hash value used by the StringTable for bucket selection
375 // and comparison (stored in the HashtableEntry structures). This
376 // is used in the String.intern() method.
377 //
378 // (b) As the hash value used by the String object itself, in
379 // String.hashCode(). This value is normally calculate in Java code
380 // in the String.hashCode method(), but is precomputed for String
381 // objects in the shared archive file.
382 //
383 // For this reason, THIS ALGORITHM MUST MATCH String.hashCode().
384
385 int StringTable::hash_string(jchar* s, int len) {
386 unsigned h = 0;
387 while (len-- > 0) {
388 h = 31*h + (unsigned) *s;
389 s++;
390 }
391 return h;
392 }
393
394
395 StringTable* StringTable::_the_table = NULL;
396
397 oop StringTable::lookup(int index, jchar* name,
398 int len, unsigned int hash) {
399 for (HashtableEntry* l = bucket(index); l != NULL; l = l->next()) {
400 if (l->hash() == hash) {
401 if (java_lang_String::equals(l->literal(), name, len)) {
402 return l->literal();
403 }
404 }
405 }
406 return NULL;
407 }
408
409
410 oop StringTable::basic_add(int index, Handle string_or_null, jchar* name,
411 int len, unsigned int hashValue, TRAPS) {
412 debug_only(StableMemoryChecker smc(name, len * sizeof(name[0])));
413 assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(),
414 "proposed name of symbol must be stable");
415
416 Handle string;
417 // try to reuse the string if possible
418 if (!string_or_null.is_null() && string_or_null()->is_perm()) {
419 string = string_or_null;
420 } else {
421 string = java_lang_String::create_tenured_from_unicode(name, len, CHECK_NULL);
422 }
423
424 // Allocation must be done before grapping the SymbolTable_lock lock
425 MutexLocker ml(StringTable_lock, THREAD);
426
427 assert(java_lang_String::equals(string(), name, len),
428 "string must be properly initialized");
429
430 // Since look-up was done lock-free, we need to check if another
431 // thread beat us in the race to insert the symbol.
432
433 oop test = lookup(index, name, len, hashValue); // calls lookup(u1*, int)
434 if (test != NULL) {
435 // Entry already added
436 return test;
437 }
438
439 HashtableEntry* entry = new_entry(hashValue, string());
440 add_entry(index, entry);
441 return string();
442 }
443
444
445 oop StringTable::lookup(symbolOop symbol) {
446 ResourceMark rm;
447 int length;
448 jchar* chars = symbol->as_unicode(length);
449 unsigned int hashValue = hash_string(chars, length);
450 int index = the_table()->hash_to_index(hashValue);
451 return the_table()->lookup(index, chars, length, hashValue);
452 }
453
454
455 oop StringTable::intern(Handle string_or_null, jchar* name,
456 int len, TRAPS) {
457 unsigned int hashValue = hash_string(name, len);
458 int index = the_table()->hash_to_index(hashValue);
459 oop string = the_table()->lookup(index, name, len, hashValue);
460
461 // Found
462 if (string != NULL) return string;
463
464 // Otherwise, add to symbol to table
465 return the_table()->basic_add(index, string_or_null, name, len,
466 hashValue, CHECK_NULL);
467 }
468
469 oop StringTable::intern(symbolOop symbol, TRAPS) {
470 if (symbol == NULL) return NULL;
471 ResourceMark rm(THREAD);
472 int length;
473 jchar* chars = symbol->as_unicode(length);
474 Handle string;
475 oop result = intern(string, chars, length, CHECK_NULL);
476 return result;
477 }
478
479
480 oop StringTable::intern(oop string, TRAPS)
481 {
482 if (string == NULL) return NULL;
483 ResourceMark rm(THREAD);
484 int length;
485 Handle h_string (THREAD, string);
486 jchar* chars = java_lang_String::as_unicode_string(string, length);
487 oop result = intern(h_string, chars, length, CHECK_NULL);
488 return result;
489 }
490
491
492 oop StringTable::intern(const char* utf8_string, TRAPS) {
493 if (utf8_string == NULL) return NULL;
494 ResourceMark rm(THREAD);
495 int length = UTF8::unicode_length(utf8_string);
496 jchar* chars = NEW_RESOURCE_ARRAY(jchar, length);
497 UTF8::convert_to_unicode(utf8_string, chars, length);
498 Handle string;
499 oop result = intern(string, chars, length, CHECK_NULL);
500 return result;
501 }
502
503 void StringTable::verify() {
504 for (int i = 0; i < the_table()->table_size(); ++i) {
505 HashtableEntry* p = the_table()->bucket(i);
506 for ( ; p != NULL; p = p->next()) {
507 oop s = p->literal();
508 guarantee(s != NULL, "interned string is NULL");
509 guarantee(s->is_perm(), "interned string not in permspace");
510
511 int length;
512 jchar* chars = java_lang_String::as_unicode_string(s, length);
513 unsigned int h = hash_string(chars, length);
514 guarantee(p->hash() == h, "broken hash in string table entry");
515 guarantee(the_table()->hash_to_index(h) == i,
516 "wrong index in string table");
517 }
518 }
519 }