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 #include "precompiled.hpp"
26 #include "classfile/altHashing.hpp"
27 #include "classfile/javaClasses.hpp"
28 #include "classfile/stringTable.hpp"
29 #include "classfile/systemDictionary.hpp"
30 #include "gc_interface/collectedHeap.inline.hpp"
31 #include "memory/allocation.inline.hpp"
32 #include "memory/filemap.hpp"
33 #include "memory/gcLocker.inline.hpp"
34 #include "oops/oop.inline.hpp"
35 #include "oops/oop.inline2.hpp"
36 #include "runtime/atomic.inline.hpp"
37 #include "runtime/mutexLocker.hpp"
38 #include "utilities/hashtable.inline.hpp"
39 #if INCLUDE_ALL_GCS
40 #include "gc_implementation/g1/g1StringDedup.hpp"
41 #endif
42
43 PRAGMA_FORMAT_MUTE_WARNINGS_FOR_GCC
44
45 // the number of buckets a thread claims
46 const int ClaimChunkSize = 32;
47
48 #ifdef ASSERT
49 class StableMemoryChecker : public StackObj {
50 enum { _bufsize = wordSize*4 };
51
52 address _region;
53 jint _size;
54 u1 _save_buf[_bufsize];
55
56 int sample(u1* save_buf) {
57 if (_size <= _bufsize) {
58 memcpy(save_buf, _region, _size);
59 return _size;
60 } else {
61 // copy head and tail
62 memcpy(&save_buf[0], _region, _bufsize/2);
63 memcpy(&save_buf[_bufsize/2], _region + _size - _bufsize/2, _bufsize/2);
64 return (_bufsize/2)*2;
65 }
66 }
67
68 public:
69 StableMemoryChecker(const void* region, jint size) {
70 _region = (address) region;
71 _size = size;
72 sample(_save_buf);
73 }
74
75 bool verify() {
76 u1 check_buf[sizeof(_save_buf)];
77 int check_size = sample(check_buf);
78 return (0 == memcmp(_save_buf, check_buf, check_size));
79 }
80
81 void set_region(const void* region) { _region = (address) region; }
82 };
83 #endif
84
85
86 // --------------------------------------------------------------------------
87 StringTable* StringTable::_the_table = NULL;
88
89 bool StringTable::_needs_rehashing = false;
90
91 volatile int StringTable::_parallel_claimed_idx = 0;
92
93 // Pick hashing algorithm
94 unsigned int StringTable::hash_string(const jchar* s, int len) {
95 return use_alternate_hashcode() ? AltHashing::murmur3_32(seed(), s, len) :
96 java_lang_String::hash_code(s, len);
97 }
98
99 oop StringTable::lookup(int index, jchar* name,
100 int len, unsigned int hash) {
101 int count = 0;
102 for (HashtableEntry<oop, mtSymbol>* l = bucket(index); l != NULL; l = l->next()) {
103 count++;
104 if (l->hash() == hash) {
105 if (java_lang_String::equals(l->literal(), name, len)) {
106 return l->literal();
107 }
108 }
109 }
110 // If the bucket size is too deep check if this hash code is insufficient.
111 if (count >= BasicHashtable<mtSymbol>::rehash_count && !needs_rehashing()) {
112 _needs_rehashing = check_rehash_table(count);
113 }
114 return NULL;
115 }
116
117
118 oop StringTable::basic_add(int index_arg, Handle string, jchar* name,
119 int len, unsigned int hashValue_arg, TRAPS) {
120
121 assert(java_lang_String::equals(string(), name, len),
122 "string must be properly initialized");
123 // Cannot hit a safepoint in this function because the "this" pointer can move.
124 No_Safepoint_Verifier nsv;
125
126 // Check if the symbol table has been rehashed, if so, need to recalculate
127 // the hash value and index before second lookup.
128 unsigned int hashValue;
129 int index;
130 if (use_alternate_hashcode()) {
131 hashValue = hash_string(name, len);
132 index = hash_to_index(hashValue);
133 } else {
134 hashValue = hashValue_arg;
135 index = index_arg;
136 }
137
138 // Since look-up was done lock-free, we need to check if another
139 // thread beat us in the race to insert the symbol.
140
141 oop test = lookup(index, name, len, hashValue); // calls lookup(u1*, int)
142 if (test != NULL) {
143 // Entry already added
144 return test;
145 }
146
147 HashtableEntry<oop, mtSymbol>* entry = new_entry(hashValue, string());
148 add_entry(index, entry);
149 return string();
150 }
151
152
153 oop StringTable::lookup(Symbol* symbol) {
154 ResourceMark rm;
155 int length;
156 jchar* chars = symbol->as_unicode(length);
157 return lookup(chars, length);
158 }
159
160
161 oop StringTable::lookup(jchar* name, int len) {
162 unsigned int hash = hash_string(name, len);
163 int index = the_table()->hash_to_index(hash);
164 return the_table()->lookup(index, name, len, hash);
165 }
166
167
168 oop StringTable::intern(Handle string_or_null, jchar* name,
169 int len, TRAPS) {
170 unsigned int hashValue = hash_string(name, len);
171 int index = the_table()->hash_to_index(hashValue);
172 oop found_string = the_table()->lookup(index, name, len, hashValue);
173
174 // Found
175 if (found_string != NULL) return found_string;
176
177 debug_only(StableMemoryChecker smc(name, len * sizeof(name[0])));
178 assert(!Universe::heap()->is_in_reserved(name),
179 "proposed name of symbol must be stable");
180
181 Handle string;
182 // try to reuse the string if possible
183 if (!string_or_null.is_null()) {
184 string = string_or_null;
185 } else {
186 string = java_lang_String::create_from_unicode(name, len, CHECK_NULL);
187 }
188
189 #if INCLUDE_ALL_GCS
190 if (G1StringDedup::is_enabled()) {
191 // Deduplicate the string before it is interned. Note that we should never
192 // deduplicate a string after it has been interned. Doing so will counteract
193 // compiler optimizations done on e.g. interned string literals.
194 G1StringDedup::deduplicate(string());
195 }
196 #endif
197
198 // Grab the StringTable_lock before getting the_table() because it could
199 // change at safepoint.
200 MutexLocker ml(StringTable_lock, THREAD);
201
202 // Otherwise, add to symbol to table
203 return the_table()->basic_add(index, string, name, len,
204 hashValue, CHECK_NULL);
205 }
206
207 oop StringTable::intern(Symbol* symbol, TRAPS) {
208 if (symbol == NULL) return NULL;
209 ResourceMark rm(THREAD);
210 int length;
211 jchar* chars = symbol->as_unicode(length);
212 Handle string;
213 oop result = intern(string, chars, length, CHECK_NULL);
214 return result;
215 }
216
217
218 oop StringTable::intern(oop string, TRAPS)
219 {
220 if (string == NULL) return NULL;
221 ResourceMark rm(THREAD);
222 int length;
223 Handle h_string (THREAD, string);
224 jchar* chars = java_lang_String::as_unicode_string(string, length, CHECK_NULL);
225 oop result = intern(h_string, chars, length, CHECK_NULL);
226 return result;
227 }
228
229
230 oop StringTable::intern(const char* utf8_string, TRAPS) {
231 if (utf8_string == NULL) return NULL;
232 ResourceMark rm(THREAD);
233 int length = UTF8::unicode_length(utf8_string);
234 jchar* chars = NEW_RESOURCE_ARRAY(jchar, length);
235 UTF8::convert_to_unicode(utf8_string, chars, length);
236 Handle string;
237 oop result = intern(string, chars, length, CHECK_NULL);
238 return result;
239 }
240
241 void StringTable::unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int* processed, int* removed) {
242 buckets_unlink_or_oops_do(is_alive, f, 0, the_table()->table_size(), processed, removed);
243 }
244
245 void StringTable::possibly_parallel_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int* processed, int* removed) {
246 // Readers of the table are unlocked, so we should only be removing
247 // entries at a safepoint.
248 assert(SafepointSynchronize::is_at_safepoint(), "must be at safepoint");
249 const int limit = the_table()->table_size();
250
251 for (;;) {
252 // Grab next set of buckets to scan
253 int start_idx = Atomic::add(ClaimChunkSize, &_parallel_claimed_idx) - ClaimChunkSize;
254 if (start_idx >= limit) {
255 // End of table
256 break;
257 }
258
259 int end_idx = MIN2(limit, start_idx + ClaimChunkSize);
260 buckets_unlink_or_oops_do(is_alive, f, start_idx, end_idx, processed, removed);
261 }
262 }
263
264 void StringTable::buckets_oops_do(OopClosure* f, int start_idx, int end_idx) {
265 const int limit = the_table()->table_size();
266
267 assert(0 <= start_idx && start_idx <= limit,
268 err_msg("start_idx (%d) is out of bounds", start_idx));
269 assert(0 <= end_idx && end_idx <= limit,
270 err_msg("end_idx (%d) is out of bounds", end_idx));
271 assert(start_idx <= end_idx,
272 err_msg("Index ordering: start_idx=%d, end_idx=%d",
273 start_idx, end_idx));
274
275 for (int i = start_idx; i < end_idx; i += 1) {
276 HashtableEntry<oop, mtSymbol>* entry = the_table()->bucket(i);
277 while (entry != NULL) {
278 assert(!entry->is_shared(), "CDS not used for the StringTable");
279
280 f->do_oop((oop*)entry->literal_addr());
281
282 entry = entry->next();
283 }
284 }
285 }
286
287 void StringTable::buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, int* processed, int* removed) {
288 const int limit = the_table()->table_size();
289
290 assert(0 <= start_idx && start_idx <= limit,
291 err_msg("start_idx (%d) is out of bounds", start_idx));
292 assert(0 <= end_idx && end_idx <= limit,
293 err_msg("end_idx (%d) is out of bounds", end_idx));
294 assert(start_idx <= end_idx,
295 err_msg("Index ordering: start_idx=%d, end_idx=%d",
296 start_idx, end_idx));
297
298 for (int i = start_idx; i < end_idx; ++i) {
299 HashtableEntry<oop, mtSymbol>** p = the_table()->bucket_addr(i);
300 HashtableEntry<oop, mtSymbol>* entry = the_table()->bucket(i);
301 while (entry != NULL) {
302 assert(!entry->is_shared(), "CDS not used for the StringTable");
303
304 if (is_alive->do_object_b(entry->literal())) {
305 if (f != NULL) {
306 f->do_oop((oop*)entry->literal_addr());
307 }
308 p = entry->next_addr();
309 } else {
310 *p = entry->next();
311 the_table()->free_entry(entry);
312 (*removed)++;
313 }
314 (*processed)++;
315 entry = *p;
316 }
317 }
318 }
319
320 void StringTable::oops_do(OopClosure* f) {
321 buckets_oops_do(f, 0, the_table()->table_size());
322 }
323
324 void StringTable::possibly_parallel_oops_do(OopClosure* f) {
325 const int limit = the_table()->table_size();
326
327 for (;;) {
328 // Grab next set of buckets to scan
329 int start_idx = Atomic::add(ClaimChunkSize, &_parallel_claimed_idx) - ClaimChunkSize;
330 if (start_idx >= limit) {
331 // End of table
332 break;
333 }
334
335 int end_idx = MIN2(limit, start_idx + ClaimChunkSize);
336 buckets_oops_do(f, start_idx, end_idx);
337 }
338 }
339
340 // This verification is part of Universe::verify() and needs to be quick.
341 // See StringTable::verify_and_compare() below for exhaustive verification.
342 void StringTable::verify() {
343 for (int i = 0; i < the_table()->table_size(); ++i) {
344 HashtableEntry<oop, mtSymbol>* p = the_table()->bucket(i);
345 for ( ; p != NULL; p = p->next()) {
346 oop s = p->literal();
347 guarantee(s != NULL, "interned string is NULL");
348 unsigned int h = java_lang_String::hash_string(s);
349 guarantee(p->hash() == h, "broken hash in string table entry");
350 guarantee(the_table()->hash_to_index(h) == i,
351 "wrong index in string table");
352 }
353 }
354 }
355
356 void StringTable::dump(outputStream* st) {
357 the_table()->dump_table(st, "StringTable");
358 }
359
360 StringTable::VerifyRetTypes StringTable::compare_entries(
361 int bkt1, int e_cnt1,
362 HashtableEntry<oop, mtSymbol>* e_ptr1,
363 int bkt2, int e_cnt2,
364 HashtableEntry<oop, mtSymbol>* e_ptr2) {
365 // These entries are sanity checked by verify_and_compare_entries()
366 // before this function is called.
367 oop str1 = e_ptr1->literal();
368 oop str2 = e_ptr2->literal();
369
370 if (str1 == str2) {
371 tty->print_cr("ERROR: identical oop values (0x" PTR_FORMAT ") "
372 "in entry @ bucket[%d][%d] and entry @ bucket[%d][%d]",
373 (void *)str1, bkt1, e_cnt1, bkt2, e_cnt2);
374 return _verify_fail_continue;
375 }
376
377 if (java_lang_String::equals(str1, str2)) {
378 tty->print_cr("ERROR: identical String values in entry @ "
379 "bucket[%d][%d] and entry @ bucket[%d][%d]",
380 bkt1, e_cnt1, bkt2, e_cnt2);
381 return _verify_fail_continue;
382 }
383
384 return _verify_pass;
385 }
386
387 StringTable::VerifyRetTypes StringTable::verify_entry(int bkt, int e_cnt,
388 HashtableEntry<oop, mtSymbol>* e_ptr,
389 StringTable::VerifyMesgModes mesg_mode) {
390
391 VerifyRetTypes ret = _verify_pass; // be optimistic
392
393 oop str = e_ptr->literal();
394 if (str == NULL) {
395 if (mesg_mode == _verify_with_mesgs) {
396 tty->print_cr("ERROR: NULL oop value in entry @ bucket[%d][%d]", bkt,
397 e_cnt);
398 }
399 // NULL oop means no more verifications are possible
400 return _verify_fail_done;
401 }
402
403 if (str->klass() != SystemDictionary::String_klass()) {
404 if (mesg_mode == _verify_with_mesgs) {
405 tty->print_cr("ERROR: oop is not a String in entry @ bucket[%d][%d]",
406 bkt, e_cnt);
407 }
408 // not a String means no more verifications are possible
409 return _verify_fail_done;
410 }
411
412 unsigned int h = java_lang_String::hash_string(str);
413 if (e_ptr->hash() != h) {
414 if (mesg_mode == _verify_with_mesgs) {
415 tty->print_cr("ERROR: broken hash value in entry @ bucket[%d][%d], "
416 "bkt_hash=%d, str_hash=%d", bkt, e_cnt, e_ptr->hash(), h);
417 }
418 ret = _verify_fail_continue;
419 }
420
421 if (the_table()->hash_to_index(h) != bkt) {
422 if (mesg_mode == _verify_with_mesgs) {
423 tty->print_cr("ERROR: wrong index value for entry @ bucket[%d][%d], "
424 "str_hash=%d, hash_to_index=%d", bkt, e_cnt, h,
425 the_table()->hash_to_index(h));
426 }
427 ret = _verify_fail_continue;
428 }
429
430 return ret;
431 }
432
433 // See StringTable::verify() above for the quick verification that is
434 // part of Universe::verify(). This verification is exhaustive and
435 // reports on every issue that is found. StringTable::verify() only
436 // reports on the first issue that is found.
437 //
438 // StringTable::verify_entry() checks:
439 // - oop value != NULL (same as verify())
440 // - oop value is a String
441 // - hash(String) == hash in entry (same as verify())
442 // - index for hash == index of entry (same as verify())
443 //
444 // StringTable::compare_entries() checks:
445 // - oops are unique across all entries
446 // - String values are unique across all entries
447 //
448 int StringTable::verify_and_compare_entries() {
449 assert(StringTable_lock->is_locked(), "sanity check");
450
451 int fail_cnt = 0;
452
453 // first, verify all the entries individually:
454 for (int bkt = 0; bkt < the_table()->table_size(); bkt++) {
455 HashtableEntry<oop, mtSymbol>* e_ptr = the_table()->bucket(bkt);
456 for (int e_cnt = 0; e_ptr != NULL; e_ptr = e_ptr->next(), e_cnt++) {
457 VerifyRetTypes ret = verify_entry(bkt, e_cnt, e_ptr, _verify_with_mesgs);
458 if (ret != _verify_pass) {
459 fail_cnt++;
460 }
461 }
462 }
463
464 // Optimization: if the above check did not find any failures, then
465 // the comparison loop below does not need to call verify_entry()
466 // before calling compare_entries(). If there were failures, then we
467 // have to call verify_entry() to see if the entry can be passed to
468 // compare_entries() safely. When we call verify_entry() in the loop
469 // below, we do so quietly to void duplicate messages and we don't
470 // increment fail_cnt because the failures have already been counted.
471 bool need_entry_verify = (fail_cnt != 0);
472
473 // second, verify all entries relative to each other:
474 for (int bkt1 = 0; bkt1 < the_table()->table_size(); bkt1++) {
475 HashtableEntry<oop, mtSymbol>* e_ptr1 = the_table()->bucket(bkt1);
476 for (int e_cnt1 = 0; e_ptr1 != NULL; e_ptr1 = e_ptr1->next(), e_cnt1++) {
477 if (need_entry_verify) {
478 VerifyRetTypes ret = verify_entry(bkt1, e_cnt1, e_ptr1,
479 _verify_quietly);
480 if (ret == _verify_fail_done) {
481 // cannot use the current entry to compare against other entries
482 continue;
483 }
484 }
485
486 for (int bkt2 = bkt1; bkt2 < the_table()->table_size(); bkt2++) {
487 HashtableEntry<oop, mtSymbol>* e_ptr2 = the_table()->bucket(bkt2);
488 int e_cnt2;
489 for (e_cnt2 = 0; e_ptr2 != NULL; e_ptr2 = e_ptr2->next(), e_cnt2++) {
490 if (bkt1 == bkt2 && e_cnt2 <= e_cnt1) {
491 // skip the entries up to and including the one that
492 // we're comparing against
493 continue;
494 }
495
496 if (need_entry_verify) {
497 VerifyRetTypes ret = verify_entry(bkt2, e_cnt2, e_ptr2,
498 _verify_quietly);
499 if (ret == _verify_fail_done) {
500 // cannot compare against this entry
501 continue;
502 }
503 }
504
505 // compare two entries, report and count any failures:
506 if (compare_entries(bkt1, e_cnt1, e_ptr1, bkt2, e_cnt2, e_ptr2)
507 != _verify_pass) {
508 fail_cnt++;
509 }
510 }
511 }
512 }
513 }
514 return fail_cnt;
515 }
516
517 // Create a new table and using alternate hash code, populate the new table
518 // with the existing strings. Set flag to use the alternate hash code afterwards.
519 void StringTable::rehash_table() {
520 assert(SafepointSynchronize::is_at_safepoint(), "must be at safepoint");
521 // This should never happen with -Xshare:dump but it might in testing mode.
522 if (DumpSharedSpaces) return;
523 StringTable* new_table = new StringTable();
524
525 // Rehash the table
526 the_table()->move_to(new_table);
527
528 // Delete the table and buckets (entries are reused in new table).
529 delete _the_table;
530 // Don't check if we need rehashing until the table gets unbalanced again.
531 // Then rehash with a new global seed.
532 _needs_rehashing = false;
533 _the_table = new_table;
534 }