1 /*
2 * Copyright (c) 1997, 2024, 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 "cds/archiveBuilder.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/stringUtils.hpp"
43 #include "utilities/utf8.hpp"
44
45 Symbol* Symbol::_vm_symbols[vmSymbols::number_of_symbols()];
46
47 uint32_t Symbol::pack_hash_and_refcount(short hash, int refcount) {
48 STATIC_ASSERT(PERM_REFCOUNT == ((1 << 16) - 1));
49 assert(refcount >= 0, "negative refcount");
50 assert(refcount <= PERM_REFCOUNT, "invalid refcount");
51 uint32_t hi = hash;
52 uint32_t lo = refcount;
53 return (hi << 16) | lo;
54 }
55
56 Symbol::Symbol(const u1* name, int length, int refcount) {
57 assert(length <= max_length(), "SymbolTable should have caught this!");
58 _hash_and_refcount = pack_hash_and_refcount((short)os::random(), refcount);
59 _length = (u2)length;
60 // _body[0..1] are allocated in the header just by coincidence in the current
61 // implementation of Symbol. They are read by identity_hash(), so make sure they
62 // are initialized.
63 // No other code should assume that _body[0..1] are always allocated. E.g., do
64 // not unconditionally read base()[0] as that will be invalid for an empty Symbol.
65 _body[0] = _body[1] = 0;
66 memcpy(_body, name, length);
67 }
68
69 // This copies the symbol when it is added to the ConcurrentHashTable.
70 Symbol::Symbol(const Symbol& s1) {
71 _hash_and_refcount = s1._hash_and_refcount;
72 _length = s1._length;
73 memcpy(_body, s1._body, _length);
74 }
75
76 #if INCLUDE_CDS
77 void Symbol::update_identity_hash() {
78 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
79 _hash_and_refcount = pack_hash_and_refcount((short)ArchiveBuilder::current()->entropy(), PERM_REFCOUNT);
80 }
81
82 void Symbol::set_permanent() {
83 // This is called at a safepoint during dumping of a dynamic CDS archive.
84 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
85 _hash_and_refcount = pack_hash_and_refcount(extract_hash(_hash_and_refcount), PERM_REFCOUNT);
86 }
87 #endif
88
89 // ------------------------------------------------------------------
90 // Symbol::index_of
91 //
92 // Test if we have the give substring at or after the i-th char of this
93 // symbol's utf8 bytes.
94 // Return -1 on failure. Otherwise return the first index where substr occurs.
95 int Symbol::index_of_at(int i, const char* substr, int substr_len) const {
96 assert(i >= 0 && i <= utf8_length(), "oob");
97 if (substr_len <= 0) return 0;
98 char first_char = substr[0];
99 address bytes = (address) ((Symbol*)this)->base();
100 address limit = bytes + utf8_length() - substr_len; // inclusive limit
101 address scan = bytes + i;
102 if (scan > limit)
103 return -1;
104 for (; scan <= limit; scan++) {
105 scan = (address) memchr(scan, first_char, (limit + 1 - scan));
106 if (scan == nullptr)
107 return -1; // not found
108 assert(scan >= bytes+i && scan <= limit, "scan oob");
109 if (substr_len <= 2
110 ? (char) scan[substr_len-1] == substr[substr_len-1]
111 : memcmp(scan+1, substr+1, substr_len-1) == 0) {
112 return (int)(scan - bytes);
113 }
114 }
115 return -1;
116 }
117
118 bool Symbol::is_star_match(const char* pattern) const {
119 if (strchr(pattern, '*') == nullptr) {
120 return equals(pattern);
121 } else {
122 ResourceMark rm;
123 char* buf = as_C_string();
124 return StringUtils::is_star_match(pattern, buf);
125 }
126 }
127
128 char* Symbol::as_C_string(char* buf, int size) const {
129 if (size > 0) {
130 int len = MIN2(size - 1, utf8_length());
131 for (int i = 0; i < len; i++) {
132 buf[i] = char_at(i);
133 }
134 buf[len] = '\0';
135 }
136 return buf;
137 }
138
139 char* Symbol::as_C_string() const {
140 int len = utf8_length();
141 char* str = NEW_RESOURCE_ARRAY(char, len + 1);
142 return as_C_string(str, len + 1);
143 }
144
145 void Symbol::print_utf8_on(outputStream* st) const {
146 st->print("%s", as_C_string());
147 }
148
149 void Symbol::print_symbol_on(outputStream* st) const {
150 char *s;
151 st = st ? st : tty;
152 {
153 // ResourceMark may not affect st->print(). If st is a string
154 // stream it could resize, using the same resource arena.
155 ResourceMark rm;
156 s = as_quoted_ascii();
157 s = os::strdup(s);
158 }
159 if (s == nullptr) {
160 st->print("(null)");
161 } else {
162 st->print("%s", s);
163 os::free(s);
164 }
165 }
166
167 char* Symbol::as_quoted_ascii() const {
168 const char *ptr = (const char *)&_body[0];
169 size_t quoted_length = UTF8::quoted_ascii_length(ptr, utf8_length());
170 char* result = NEW_RESOURCE_ARRAY(char, quoted_length + 1);
171 UTF8::as_quoted_ascii(ptr, utf8_length(), result, quoted_length + 1);
172 return result;
173 }
174
175 jchar* Symbol::as_unicode(int& length) const {
176 Symbol* this_ptr = (Symbol*)this;
177 length = UTF8::unicode_length((char*)this_ptr->bytes(), utf8_length());
178 jchar* result = NEW_RESOURCE_ARRAY(jchar, length);
179 if (length > 0) {
180 UTF8::convert_to_unicode((char*)this_ptr->bytes(), result, length);
181 }
182 return result;
183 }
184
185 const char* Symbol::as_klass_external_name(char* buf, int size) const {
186 if (size > 0) {
187 char* str = as_C_string(buf, size);
188 int length = (int)strlen(str);
189 // Turn all '/'s into '.'s (also for array klasses)
190 for (int index = 0; index < length; index++) {
191 if (str[index] == JVM_SIGNATURE_SLASH) {
192 str[index] = JVM_SIGNATURE_DOT;
193 }
194 }
195 return str;
196 } else {
197 return buf;
198 }
199 }
200
201 const char* Symbol::as_klass_external_name() const {
202 char* str = as_C_string();
203 int length = (int)strlen(str);
204 // Turn all '/'s into '.'s (also for array klasses)
205 for (int index = 0; index < length; index++) {
206 if (str[index] == JVM_SIGNATURE_SLASH) {
207 str[index] = JVM_SIGNATURE_DOT;
208 }
209 }
210 return str;
211 }
212
213 static void print_class(outputStream *os, const SignatureStream& ss) {
214 int sb = ss.raw_symbol_begin(), se = ss.raw_symbol_end();
215 for (int i = sb; i < se; ++i) {
216 char ch = ss.raw_char_at(i);
217 if (ch == JVM_SIGNATURE_SLASH) {
218 os->put(JVM_SIGNATURE_DOT);
219 } else {
220 os->put(ch);
221 }
222 }
223 }
224
225 static void print_array(outputStream *os, SignatureStream& ss) {
226 int dimensions = ss.skip_array_prefix();
227 assert(dimensions > 0, "");
228 if (ss.is_reference()) {
229 print_class(os, ss);
230 } else {
231 os->print("%s", type2name(ss.type()));
232 }
233 for (int i = 0; i < dimensions; ++i) {
234 os->print("[]");
235 }
236 }
237
238 void Symbol::print_as_signature_external_return_type(outputStream *os) {
239 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
240 if (ss.at_return_type()) {
241 if (ss.is_array()) {
242 print_array(os, ss);
243 } else if (ss.is_reference()) {
244 print_class(os, ss);
245 } else {
246 os->print("%s", type2name(ss.type()));
247 }
248 }
249 }
250 }
251
252 void Symbol::print_as_signature_external_parameters(outputStream *os) {
253 bool first = true;
254 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
255 if (ss.at_return_type()) break;
256 if (!first) { os->print(", "); }
257 if (ss.is_array()) {
258 print_array(os, ss);
259 } else if (ss.is_reference()) {
260 print_class(os, ss);
261 } else {
262 os->print("%s", type2name(ss.type()));
263 }
264 first = false;
265 }
266 }
267
268 void Symbol::print_as_field_external_type(outputStream *os) {
269 SignatureStream ss(this, false);
270 assert(!ss.is_done(), "must have at least one element in field ref");
271 assert(!ss.at_return_type(), "field ref cannot be a return type");
272 assert(!Signature::is_method(this), "field ref cannot be a method");
273
274 if (ss.is_array()) {
275 print_array(os, ss);
276 } else if (ss.is_reference()) {
277 print_class(os, ss);
278 } else {
279 os->print("%s", type2name(ss.type()));
280 }
281 #ifdef ASSERT
282 ss.next();
283 assert(ss.is_done(), "must have at most one element in field ref");
284 #endif
285 }
286
287 // Increment refcount while checking for zero. If the Symbol's refcount becomes zero
288 // a thread could be concurrently removing the Symbol. This is used during SymbolTable
289 // lookup to avoid reviving a dead Symbol.
290 bool Symbol::try_increment_refcount() {
291 uint32_t found = _hash_and_refcount;
292 while (true) {
293 uint32_t old_value = found;
294 int refc = extract_refcount(old_value);
295 if (refc == PERM_REFCOUNT) {
296 return true; // sticky max or created permanent
297 } else if (refc == 0) {
298 return false; // dead, can't revive.
299 } else {
300 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value + 1);
301 if (found == old_value) {
302 return true; // successfully updated.
303 }
304 // refcount changed, try again.
305 }
306 }
307 }
308
309 // The increment_refcount() is called when not doing lookup. It is assumed that you
310 // have a symbol with a non-zero refcount and it can't become zero while referenced by
311 // this caller.
312 void Symbol::increment_refcount() {
313 if (!try_increment_refcount()) {
314 print();
315 fatal("refcount has gone to zero");
316 }
317 #ifndef PRODUCT
318 if (refcount() != PERM_REFCOUNT) { // not a permanent symbol
319 NOT_PRODUCT(Atomic::inc(&_total_count);)
320 }
321 #endif
322 }
323
324 // Decrement refcount potentially while racing increment, so we need
325 // to check the value after attempting to decrement so that if another
326 // thread increments to PERM_REFCOUNT the value is not decremented.
327 void Symbol::decrement_refcount() {
328 uint32_t found = _hash_and_refcount;
329 while (true) {
330 uint32_t old_value = found;
331 int refc = extract_refcount(old_value);
332 if (refc == PERM_REFCOUNT) {
333 return; // refcount is permanent, permanent is sticky
334 } else if (refc == 0) {
335 print();
336 fatal("refcount underflow");
337 return;
338 } else {
339 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value - 1);
340 if (found == old_value) {
341 return; // successfully updated.
342 }
343 // refcount changed, try again.
344 }
345 }
346 }
347
348 void Symbol::make_permanent() {
349 uint32_t found = _hash_and_refcount;
350 while (true) {
351 uint32_t old_value = found;
352 int refc = extract_refcount(old_value);
353 if (refc == PERM_REFCOUNT) {
354 return; // refcount is permanent, permanent is sticky
355 } else if (refc == 0) {
356 print();
357 fatal("refcount underflow");
358 return;
359 } else {
360 short hash = extract_hash(old_value);
361 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, pack_hash_and_refcount(hash, PERM_REFCOUNT));
362 if (found == old_value) {
363 return; // successfully updated.
364 }
365 // refcount changed, try again.
366 }
367 }
368 }
369
370 void Symbol::metaspace_pointers_do(MetaspaceClosure* it) {
371 if (log_is_enabled(Trace, cds)) {
372 LogStream trace_stream(Log(cds)::trace());
373 trace_stream.print("Iter(Symbol): %p ", this);
374 print_value_on(&trace_stream);
375 trace_stream.cr();
376 }
377 }
378
379 void Symbol::print_on(outputStream* st) const {
380 st->print("Symbol: '");
381 print_symbol_on(st);
382 st->print("'");
383 st->print(" count %d", refcount());
384 }
385
386 void Symbol::print() const { print_on(tty); }
387
388 // The print_value functions are present in all builds, to support the
389 // disassembler and error reporting.
390 void Symbol::print_value_on(outputStream* st) const {
391 st->print_raw("'", 1);
392 st->print_raw((const char*)base(), utf8_length());
393 st->print_raw("'", 1);
394 }
395
396 void Symbol::print_value() const { print_value_on(tty); }
397
398 bool Symbol::is_valid(Symbol* s) {
399 if (!is_aligned(s, sizeof(MetaWord))) return false;
400 if ((size_t)s < os::min_page_size()) return false;
401
402 if (!os::is_readable_range(s, s + 1)) return false;
403
404 // Symbols are not allocated in Java heap.
405 if (Universe::heap()->is_in(s)) return false;
406
407 int len = s->utf8_length();
408 if (len < 0) return false;
409
410 jbyte* bytes = (jbyte*) s->bytes();
411 return os::is_readable_range(bytes, bytes + len);
412 }
413
414 // SymbolTable prints this in its statistics
415 NOT_PRODUCT(size_t Symbol::_total_count = 0;)
416
417 #ifndef PRODUCT
418 bool Symbol::is_valid_id(vmSymbolID vm_symbol_id) {
419 return vmSymbols::is_valid_id(vm_symbol_id);
420 }
421 #endif