1 /*
2 * Copyright (c) 1997, 2020, 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 "classfile/altHashing.hpp"
28 #include "classfile/classLoaderData.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/metaspaceShared.hpp"
34 #include "memory/resourceArea.hpp"
35 #include "memory/universe.hpp"
36 #include "oops/symbol.hpp"
37 #include "runtime/atomic.hpp"
38 #include "runtime/os.hpp"
39 #include "runtime/signature.hpp"
40 #include "utilities/utf8.hpp"
41
42 uint32_t Symbol::pack_hash_and_refcount(short hash, int refcount) {
43 STATIC_ASSERT(PERM_REFCOUNT == ((1 << 16) - 1));
44 assert(refcount >= 0, "negative refcount");
45 assert(refcount <= PERM_REFCOUNT, "invalid refcount");
46 uint32_t hi = hash;
47 uint32_t lo = refcount;
48 return (hi << 16) | lo;
49 }
50
51 Symbol::Symbol(const u1* name, int length, int refcount) {
52 _hash_and_refcount = pack_hash_and_refcount((short)os::random(), refcount);
53 _length = length;
54 _body[0] = 0; // in case length == 0
55 for (int i = 0; i < length; i++) {
56 byte_at_put(i, name[i]);
57 }
58 }
59
60 void* Symbol::operator new(size_t sz, int len) throw() {
61 #if INCLUDE_CDS
62 if (DumpSharedSpaces) {
63 // To get deterministic output from -Xshare:dump, we ensure that Symbols are allocated in
64 // increasing addresses. When the symbols are copied into the archive, we preserve their
65 // relative address order (see SortedSymbolClosure in metaspaceShared.cpp)
66 //
67 // We cannot use arena because arena chunks are allocated by the OS. As a result, for example,
68 // the archived symbol of "java/lang/Object" may sometimes be lower than "java/lang/String", and
69 // sometimes be higher. This would cause non-deterministic contents in the archive.
70 DEBUG_ONLY(static void* last = 0);
71 void* p = (void*)MetaspaceShared::symbol_space_alloc(size(len)*wordSize);
72 assert(p > last, "must increase monotonically");
73 DEBUG_ONLY(last = p);
74 return p;
75 }
76 #endif
77 int alloc_size = size(len)*wordSize;
78 address res = (address) AllocateHeap(alloc_size, mtSymbol);
79 return res;
80 }
81
82 void* Symbol::operator new(size_t sz, int len, Arena* arena) throw() {
83 int alloc_size = size(len)*wordSize;
84 address res = (address)arena->Amalloc_4(alloc_size);
85 return res;
86 }
87
88 void Symbol::operator delete(void *p) {
89 assert(((Symbol*)p)->refcount() == 0, "should not call this");
90 FreeHeap(p);
91 }
92
93 #if INCLUDE_CDS
94 void Symbol::update_identity_hash() {
95 // This is called at a safepoint during dumping of a static CDS archive. The caller should have
96 // called os::init_random() with a deterministic seed and then iterate all archived Symbols in
97 // a deterministic order.
98 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
99 _hash_and_refcount = pack_hash_and_refcount((short)os::random(), PERM_REFCOUNT);
100 }
101
102 void Symbol::set_permanent() {
103 // This is called at a safepoint during dumping of a dynamic CDS archive.
104 assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
105 _hash_and_refcount = pack_hash_and_refcount(extract_hash(_hash_and_refcount), PERM_REFCOUNT);
106 }
107 #endif
108
109 // ------------------------------------------------------------------
110 // Symbol::index_of
111 //
112 // Finds if the given string is a substring of this symbol's utf8 bytes.
113 // Return -1 on failure. Otherwise return the first index where str occurs.
114 int Symbol::index_of_at(int i, const char* str, int len) const {
115 assert(i >= 0 && i <= utf8_length(), "oob");
116 if (len <= 0) return 0;
117 char first_char = str[0];
118 address bytes = (address) ((Symbol*)this)->base();
119 address limit = bytes + utf8_length() - len; // inclusive limit
120 address scan = bytes + i;
121 if (scan > limit)
122 return -1;
123 for (; scan <= limit; scan++) {
124 scan = (address) memchr(scan, first_char, (limit + 1 - scan));
125 if (scan == NULL)
126 return -1; // not found
127 assert(scan >= bytes+i && scan <= limit, "scan oob");
128 if (len <= 2
129 ? (char) scan[len-1] == str[len-1]
130 : memcmp(scan+1, str+1, len-1) == 0) {
131 return (int)(scan - bytes);
132 }
133 }
134 return -1;
135 }
136
137
138 char* Symbol::as_C_string(char* buf, int size) const {
139 if (size > 0) {
140 int len = MIN2(size - 1, utf8_length());
141 for (int i = 0; i < len; i++) {
142 buf[i] = char_at(i);
143 }
144 buf[len] = '\0';
145 }
146 return buf;
147 }
148
149 char* Symbol::as_C_string() const {
150 int len = utf8_length();
151 char* str = NEW_RESOURCE_ARRAY(char, len + 1);
152 return as_C_string(str, len + 1);
153 }
154
155 void Symbol::print_utf8_on(outputStream* st) const {
156 st->print("%s", as_C_string());
157 }
158
159 void Symbol::print_symbol_on(outputStream* st) const {
160 char *s;
161 st = st ? st : tty;
162 {
163 // ResourceMark may not affect st->print(). If st is a string
164 // stream it could resize, using the same resource arena.
165 ResourceMark rm;
166 s = as_quoted_ascii();
167 s = os::strdup(s);
168 }
169 if (s == NULL) {
170 st->print("(null)");
171 } else {
172 st->print("%s", s);
173 os::free(s);
174 }
175 }
176
177 char* Symbol::as_quoted_ascii() const {
178 const char *ptr = (const char *)&_body[0];
179 int quoted_length = UTF8::quoted_ascii_length(ptr, utf8_length());
180 char* result = NEW_RESOURCE_ARRAY(char, quoted_length + 1);
181 UTF8::as_quoted_ascii(ptr, utf8_length(), result, quoted_length + 1);
182 return result;
183 }
184
185 jchar* Symbol::as_unicode(int& length) const {
186 Symbol* this_ptr = (Symbol*)this;
187 length = UTF8::unicode_length((char*)this_ptr->bytes(), utf8_length());
188 jchar* result = NEW_RESOURCE_ARRAY(jchar, length);
189 if (length > 0) {
190 UTF8::convert_to_unicode((char*)this_ptr->bytes(), result, length);
191 }
192 return result;
193 }
194
195 const char* Symbol::as_klass_external_name(char* buf, int size) const {
196 if (size > 0) {
197 char* str = as_C_string(buf, size);
198 int length = (int)strlen(str);
199 // Turn all '/'s into '.'s (also for array klasses)
200 for (int index = 0; index < length; index++) {
201 if (str[index] == JVM_SIGNATURE_SLASH) {
202 str[index] = JVM_SIGNATURE_DOT;
203 }
204 }
205 return str;
206 } else {
207 return buf;
208 }
209 }
210
211 const char* Symbol::as_klass_external_name() const {
212 char* str = as_C_string();
213 int length = (int)strlen(str);
214 // Turn all '/'s into '.'s (also for array klasses)
215 for (int index = 0; index < length; index++) {
216 if (str[index] == JVM_SIGNATURE_SLASH) {
217 str[index] = JVM_SIGNATURE_DOT;
218 }
219 }
220 return str;
221 }
222
223 static void print_class(outputStream *os, const SignatureStream& ss) {
224 int sb = ss.raw_symbol_begin(), se = ss.raw_symbol_end();
225 for (int i = sb; i < se; ++i) {
226 int ch = ss.raw_char_at(i);
227 if (ch == JVM_SIGNATURE_SLASH) {
228 os->put(JVM_SIGNATURE_DOT);
229 } else {
230 os->put(ch);
231 }
232 }
233 }
234
235 static void print_array(outputStream *os, SignatureStream& ss) {
236 int dimensions = ss.skip_array_prefix();
237 assert(dimensions > 0, "");
238 if (ss.is_reference()) {
239 print_class(os, ss);
240 } else {
241 os->print("%s", type2name(ss.type()));
242 }
243 for (int i = 0; i < dimensions; ++i) {
244 os->print("[]");
245 }
246 }
247
248 void Symbol::print_as_signature_external_return_type(outputStream *os) {
249 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
250 if (ss.at_return_type()) {
251 if (ss.is_array()) {
252 print_array(os, ss);
253 } else if (ss.is_reference()) {
254 print_class(os, ss);
255 } else {
256 os->print("%s", type2name(ss.type()));
257 }
258 }
259 }
260 }
261
262 void Symbol::print_as_signature_external_parameters(outputStream *os) {
263 bool first = true;
264 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
265 if (ss.at_return_type()) break;
266 if (!first) { os->print(", "); }
267 if (ss.is_array()) {
268 print_array(os, ss);
269 } else if (ss.is_reference()) {
270 print_class(os, ss);
271 } else {
272 os->print("%s", type2name(ss.type()));
273 }
274 first = false;
275 }
276 }
277
278 // Increment refcount while checking for zero. If the Symbol's refcount becomes zero
279 // a thread could be concurrently removing the Symbol. This is used during SymbolTable
280 // lookup to avoid reviving a dead Symbol.
281 bool Symbol::try_increment_refcount() {
282 uint32_t found = _hash_and_refcount;
283 while (true) {
284 uint32_t old_value = found;
285 int refc = extract_refcount(old_value);
286 if (refc == PERM_REFCOUNT) {
287 return true; // sticky max or created permanent
288 } else if (refc == 0) {
289 return false; // dead, can't revive.
290 } else {
291 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value + 1);
292 if (found == old_value) {
293 return true; // successfully updated.
294 }
295 // refcount changed, try again.
296 }
297 }
298 }
299
300 // The increment_refcount() is called when not doing lookup. It is assumed that you
301 // have a symbol with a non-zero refcount and it can't become zero while referenced by
302 // this caller.
303 void Symbol::increment_refcount() {
304 if (!try_increment_refcount()) {
305 #ifdef ASSERT
306 print();
307 fatal("refcount has gone to zero");
308 #endif
309 }
310 #ifndef PRODUCT
311 if (refcount() != PERM_REFCOUNT) { // not a permanent symbol
312 NOT_PRODUCT(Atomic::inc(&_total_count);)
313 }
314 #endif
315 }
316
317 // Decrement refcount potentially while racing increment, so we need
318 // to check the value after attempting to decrement so that if another
319 // thread increments to PERM_REFCOUNT the value is not decremented.
320 void Symbol::decrement_refcount() {
321 uint32_t found = _hash_and_refcount;
322 while (true) {
323 uint32_t old_value = found;
324 int refc = extract_refcount(old_value);
325 if (refc == PERM_REFCOUNT) {
326 return; // refcount is permanent, permanent is sticky
327 } else if (refc == 0) {
328 #ifdef ASSERT
329 print();
330 fatal("refcount underflow");
331 #endif
332 return;
333 } else {
334 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value - 1);
335 if (found == old_value) {
336 return; // successfully updated.
337 }
338 // refcount changed, try again.
339 }
340 }
341 }
342
343 void Symbol::make_permanent() {
344 uint32_t found = _hash_and_refcount;
345 while (true) {
346 uint32_t old_value = found;
347 int refc = extract_refcount(old_value);
348 if (refc == PERM_REFCOUNT) {
349 return; // refcount is permanent, permanent is sticky
350 } else if (refc == 0) {
351 #ifdef ASSERT
352 print();
353 fatal("refcount underflow");
354 #endif
355 return;
356 } else {
357 int hash = extract_hash(old_value);
358 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, pack_hash_and_refcount(hash, PERM_REFCOUNT));
359 if (found == old_value) {
360 return; // successfully updated.
361 }
362 // refcount changed, try again.
363 }
364 }
365 }
366
367 void Symbol::metaspace_pointers_do(MetaspaceClosure* it) {
368 if (log_is_enabled(Trace, cds)) {
369 LogStream trace_stream(Log(cds)::trace());
370 trace_stream.print("Iter(Symbol): %p ", this);
371 print_value_on(&trace_stream);
372 trace_stream.cr();
373 }
374 }
375
376 void Symbol::print_on(outputStream* st) const {
377 st->print("Symbol: '");
378 print_symbol_on(st);
379 st->print("'");
380 st->print(" count %d", refcount());
381 }
382
383 void Symbol::print() const { print_on(tty); }
384
385 // The print_value functions are present in all builds, to support the
386 // disassembler and error reporting.
387 void Symbol::print_value_on(outputStream* st) const {
388 st->print("'");
389 for (int i = 0; i < utf8_length(); i++) {
390 st->print("%c", char_at(i));
391 }
392 st->print("'");
393 }
394
395 void Symbol::print_value() const { print_value_on(tty); }
396
397 bool Symbol::is_valid(Symbol* s) {
398 if (!is_aligned(s, sizeof(MetaWord))) return false;
399 if ((size_t)s < os::min_page_size()) return false;
400
401 if (!os::is_readable_range(s, s + 1)) return false;
402
403 // Symbols are not allocated in Java heap.
404 if (Universe::heap()->is_in(s)) return false;
405
406 int len = s->utf8_length();
407 if (len < 0) return false;
408
409 jbyte* bytes = (jbyte*) s->bytes();
410 return os::is_readable_range(bytes, bytes + len);
411 }
412
413 // SymbolTable prints this in its statistics
414 NOT_PRODUCT(size_t Symbol::_total_count = 0;)