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
pack_hash_and_refcount(short hash,int refcount)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
Symbol(const u1 * name,int length,int refcount)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
operator new(size_t sz,int len)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
operator new(size_t sz,int len,Arena * arena)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->Amalloc_4(alloc_size);
93 return res;
94 }
95
operator delete(void * p)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
update_identity_hash()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
set_permanent()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 // ------------------------------------------------------------------
118 // Symbol::index_of
119 //
120 // Finds if the given string is a substring of this symbol's utf8 bytes.
121 // Return -1 on failure. Otherwise return the first index where str occurs.
index_of_at(int i,const char * str,int len) const122 int Symbol::index_of_at(int i, const char* str, int len) const {
123 assert(i >= 0 && i <= utf8_length(), "oob");
124 if (len <= 0) return 0;
125 char first_char = str[0];
126 address bytes = (address) ((Symbol*)this)->base();
127 address limit = bytes + utf8_length() - len; // inclusive limit
128 address scan = bytes + i;
129 if (scan > limit)
130 return -1;
131 for (; scan <= limit; scan++) {
132 scan = (address) memchr(scan, first_char, (limit + 1 - scan));
133 if (scan == NULL)
134 return -1; // not found
135 assert(scan >= bytes+i && scan <= limit, "scan oob");
136 if (len <= 2
137 ? (char) scan[len-1] == str[len-1]
138 : memcmp(scan+1, str+1, len-1) == 0) {
139 return (int)(scan - bytes);
140 }
141 }
142 return -1;
143 }
144
145
as_C_string(char * buf,int size) const146 char* Symbol::as_C_string(char* buf, int size) const {
147 if (size > 0) {
148 int len = MIN2(size - 1, utf8_length());
149 for (int i = 0; i < len; i++) {
150 buf[i] = char_at(i);
151 }
152 buf[len] = '\0';
153 }
154 return buf;
155 }
156
as_C_string() const157 char* Symbol::as_C_string() const {
158 int len = utf8_length();
159 char* str = NEW_RESOURCE_ARRAY(char, len + 1);
160 return as_C_string(str, len + 1);
161 }
162
print_utf8_on(outputStream * st) const163 void Symbol::print_utf8_on(outputStream* st) const {
164 st->print("%s", as_C_string());
165 }
166
print_symbol_on(outputStream * st) const167 void Symbol::print_symbol_on(outputStream* st) const {
168 char *s;
169 st = st ? st : tty;
170 {
171 // ResourceMark may not affect st->print(). If st is a string
172 // stream it could resize, using the same resource arena.
173 ResourceMark rm;
174 s = as_quoted_ascii();
175 s = os::strdup(s);
176 }
177 if (s == NULL) {
178 st->print("(null)");
179 } else {
180 st->print("%s", s);
181 os::free(s);
182 }
183 }
184
as_quoted_ascii() const185 char* Symbol::as_quoted_ascii() const {
186 const char *ptr = (const char *)&_body[0];
187 int quoted_length = UTF8::quoted_ascii_length(ptr, utf8_length());
188 char* result = NEW_RESOURCE_ARRAY(char, quoted_length + 1);
189 UTF8::as_quoted_ascii(ptr, utf8_length(), result, quoted_length + 1);
190 return result;
191 }
192
as_unicode(int & length) const193 jchar* Symbol::as_unicode(int& length) const {
194 Symbol* this_ptr = (Symbol*)this;
195 length = UTF8::unicode_length((char*)this_ptr->bytes(), utf8_length());
196 jchar* result = NEW_RESOURCE_ARRAY(jchar, length);
197 if (length > 0) {
198 UTF8::convert_to_unicode((char*)this_ptr->bytes(), result, length);
199 }
200 return result;
201 }
202
as_klass_external_name(char * buf,int size) const203 const char* Symbol::as_klass_external_name(char* buf, int size) const {
204 if (size > 0) {
205 char* str = as_C_string(buf, size);
206 int length = (int)strlen(str);
207 // Turn all '/'s into '.'s (also for array klasses)
208 for (int index = 0; index < length; index++) {
209 if (str[index] == JVM_SIGNATURE_SLASH) {
210 str[index] = JVM_SIGNATURE_DOT;
211 }
212 }
213 return str;
214 } else {
215 return buf;
216 }
217 }
218
as_klass_external_name() const219 const char* Symbol::as_klass_external_name() const {
220 char* str = as_C_string();
221 int length = (int)strlen(str);
222 // Turn all '/'s into '.'s (also for array klasses)
223 for (int index = 0; index < length; index++) {
224 if (str[index] == JVM_SIGNATURE_SLASH) {
225 str[index] = JVM_SIGNATURE_DOT;
226 }
227 }
228 return str;
229 }
230
print_class(outputStream * os,const SignatureStream & ss)231 static void print_class(outputStream *os, const SignatureStream& ss) {
232 int sb = ss.raw_symbol_begin(), se = ss.raw_symbol_end();
233 for (int i = sb; i < se; ++i) {
234 int ch = ss.raw_char_at(i);
235 if (ch == JVM_SIGNATURE_SLASH) {
236 os->put(JVM_SIGNATURE_DOT);
237 } else {
238 os->put(ch);
239 }
240 }
241 }
242
print_array(outputStream * os,SignatureStream & ss)243 static void print_array(outputStream *os, SignatureStream& ss) {
244 int dimensions = ss.skip_array_prefix();
245 assert(dimensions > 0, "");
246 if (ss.is_reference()) {
247 print_class(os, ss);
248 } else {
249 os->print("%s", type2name(ss.type()));
250 }
251 for (int i = 0; i < dimensions; ++i) {
252 os->print("[]");
253 }
254 }
255
print_as_signature_external_return_type(outputStream * os)256 void Symbol::print_as_signature_external_return_type(outputStream *os) {
257 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
258 if (ss.at_return_type()) {
259 if (ss.is_array()) {
260 print_array(os, ss);
261 } else if (ss.is_reference()) {
262 print_class(os, ss);
263 } else {
264 os->print("%s", type2name(ss.type()));
265 }
266 }
267 }
268 }
269
print_as_signature_external_parameters(outputStream * os)270 void Symbol::print_as_signature_external_parameters(outputStream *os) {
271 bool first = true;
272 for (SignatureStream ss(this); !ss.is_done(); ss.next()) {
273 if (ss.at_return_type()) break;
274 if (!first) { os->print(", "); }
275 if (ss.is_array()) {
276 print_array(os, ss);
277 } else if (ss.is_reference()) {
278 print_class(os, ss);
279 } else {
280 os->print("%s", type2name(ss.type()));
281 }
282 first = false;
283 }
284 }
285
286 // Increment refcount while checking for zero. If the Symbol's refcount becomes zero
287 // a thread could be concurrently removing the Symbol. This is used during SymbolTable
288 // lookup to avoid reviving a dead Symbol.
try_increment_refcount()289 bool Symbol::try_increment_refcount() {
290 uint32_t found = _hash_and_refcount;
291 while (true) {
292 uint32_t old_value = found;
293 int refc = extract_refcount(old_value);
294 if (refc == PERM_REFCOUNT) {
295 return true; // sticky max or created permanent
296 } else if (refc == 0) {
297 return false; // dead, can't revive.
298 } else {
299 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value + 1);
300 if (found == old_value) {
301 return true; // successfully updated.
302 }
303 // refcount changed, try again.
304 }
305 }
306 }
307
308 // The increment_refcount() is called when not doing lookup. It is assumed that you
309 // have a symbol with a non-zero refcount and it can't become zero while referenced by
310 // this caller.
increment_refcount()311 void Symbol::increment_refcount() {
312 if (!try_increment_refcount()) {
313 #ifdef ASSERT
314 print();
315 fatal("refcount has gone to zero");
316 #endif
317 }
318 #ifndef PRODUCT
319 if (refcount() != PERM_REFCOUNT) { // not a permanent symbol
320 NOT_PRODUCT(Atomic::inc(&_total_count);)
321 }
322 #endif
323 }
324
325 // Decrement refcount potentially while racing increment, so we need
326 // to check the value after attempting to decrement so that if another
327 // thread increments to PERM_REFCOUNT the value is not decremented.
decrement_refcount()328 void Symbol::decrement_refcount() {
329 uint32_t found = _hash_and_refcount;
330 while (true) {
331 uint32_t old_value = found;
332 int refc = extract_refcount(old_value);
333 if (refc == PERM_REFCOUNT) {
334 return; // refcount is permanent, permanent is sticky
335 } else if (refc == 0) {
336 #ifdef ASSERT
337 print();
338 fatal("refcount underflow");
339 #endif
340 return;
341 } else {
342 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, old_value - 1);
343 if (found == old_value) {
344 return; // successfully updated.
345 }
346 // refcount changed, try again.
347 }
348 }
349 }
350
make_permanent()351 void Symbol::make_permanent() {
352 uint32_t found = _hash_and_refcount;
353 while (true) {
354 uint32_t old_value = found;
355 int refc = extract_refcount(old_value);
356 if (refc == PERM_REFCOUNT) {
357 return; // refcount is permanent, permanent is sticky
358 } else if (refc == 0) {
359 #ifdef ASSERT
360 print();
361 fatal("refcount underflow");
362 #endif
363 return;
364 } else {
365 int hash = extract_hash(old_value);
366 found = Atomic::cmpxchg(&_hash_and_refcount, old_value, pack_hash_and_refcount(hash, PERM_REFCOUNT));
367 if (found == old_value) {
368 return; // successfully updated.
369 }
370 // refcount changed, try again.
371 }
372 }
373 }
374
metaspace_pointers_do(MetaspaceClosure * it)375 void Symbol::metaspace_pointers_do(MetaspaceClosure* it) {
376 if (log_is_enabled(Trace, cds)) {
377 LogStream trace_stream(Log(cds)::trace());
378 trace_stream.print("Iter(Symbol): %p ", this);
379 print_value_on(&trace_stream);
380 trace_stream.cr();
381 }
382 }
383
print_on(outputStream * st) const384 void Symbol::print_on(outputStream* st) const {
385 st->print("Symbol: '");
386 print_symbol_on(st);
387 st->print("'");
388 st->print(" count %d", refcount());
389 }
390
print() const391 void Symbol::print() const { print_on(tty); }
392
393 // The print_value functions are present in all builds, to support the
394 // disassembler and error reporting.
print_value_on(outputStream * st) const395 void Symbol::print_value_on(outputStream* st) const {
396 st->print("'");
397 for (int i = 0; i < utf8_length(); i++) {
398 st->print("%c", char_at(i));
399 }
400 st->print("'");
401 }
402
print_value() const403 void Symbol::print_value() const { print_value_on(tty); }
404
is_valid(Symbol * s)405 bool Symbol::is_valid(Symbol* s) {
406 if (!is_aligned(s, sizeof(MetaWord))) return false;
407 if ((size_t)s < os::min_page_size()) return false;
408
409 if (!os::is_readable_range(s, s + 1)) return false;
410
411 // Symbols are not allocated in Java heap.
412 if (Universe::heap()->is_in(s)) return false;
413
414 int len = s->utf8_length();
415 if (len < 0) return false;
416
417 jbyte* bytes = (jbyte*) s->bytes();
418 return os::is_readable_range(bytes, bytes + len);
419 }
420
421 // SymbolTable prints this in its statistics
NOT_PRODUCT(size_t Symbol::_total_count=0;)422 NOT_PRODUCT(size_t Symbol::_total_count = 0;)
423
424 #ifndef PRODUCT
425 bool Symbol::is_valid_id(vmSymbolID vm_symbol_id) {
426 return vmSymbols::is_valid_id(vm_symbol_id);
427 }
428 #endif
429