1 /*
2 * Copyright (c) 1997, 2025, 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 "cds/archiveBuilder.hpp"
26 #include "classfile/altHashing.hpp"
27 #include "classfile/classLoaderData.hpp"
28 #include "classfile/vmSymbols.hpp"
29 #include "gc/shared/collectedHeap.hpp"
30 #include "logging/log.hpp"
31 #include "logging/logStream.hpp"
32 #include "memory/allocation.inline.hpp"
33 #include "memory/resourceArea.hpp"
34 #include "memory/universe.hpp"
35 #include "oops/symbol.hpp"
36 #include "runtime/atomicAccess.hpp"
37 #include "runtime/mutexLocker.hpp"
38 #include "runtime/os.hpp"
39 #include "runtime/signature.hpp"
40 #include "utilities/stringUtils.hpp"
41 #include "utilities/utf8.hpp"
42
43 Symbol* Symbol::_vm_symbols[vmSymbols::number_of_symbols()];
44
45 uint32_t Symbol::pack_hash_and_refcount(short hash, int refcount) {
46 STATIC_ASSERT(PERM_REFCOUNT == ((1 << 16) - 1));
47 assert(refcount >= 0, "negative refcount");
48 assert(refcount <= PERM_REFCOUNT, "invalid refcount");
49 uint32_t hi = hash;
50 uint32_t lo = refcount;
51 return (hi << 16) | lo;
52 }
53
54 Symbol::Symbol(const u1* name, int length, int refcount) {
55 assert(length <= max_length(), "SymbolTable should have caught this!");
56 _hash_and_refcount = pack_hash_and_refcount((short)os::random(), refcount);
57 _length = (u2)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 // This copies the symbol when it is added to the ConcurrentHashTable.
68 Symbol::Symbol(const Symbol& s1) {
69 _hash_and_refcount = s1._hash_and_refcount;
70 _length = s1._length;
71 memcpy(_body, s1._body, _length);
72 }
73
74 #if INCLUDE_CDS
75 void Symbol::update_identity_hash() {
76 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
77 _hash_and_refcount = pack_hash_and_refcount((short)ArchiveBuilder::current()->entropy(), PERM_REFCOUNT);
78 }
79
80 void Symbol::set_permanent() {
81 // This is called at a safepoint during dumping of a dynamic CDS archive.
82 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
83 _hash_and_refcount = pack_hash_and_refcount(extract_hash(_hash_and_refcount), PERM_REFCOUNT);
84 }
85 #endif
86
87 // ------------------------------------------------------------------
88 // Symbol::index_of
89 //
90 // Test if we have the give substring at or after the i-th char of this
91 // symbol's utf8 bytes.
92 // Return -1 on failure. Otherwise return the first index where substr occurs.
93 int Symbol::index_of_at(int i, const char* substr, int substr_len) const {
94 assert(i >= 0 && i <= utf8_length(), "oob");
95 if (substr_len <= 0) return 0;
96 char first_char = substr[0];
97 address bytes = (address) ((Symbol*)this)->base();
98 address limit = bytes + utf8_length() - substr_len; // inclusive limit
99 address scan = bytes + i;
100 if (scan > limit)
101 return -1;
102 for (; scan <= limit; scan++) {
103 scan = (address) memchr(scan, first_char, (limit + 1 - scan));
104 if (scan == nullptr)
105 return -1; // not found
106 assert(scan >= bytes+i && scan <= limit, "scan oob");
107 if (substr_len <= 2
108 ? (char) scan[substr_len-1] == substr[substr_len-1]
109 : memcmp(scan+1, substr+1, substr_len-1) == 0) {
110 return (int)(scan - bytes);
111 }
112 }
113 return -1;
114 }
115
116 bool Symbol::is_star_match(const char* pattern) const {
117 if (strchr(pattern, '*') == nullptr) {
118 return equals(pattern);
119 } else {
120 ResourceMark rm;
121 char* buf = as_C_string();
122 return StringUtils::is_star_match(pattern, buf);
123 }
124 }
125
126 char* Symbol::as_C_string(char* buf, int size) const {
127 if (size > 0) {
128 int len = MIN2(size - 1, utf8_length());
129 for (int i = 0; i < len; i++) {
130 buf[i] = char_at(i);
131 }
132 buf[len] = '\0';
133 }
134 return buf;
135 }
136
137 char* Symbol::as_C_string() const {
138 int len = utf8_length();
139 char* str = NEW_RESOURCE_ARRAY(char, len + 1);
140 return as_C_string(str, len + 1);
141 }
142
143 void Symbol::print_utf8_on(outputStream* st) const {
144 st->print("%s", as_C_string());
145 }
146
147 void Symbol::print_symbol_on(outputStream* st) const {
148 char *s;
149 st = st ? st : tty;
150 {
151 // ResourceMark may not affect st->print(). If st is a string
152 // stream it could resize, using the same resource arena.
153 ResourceMark rm;
154 s = as_quoted_ascii();
155 s = os::strdup(s);
156 }
157 if (s == nullptr) {
158 st->print("(null)");
159 } else {
160 st->print("%s", s);
161 os::free(s);
162 }
163 }
164
165 char* Symbol::as_quoted_ascii() const {
166 const char *ptr = (const char *)&_body[0];
167 size_t quoted_length = UTF8::quoted_ascii_length(ptr, utf8_length());
168 char* result = NEW_RESOURCE_ARRAY(char, quoted_length + 1);
169 UTF8::as_quoted_ascii(ptr, utf8_length(), result, quoted_length + 1);
170 return result;
171 }
172
173 jchar* Symbol::as_unicode(int& length) const {
174 Symbol* this_ptr = (Symbol*)this;
175 length = UTF8::unicode_length((char*)this_ptr->bytes(), utf8_length());
176 jchar* result = NEW_RESOURCE_ARRAY(jchar, length);
177 if (length > 0) {
178 UTF8::convert_to_unicode((char*)this_ptr->bytes(), result, length);
179 }
180 return result;
181 }
182
183 const char* Symbol::as_klass_external_name(char* buf, int size) const {
184 if (size > 0) {
185 char* str = as_C_string(buf, size);
186 int length = (int)strlen(str);
187 // Turn all '/'s into '.'s (also for array klasses)
188 for (int index = 0; index < length; index++) {
189 if (str[index] == JVM_SIGNATURE_SLASH) {
190 str[index] = JVM_SIGNATURE_DOT;
191 }
192 }
193 return str;
194 } else {
195 return buf;
196 }
197 }
198
199 const char* Symbol::as_klass_external_name() const {
200 char* str = as_C_string();
201 int length = (int)strlen(str);
202 // Turn all '/'s into '.'s (also for array klasses)
203 for (int index = 0; index < length; index++) {
204 if (str[index] == JVM_SIGNATURE_SLASH) {
205 str[index] = JVM_SIGNATURE_DOT;
206 }
207 }
208 return str;
209 }
210
211 static void print_class(outputStream *os, const SignatureStream& ss) {
212 int sb = ss.raw_symbol_begin(), se = ss.raw_symbol_end();
213 for (int i = sb; i < se; ++i) {
214 char ch = ss.raw_char_at(i);
215 if (ch == JVM_SIGNATURE_SLASH) {
216 os->put(JVM_SIGNATURE_DOT);
217 } else {
218 os->put(ch);
219 }
220 }
221 }
222
223 static void print_array(outputStream *os, SignatureStream& ss) {
224 int dimensions = ss.skip_array_prefix();
225 assert(dimensions > 0, "");
226 if (ss.is_reference()) {
227 print_class(os, ss);
228 } else {
229 os->print("%s", type2name(ss.type()));
230 }
231 for (int i = 0; i < dimensions; ++i) {
232 os->print("[]");
233 }
234 }
235
236 void Symbol::print_as_signature_external_return_type(outputStream *os) {
237 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
238 if (ss.at_return_type()) {
239 if (ss.is_array()) {
240 print_array(os, ss);
241 } else if (ss.is_reference()) {
242 print_class(os, ss);
243 } else {
244 os->print("%s", type2name(ss.type()));
245 }
246 }
247 }
248 }
249
250 void Symbol::print_as_signature_external_parameters(outputStream *os) {
251 bool first = true;
252 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
253 if (ss.at_return_type()) break;
254 if (!first) { os->print(", "); }
255 if (ss.is_array()) {
256 print_array(os, ss);
257 } else if (ss.is_reference()) {
258 print_class(os, ss);
259 } else {
260 os->print("%s", type2name(ss.type()));
261 }
262 first = false;
263 }
264 }
265
266 void Symbol::print_as_field_external_type(outputStream *os) {
267 SignatureStream ss(this, false);
268 assert(!ss.is_done(), "must have at least one element in field ref");
269 assert(!ss.at_return_type(), "field ref cannot be a return type");
270 assert(!Signature::is_method(this), "field ref cannot be a method");
271
272 if (ss.is_array()) {
273 print_array(os, ss);
274 } else if (ss.is_reference()) {
275 print_class(os, ss);
276 } else {
277 os->print("%s", type2name(ss.type()));
278 }
279 #ifdef ASSERT
280 ss.next();
281 assert(ss.is_done(), "must have at most one element in field ref");
282 #endif
283 }
284
285 // Increment refcount while checking for zero. If the Symbol's refcount becomes zero
286 // a thread could be concurrently removing the Symbol. This is used during SymbolTable
287 // lookup to avoid reviving a dead Symbol.
288 bool Symbol::try_increment_refcount() {
289 uint32_t found = _hash_and_refcount;
290 while (true) {
291 uint32_t old_value = found;
292 int refc = extract_refcount(old_value);
293 if (refc == PERM_REFCOUNT) {
294 return true; // sticky max or created permanent
295 } else if (refc == 0) {
296 return false; // dead, can't revive.
297 } else {
298 found = AtomicAccess::cmpxchg(&_hash_and_refcount, old_value, old_value + 1);
299 if (found == old_value) {
300 return true; // successfully updated.
301 }
302 // refcount changed, try again.
303 }
304 }
305 }
306
307 // The increment_refcount() is called when not doing lookup. It is assumed that you
308 // have a symbol with a non-zero refcount and it can't become zero while referenced by
309 // this caller.
310 void Symbol::increment_refcount() {
311 if (!try_increment_refcount()) {
312 print();
313 fatal("refcount has gone to zero");
314 }
315 #ifndef PRODUCT
316 if (refcount() != PERM_REFCOUNT) { // not a permanent symbol
317 NOT_PRODUCT(AtomicAccess::inc(&_total_count);)
318 }
319 #endif
320 }
321
322 // Decrement refcount potentially while racing increment, so we need
323 // to check the value after attempting to decrement so that if another
324 // thread increments to PERM_REFCOUNT the value is not decremented.
325 void Symbol::decrement_refcount() {
326 uint32_t found = _hash_and_refcount;
327 while (true) {
328 uint32_t old_value = found;
329 int refc = extract_refcount(old_value);
330 if (refc == PERM_REFCOUNT) {
331 return; // refcount is permanent, permanent is sticky
332 } else if (refc == 0) {
333 print();
334 fatal("refcount underflow");
335 return;
336 } else {
337 found = AtomicAccess::cmpxchg(&_hash_and_refcount, old_value, old_value - 1);
338 if (found == old_value) {
339 return; // successfully updated.
340 }
341 // refcount changed, try again.
342 }
343 }
344 }
345
346 void Symbol::make_permanent() {
347 uint32_t found = _hash_and_refcount;
348 while (true) {
349 uint32_t old_value = found;
350 int refc = extract_refcount(old_value);
351 if (refc == PERM_REFCOUNT) {
352 return; // refcount is permanent, permanent is sticky
353 } else if (refc == 0) {
354 print();
355 fatal("refcount underflow");
356 return;
357 } else {
358 short hash = extract_hash(old_value);
359 found = AtomicAccess::cmpxchg(&_hash_and_refcount, old_value, pack_hash_and_refcount(hash, PERM_REFCOUNT));
360 if (found == old_value) {
361 return; // successfully updated.
362 }
363 // refcount changed, try again.
364 }
365 }
366 }
367
368 void Symbol::metaspace_pointers_do(MetaspaceClosure* it) {
369 if (log_is_enabled(Trace, aot)) {
370 LogStream trace_stream(Log(aot)::trace());
371 trace_stream.print("Iter(Symbol): %p ", this);
372 print_value_on(&trace_stream);
373 trace_stream.cr();
374 }
375 }
376
377 void Symbol::print_on(outputStream* st) const {
378 st->print("Symbol: '");
379 print_symbol_on(st);
380 st->print("'");
381 st->print(" count %d", refcount());
382 }
383
384 void Symbol::print() const { print_on(tty); }
385
386 // The print_value functions are present in all builds, to support the
387 // disassembler and error reporting.
388 void Symbol::print_value_on(outputStream* st) const {
389 st->print_raw("'", 1);
390 st->print_raw((const char*)base(), utf8_length());
391 st->print_raw("'", 1);
392 }
393
394 void Symbol::print_value() const { print_value_on(tty); }
395
396 bool Symbol::is_valid(Symbol* s) {
397 if (!is_aligned(s, sizeof(MetaWord))) return false;
398 if ((size_t)s < os::min_page_size()) return false;
399
400 if (!os::is_readable_range(s, s + 1)) return false;
401
402 // Symbols are not allocated in Java heap.
403 if (Universe::heap()->is_in(s)) return false;
404
405 int len = s->utf8_length();
406 if (len < 0) return false;
407
408 jbyte* bytes = (jbyte*) s->bytes();
409 return os::is_readable_range(bytes, bytes + len);
410 }
411
412 // SymbolTable prints this in its statistics
413 NOT_PRODUCT(size_t Symbol::_total_count = 0;)
414
415 #ifndef PRODUCT
416 bool Symbol::is_valid_id(vmSymbolID vm_symbol_id) {
417 return vmSymbols::is_valid_id(vm_symbol_id);
418 }
419 #endif