1 /*
2 * Copyright (c) 1997, 2021, 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
26 #include "precompiled.hpp"
27 #include "cds/metaspaceShared.hpp"
28 #include "classfile/altHashing.hpp"
29 #include "classfile/classLoaderData.hpp"
30 #include "classfile/vmSymbols.hpp"
31 #include "gc/shared/collectedHeap.hpp"
32 #include "logging/log.hpp"
33 #include "logging/logStream.hpp"
34 #include "memory/allocation.inline.hpp"
35 #include "memory/resourceArea.hpp"
36 #include "memory/universe.hpp"
37 #include "oops/symbol.hpp"
38 #include "runtime/atomic.hpp"
39 #include "runtime/mutexLocker.hpp"
40 #include "runtime/os.hpp"
41 #include "runtime/signature.hpp"
42 #include "utilities/utf8.hpp"
43
44 Symbol* Symbol::_vm_symbols[vmSymbols::number_of_symbols()];
45
46 uint32_t Symbol::pack_hash_and_refcount(short hash, int refcount) {
47 STATIC_ASSERT(PERM_REFCOUNT == ((1 << 16) - 1));
48 assert(refcount >= 0, "negative refcount");
49 assert(refcount <= PERM_REFCOUNT, "invalid refcount");
50 uint32_t hi = hash;
51 uint32_t lo = refcount;
52 return (hi << 16) | lo;
53 }
54
55 Symbol::Symbol(const u1* name, int length, int refcount) {
56 _hash_and_refcount = pack_hash_and_refcount((short)os::random(), refcount);
57 _length = length;
58 // _body[0..1] are allocated in the header just by coincidence in the current
59 // implementation of Symbol. They are read by identity_hash(), so make sure they
60 // are initialized.
61 // No other code should assume that _body[0..1] are always allocated. E.g., do
62 // not unconditionally read base()[0] as that will be invalid for an empty Symbol.
63 _body[0] = _body[1] = 0;
64 memcpy(_body, name, length);
65 }
66
67 void* Symbol::operator new(size_t sz, int len) throw() {
68 #if INCLUDE_CDS
69 if (DumpSharedSpaces) {
70 MutexLocker ml(DumpRegion_lock, Mutex::_no_safepoint_check_flag);
71 // To get deterministic output from -Xshare:dump, we ensure that Symbols are allocated in
72 // increasing addresses. When the symbols are copied into the archive, we preserve their
73 // relative address order (sorted, see ArchiveBuilder::gather_klasses_and_symbols).
74 //
75 // We cannot use arena because arena chunks are allocated by the OS. As a result, for example,
76 // the archived symbol of "java/lang/Object" may sometimes be lower than "java/lang/String", and
77 // sometimes be higher. This would cause non-deterministic contents in the archive.
78 DEBUG_ONLY(static void* last = 0);
79 void* p = (void*)MetaspaceShared::symbol_space_alloc(size(len)*wordSize);
80 assert(p > last, "must increase monotonically");
81 DEBUG_ONLY(last = p);
82 return p;
83 }
84 #endif
85 int alloc_size = size(len)*wordSize;
86 address res = (address) AllocateHeap(alloc_size, mtSymbol);
87 return res;
88 }
89
90 void* Symbol::operator new(size_t sz, int len, Arena* arena) throw() {
91 int alloc_size = size(len)*wordSize;
92 address res = (address)arena->AmallocWords(alloc_size);
93 return res;
94 }
95
96 void Symbol::operator delete(void *p) {
97 assert(((Symbol*)p)->refcount() == 0, "should not call this");
98 FreeHeap(p);
99 }
100
101 #if INCLUDE_CDS
102 void Symbol::update_identity_hash() {
103 // This is called at a safepoint during dumping of a static CDS archive. The caller should have
104 // called os::init_random() with a deterministic seed and then iterate all archived Symbols in
105 // a deterministic order.
106 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
107 _hash_and_refcount = pack_hash_and_refcount((short)os::random(), PERM_REFCOUNT);
108 }
109
110 void Symbol::set_permanent() {
111 // This is called at a safepoint during dumping of a dynamic CDS archive.
112 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
113 _hash_and_refcount = pack_hash_and_refcount(extract_hash(_hash_and_refcount), PERM_REFCOUNT);
114 }
115 #endif
116
117 bool Symbol::is_Q_signature() const {
118 int len = utf8_length();
119 return len > 2 && char_at(0) == JVM_SIGNATURE_INLINE_TYPE && char_at(len - 1) == JVM_SIGNATURE_ENDCLASS;
120 }
121
122 bool Symbol::is_Q_array_signature() const {
123 int l = utf8_length();
124 if (l < 2 || char_at(0) != JVM_SIGNATURE_ARRAY || char_at(l - 1) != JVM_SIGNATURE_ENDCLASS) {
125 return false;
126 }
127 for (int i = 1; i < (l - 2); i++) {
128 char c = char_at(i);
129 if (c == JVM_SIGNATURE_INLINE_TYPE) {
130 return true;
131 }
132 if (c != JVM_SIGNATURE_ARRAY) {
133 return false;
134 }
135 }
136 return false;
137 }
138
139 bool Symbol::is_Q_method_signature() const {
140 assert(SignatureVerifier::is_valid_method_signature(this), "must be");
141 int len = utf8_length();
142 if (len > 4 && char_at(0) == JVM_SIGNATURE_FUNC) {
143 for (int i=1; i<len-3; i++) { // Must end with ")Qx;", where x is at least one character or more.
144 if (char_at(i) == JVM_SIGNATURE_ENDFUNC && char_at(i+1) == JVM_SIGNATURE_INLINE_TYPE) {
145 return true;
146 }
147 }
148 }
149 return false;
150 }
151
152 Symbol* Symbol::fundamental_name(TRAPS) {
153 if ((char_at(0) == JVM_SIGNATURE_INLINE_TYPE || char_at(0) == JVM_SIGNATURE_CLASS) && ends_with(JVM_SIGNATURE_ENDCLASS)) {
154 return SymbolTable::new_symbol(this, 1, utf8_length() - 1);
155 } else {
156 // reference count is incremented to be consistent with the behavior with
157 // the SymbolTable::new_symbol() call above
158 this->increment_refcount();
159 return this;
160 }
161 }
162
163 bool Symbol::is_same_fundamental_type(Symbol* s) const {
164 if (this == s) return true;
165 if (utf8_length() < 3) return false;
166 int offset1, offset2, len;
167 if (ends_with(JVM_SIGNATURE_ENDCLASS)) {
168 if (char_at(0) != JVM_SIGNATURE_INLINE_TYPE && char_at(0) != JVM_SIGNATURE_CLASS) return false;
169 offset1 = 1;
170 len = utf8_length() - 2;
171 } else {
172 offset1 = 0;
173 len = utf8_length();
174 }
175 if (ends_with(JVM_SIGNATURE_ENDCLASS)) {
176 if (s->char_at(0) != JVM_SIGNATURE_INLINE_TYPE && s->char_at(0) != JVM_SIGNATURE_CLASS) return false;
177 offset2 = 1;
178 } else {
179 offset2 = 0;
180 }
181 if ((offset2 + len) > s->utf8_length()) return false;
182 if ((utf8_length() - offset1 * 2) != (s->utf8_length() - offset2 * 2))
183 return false;
184 int l = len;
185 while (l-- > 0) {
186 if (char_at(offset1 + l) != s->char_at(offset2 + l))
187 return false;
188 }
189 return true;
190 }
191
192 // ------------------------------------------------------------------
193 // Symbol::index_of
194 //
195 // Finds if the given string is a substring of this symbol's utf8 bytes.
196 // Return -1 on failure. Otherwise return the first index where str occurs.
197 int Symbol::index_of_at(int i, const char* str, int len) const {
198 assert(i >= 0 && i <= utf8_length(), "oob");
199 if (len <= 0) return 0;
200 char first_char = str[0];
201 address bytes = (address) ((Symbol*)this)->base();
202 address limit = bytes + utf8_length() - len; // inclusive limit
203 address scan = bytes + i;
204 if (scan > limit)
205 return -1;
206 for (; scan <= limit; scan++) {
207 scan = (address) memchr(scan, first_char, (limit + 1 - scan));
208 if (scan == NULL)
209 return -1; // not found
210 assert(scan >= bytes+i && scan <= limit, "scan oob");
211 if (len <= 2
212 ? (char) scan[len-1] == str[len-1]
213 : memcmp(scan+1, str+1, len-1) == 0) {
214 return (int)(scan - bytes);
215 }
216 }
217 return -1;
218 }
219
220
221 char* Symbol::as_C_string(char* buf, int size) const {
222 if (size > 0) {
223 int len = MIN2(size - 1, utf8_length());
224 for (int i = 0; i < len; i++) {
225 buf[i] = char_at(i);
226 }
227 buf[len] = '\0';
228 }
229 return buf;
230 }
231
232 char* Symbol::as_C_string() const {
233 int len = utf8_length();
234 char* str = NEW_RESOURCE_ARRAY(char, len + 1);
235 return as_C_string(str, len + 1);
236 }
237
238 void Symbol::print_utf8_on(outputStream* st) const {
239 st->print("%s", as_C_string());
240 }
241
242 void Symbol::print_symbol_on(outputStream* st) const {
243 char *s;
244 st = st ? st : tty;
245 {
246 // ResourceMark may not affect st->print(). If st is a string
247 // stream it could resize, using the same resource arena.
248 ResourceMark rm;
249 s = as_quoted_ascii();
250 s = os::strdup(s);
251 }
252 if (s == NULL) {
253 st->print("(null)");
254 } else {
255 st->print("%s", s);
256 os::free(s);
257 }
258 }
259
260 char* Symbol::as_quoted_ascii() const {
261 const char *ptr = (const char *)&_body[0];
262 int quoted_length = UTF8::quoted_ascii_length(ptr, utf8_length());
263 char* result = NEW_RESOURCE_ARRAY(char, quoted_length + 1);
264 UTF8::as_quoted_ascii(ptr, utf8_length(), result, quoted_length + 1);
265 return result;
266 }
267
268 jchar* Symbol::as_unicode(int& length) const {
269 Symbol* this_ptr = (Symbol*)this;
270 length = UTF8::unicode_length((char*)this_ptr->bytes(), utf8_length());
271 jchar* result = NEW_RESOURCE_ARRAY(jchar, length);
272 if (length > 0) {
273 UTF8::convert_to_unicode((char*)this_ptr->bytes(), result, length);
274 }
275 return result;
276 }
277
278 const char* Symbol::as_klass_external_name(char* buf, int size) const {
279 if (size > 0) {
280 char* str = as_C_string(buf, size);
281 int length = (int)strlen(str);
282 // Turn all '/'s into '.'s (also for array klasses)
283 for (int index = 0; index < length; index++) {
284 if (str[index] == JVM_SIGNATURE_SLASH) {
285 str[index] = JVM_SIGNATURE_DOT;
286 }
287 }
288 return str;
289 } else {
290 return buf;
291 }
292 }
293
294 const char* Symbol::as_klass_external_name() const {
295 char* str = as_C_string();
296 int length = (int)strlen(str);
297 // Turn all '/'s into '.'s (also for array klasses)
298 for (int index = 0; index < length; index++) {
299 if (str[index] == JVM_SIGNATURE_SLASH) {
300 str[index] = JVM_SIGNATURE_DOT;
301 }
302 }
303 return str;
304 }
305
306 static void print_class(outputStream *os, const SignatureStream& ss) {
307 int sb = ss.raw_symbol_begin(), se = ss.raw_symbol_end();
308 for (int i = sb; i < se; ++i) {
309 int ch = ss.raw_char_at(i);
310 if (ch == JVM_SIGNATURE_SLASH) {
311 os->put(JVM_SIGNATURE_DOT);
312 } else {
313 os->put(ch);
314 }
315 }
316 }
317
318 static void print_array(outputStream *os, SignatureStream& ss) {
319 int dimensions = ss.skip_array_prefix();
320 assert(dimensions > 0, "");
321 if (ss.is_reference()) {
322 print_class(os, ss);
323 } else {
324 os->print("%s", type2name(ss.type()));
325 }
326 for (int i = 0; i < dimensions; ++i) {
327 os->print("[]");
328 }
329 }
330
331 void Symbol::print_as_signature_external_return_type(outputStream *os) {
332 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
333 if (ss.at_return_type()) {
334 if (ss.is_array()) {
335 print_array(os, ss);
336 } else if (ss.is_reference()) {
337 print_class(os, ss);
338 } else {
339 os->print("%s", type2name(ss.type()));
340 }
341 }
342 }
343 }
344
345 void Symbol::print_as_signature_external_parameters(outputStream *os) {
346 bool first = true;
347 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
348 if (ss.at_return_type()) break;
349 if (!first) { os->print(", "); }
350 if (ss.is_array()) {
351 print_array(os, ss);
352 } else if (ss.is_reference()) {
353 print_class(os, ss);
354 } else {
355 os->print("%s", type2name(ss.type()));
356 }
357 first = false;
358 }
359 }
360
361 // Increment refcount while checking for zero. If the Symbol's refcount becomes zero
362 // a thread could be concurrently removing the Symbol. This is used during SymbolTable
363 // lookup to avoid reviving a dead Symbol.
364 bool Symbol::try_increment_refcount() {
365 uint32_t found = _hash_and_refcount;
366 while (true) {
367 uint32_t old_value = found;
368 int refc = extract_refcount(old_value);
369 if (refc == PERM_REFCOUNT) {
370 return true; // sticky max or created permanent
371 } else if (refc == 0) {
372 return false; // dead, can't revive.
373 } else {
374 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value + 1);
375 if (found == old_value) {
376 return true; // successfully updated.
377 }
378 // refcount changed, try again.
379 }
380 }
381 }
382
383 // The increment_refcount() is called when not doing lookup. It is assumed that you
384 // have a symbol with a non-zero refcount and it can't become zero while referenced by
385 // this caller.
386 void Symbol::increment_refcount() {
387 if (!try_increment_refcount()) {
388 #ifdef ASSERT
389 print();
390 fatal("refcount has gone to zero");
391 #endif
392 }
393 #ifndef PRODUCT
394 if (refcount() != PERM_REFCOUNT) { // not a permanent symbol
395 NOT_PRODUCT(Atomic::inc(&_total_count);)
396 }
397 #endif
398 }
399
400 // Decrement refcount potentially while racing increment, so we need
401 // to check the value after attempting to decrement so that if another
402 // thread increments to PERM_REFCOUNT the value is not decremented.
403 void Symbol::decrement_refcount() {
404 uint32_t found = _hash_and_refcount;
405 while (true) {
406 uint32_t old_value = found;
407 int refc = extract_refcount(old_value);
408 if (refc == PERM_REFCOUNT) {
409 return; // refcount is permanent, permanent is sticky
410 } else if (refc == 0) {
411 #ifdef ASSERT
412 print();
413 fatal("refcount underflow");
414 #endif
415 return;
416 } else {
417 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value - 1);
418 if (found == old_value) {
419 return; // successfully updated.
420 }
421 // refcount changed, try again.
422 }
423 }
424 }
425
426 void Symbol::make_permanent() {
427 uint32_t found = _hash_and_refcount;
428 while (true) {
429 uint32_t old_value = found;
430 int refc = extract_refcount(old_value);
431 if (refc == PERM_REFCOUNT) {
432 return; // refcount is permanent, permanent is sticky
433 } else if (refc == 0) {
434 #ifdef ASSERT
435 print();
436 fatal("refcount underflow");
437 #endif
438 return;
439 } else {
440 int hash = extract_hash(old_value);
441 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, pack_hash_and_refcount(hash, PERM_REFCOUNT));
442 if (found == old_value) {
443 return; // successfully updated.
444 }
445 // refcount changed, try again.
446 }
447 }
448 }
449
450 void Symbol::metaspace_pointers_do(MetaspaceClosure* it) {
451 if (log_is_enabled(Trace, cds)) {
452 LogStream trace_stream(Log(cds)::trace());
453 trace_stream.print("Iter(Symbol): %p ", this);
454 print_value_on(&trace_stream);
455 trace_stream.cr();
456 }
457 }
458
459 void Symbol::print_on(outputStream* st) const {
460 st->print("Symbol: '");
461 print_symbol_on(st);
462 st->print("'");
463 st->print(" count %d", refcount());
464 }
465
466 void Symbol::print() const { print_on(tty); }
467
468 // The print_value functions are present in all builds, to support the
469 // disassembler and error reporting.
470 void Symbol::print_value_on(outputStream* st) const {
471 st->print("'");
472 for (int i = 0; i < utf8_length(); i++) {
473 st->print("%c", char_at(i));
474 }
475 st->print("'");
476 }
477
478 void Symbol::print_value() const { print_value_on(tty); }
479
480 bool Symbol::is_valid(Symbol* s) {
481 if (!is_aligned(s, sizeof(MetaWord))) return false;
482 if ((size_t)s < os::min_page_size()) return false;
483
484 if (!os::is_readable_range(s, s + 1)) return false;
485
486 // Symbols are not allocated in Java heap.
487 if (Universe::heap()->is_in(s)) return false;
488
489 int len = s->utf8_length();
490 if (len < 0) return false;
491
492 jbyte* bytes = (jbyte*) s->bytes();
493 return os::is_readable_range(bytes, bytes + len);
494 }
495
496 void Symbol::print_Qvalue_on(outputStream* st) const {
497 if (this == NULL) {
498 st->print("NULL");
499 } else {
500 st->print("'Q");
501 for (int i = 0; i < utf8_length(); i++) {
502 st->print("%c", char_at(i));
503 }
504 st->print(";'");
505 }
506 }
507
508 // SymbolTable prints this in its statistics
509 NOT_PRODUCT(size_t Symbol::_total_count = 0;)
510
511 #ifndef PRODUCT
512 bool Symbol::is_valid_id(vmSymbolID vm_symbol_id) {
513 return vmSymbols::is_valid_id(vm_symbol_id);
514 }
515 #endif