1 /*
2 * Copyright (c) 1997, 2019, 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 #ifndef SHARE_OOPS_MARKOOP_HPP
26 #define SHARE_OOPS_MARKOOP_HPP
27
28 #include "oops/oop.hpp"
29
30 // The markOop describes the header of an object.
31 //
32 // Note that the mark is not a real oop but just a word.
33 // It is placed in the oop hierarchy for historical reasons.
34 //
35 // Bit-format of an object header (most significant first, big endian layout below):
36 //
37 // 32 bits:
38 // --------
39 // hash:25 ------------>| age:4 biased_lock:1 lock:2 (normal object)
40 // JavaThread*:23 epoch:2 age:4 biased_lock:1 lock:2 (biased object)
41 // size:32 ------------------------------------------>| (CMS free block)
42 // PromotedObject*:29 ---------->| promo_bits:3 ----->| (CMS promoted object)
43 //
44 // 64 bits:
45 // --------
46 // unused:25 hash:31 -->| unused:1 age:4 biased_lock:1 lock:2 (normal object)
47 // JavaThread*:54 epoch:2 unused:1 age:4 biased_lock:1 lock:2 (biased object)
48 // PromotedObject*:61 --------------------->| promo_bits:3 ----->| (CMS promoted object)
49 // size:64 ----------------------------------------------------->| (CMS free block)
50 //
51 // unused:25 hash:31 -->| cms_free:1 age:4 biased_lock:1 lock:2 (COOPs && normal object)
52 // JavaThread*:54 epoch:2 cms_free:1 age:4 biased_lock:1 lock:2 (COOPs && biased object)
53 // narrowOop:32 unused:24 cms_free:1 unused:4 promo_bits:3 ----->| (COOPs && CMS promoted object)
54 // unused:21 size:35 -->| cms_free:1 unused:7 ------------------>| (COOPs && CMS free block)
55 //
56 // - hash contains the identity hash value: largest value is
57 // 31 bits, see os::random(). Also, 64-bit vm's require
58 // a hash value no bigger than 32 bits because they will not
59 // properly generate a mask larger than that: see library_call.cpp
60 // and c1_CodePatterns_sparc.cpp.
61 //
62 // - the biased lock pattern is used to bias a lock toward a given
63 // thread. When this pattern is set in the low three bits, the lock
64 // is either biased toward a given thread or "anonymously" biased,
65 // indicating that it is possible for it to be biased. When the
66 // lock is biased toward a given thread, locking and unlocking can
67 // be performed by that thread without using atomic operations.
68 // When a lock's bias is revoked, it reverts back to the normal
69 // locking scheme described below.
70 //
71 // Note that we are overloading the meaning of the "unlocked" state
72 // of the header. Because we steal a bit from the age we can
73 // guarantee that the bias pattern will never be seen for a truly
74 // unlocked object.
75 //
76 // Note also that the biased state contains the age bits normally
77 // contained in the object header. Large increases in scavenge
78 // times were seen when these bits were absent and an arbitrary age
79 // assigned to all biased objects, because they tended to consume a
80 // significant fraction of the eden semispaces and were not
81 // promoted promptly, causing an increase in the amount of copying
82 // performed. The runtime system aligns all JavaThread* pointers to
83 // a very large value (currently 128 bytes (32bVM) or 256 bytes (64bVM))
84 // to make room for the age bits & the epoch bits (used in support of
85 // biased locking), and for the CMS "freeness" bit in the 64bVM (+COOPs).
86 //
87 // [JavaThread* | epoch | age | 1 | 01] lock is biased toward given thread
88 // [0 | epoch | age | 1 | 01] lock is anonymously biased
89 //
90 // - the two lock bits are used to describe three states: locked/unlocked and monitor.
91 //
92 // [ptr | 00] locked ptr points to real header on stack
93 // [header | 0 | 01] unlocked regular object header
94 // [ptr | 10] monitor inflated lock (header is wapped out)
95 // [ptr | 11] marked used by markSweep to mark an object
96 // not valid at any other time
97 //
98 // We assume that stack/thread pointers have the lowest two bits cleared.
99
100 class BasicLock;
101 class ObjectMonitor;
102 class JavaThread;
103
104 class markOopDesc: public oopDesc {
105 private:
106 // Conversion
107 uintptr_t value() const { return (uintptr_t) this; }
108
109 public:
110 // Constants
111 enum { age_bits = 4,
112 lock_bits = 2,
113 biased_lock_bits = 1,
114 max_hash_bits = BitsPerWord - age_bits - lock_bits - biased_lock_bits,
115 hash_bits = max_hash_bits > 31 ? 31 : max_hash_bits,
116 cms_bits = LP64_ONLY(1) NOT_LP64(0),
117 epoch_bits = 2
118 };
119
120 // The biased locking code currently requires that the age bits be
121 // contiguous to the lock bits.
122 enum { lock_shift = 0,
123 biased_lock_shift = lock_bits,
124 age_shift = lock_bits + biased_lock_bits,
125 cms_shift = age_shift + age_bits,
126 hash_shift = cms_shift + cms_bits,
127 epoch_shift = hash_shift
128 };
129
130 enum { lock_mask = right_n_bits(lock_bits),
131 lock_mask_in_place = lock_mask << lock_shift,
132 biased_lock_mask = right_n_bits(lock_bits + biased_lock_bits),
133 biased_lock_mask_in_place= biased_lock_mask << lock_shift,
134 biased_lock_bit_in_place = 1 << biased_lock_shift,
135 age_mask = right_n_bits(age_bits),
136 age_mask_in_place = age_mask << age_shift,
137 epoch_mask = right_n_bits(epoch_bits),
138 epoch_mask_in_place = epoch_mask << epoch_shift,
139 cms_mask = right_n_bits(cms_bits),
140 cms_mask_in_place = cms_mask << cms_shift
141 };
142
143 const static uintptr_t hash_mask = right_n_bits(hash_bits);
144 const static uintptr_t hash_mask_in_place = hash_mask << hash_shift;
145
146 // Alignment of JavaThread pointers encoded in object header required by biased locking
147 enum { biased_lock_alignment = 2 << (epoch_shift + epoch_bits)
148 };
149
150 enum { locked_value = 0,
151 unlocked_value = 1,
152 monitor_value = 2,
153 marked_value = 3,
154 biased_lock_pattern = 5
155 };
156
157 enum { no_hash = 0 }; // no hash value assigned
158
159 enum { no_hash_in_place = (address_word)no_hash << hash_shift,
160 no_lock_in_place = unlocked_value
161 };
162
163 enum { max_age = age_mask };
164
165 enum { max_bias_epoch = epoch_mask };
166
167 // Biased Locking accessors.
168 // These must be checked by all code which calls into the
169 // ObjectSynchronizer and other code. The biasing is not understood
170 // by the lower-level CAS-based locking code, although the runtime
171 // fixes up biased locks to be compatible with it when a bias is
172 // revoked.
173 bool has_bias_pattern() const {
174 return (mask_bits(value(), biased_lock_mask_in_place) == biased_lock_pattern);
175 }
176 JavaThread* biased_locker() const {
177 assert(has_bias_pattern(), "should not call this otherwise");
178 return (JavaThread*) ((intptr_t) (mask_bits(value(), ~(biased_lock_mask_in_place | age_mask_in_place | epoch_mask_in_place))));
179 }
180 // Indicates that the mark has the bias bit set but that it has not
181 // yet been biased toward a particular thread
182 bool is_biased_anonymously() const {
183 return (has_bias_pattern() && (biased_locker() == NULL));
184 }
185 // Indicates epoch in which this bias was acquired. If the epoch
186 // changes due to too many bias revocations occurring, the biases
187 // from the previous epochs are all considered invalid.
188 int bias_epoch() const {
189 assert(has_bias_pattern(), "should not call this otherwise");
190 return (mask_bits(value(), epoch_mask_in_place) >> epoch_shift);
191 }
192 markOop set_bias_epoch(int epoch) {
193 assert(has_bias_pattern(), "should not call this otherwise");
194 assert((epoch & (~epoch_mask)) == 0, "epoch overflow");
195 return markOop(mask_bits(value(), ~epoch_mask_in_place) | (epoch << epoch_shift));
196 }
197 markOop incr_bias_epoch() {
198 return set_bias_epoch((1 + bias_epoch()) & epoch_mask);
199 }
200 // Prototype mark for initialization
201 static markOop biased_locking_prototype() {
202 return markOop( biased_lock_pattern );
203 }
204
205 // lock accessors (note that these assume lock_shift == 0)
206 bool is_locked() const {
207 return (mask_bits(value(), lock_mask_in_place) != unlocked_value);
208 }
209 bool is_unlocked() const {
210 return (mask_bits(value(), biased_lock_mask_in_place) == unlocked_value);
211 }
212 bool is_marked() const {
213 return (mask_bits(value(), lock_mask_in_place) == marked_value);
214 }
215 bool is_neutral() const { return (mask_bits(value(), biased_lock_mask_in_place) == unlocked_value); }
216
217 // Special temporary state of the markOop while being inflated.
218 // Code that looks at mark outside a lock need to take this into account.
219 bool is_being_inflated() const { return (value() == 0); }
220
221 // Distinguished markword value - used when inflating over
222 // an existing stacklock. 0 indicates the markword is "BUSY".
223 // Lockword mutators that use a LD...CAS idiom should always
224 // check for and avoid overwriting a 0 value installed by some
225 // other thread. (They should spin or block instead. The 0 value
226 // is transient and *should* be short-lived).
227 static markOop INFLATING() { return (markOop) 0; } // inflate-in-progress
228
229 // Should this header be preserved during GC?
230 inline bool must_be_preserved(oop obj_containing_mark) const;
231 inline bool must_be_preserved_with_bias(oop obj_containing_mark) const;
232
233 // Should this header (including its age bits) be preserved in the
234 // case of a promotion failure during scavenge?
235 // Note that we special case this situation. We want to avoid
236 // calling BiasedLocking::preserve_marks()/restore_marks() (which
237 // decrease the number of mark words that need to be preserved
238 // during GC) during each scavenge. During scavenges in which there
239 // is no promotion failure, we actually don't need to call the above
240 // routines at all, since we don't mutate and re-initialize the
241 // marks of promoted objects using init_mark(). However, during
242 // scavenges which result in promotion failure, we do re-initialize
243 // the mark words of objects, meaning that we should have called
244 // these mark word preservation routines. Currently there's no good
245 // place in which to call them in any of the scavengers (although
246 // guarded by appropriate locks we could make one), but the
247 // observation is that promotion failures are quite rare and
248 // reducing the number of mark words preserved during them isn't a
249 // high priority.
250 inline bool must_be_preserved_for_promotion_failure(oop obj_containing_mark) const;
251 inline bool must_be_preserved_with_bias_for_promotion_failure(oop obj_containing_mark) const;
252
253 // Should this header be preserved during a scavenge where CMS is
254 // the old generation?
255 // (This is basically the same body as must_be_preserved_for_promotion_failure(),
256 // but takes the Klass* as argument instead)
257 inline bool must_be_preserved_for_cms_scavenge(Klass* klass_of_obj_containing_mark) const;
258 inline bool must_be_preserved_with_bias_for_cms_scavenge(Klass* klass_of_obj_containing_mark) const;
259
260 // WARNING: The following routines are used EXCLUSIVELY by
261 // synchronization functions. They are not really gc safe.
262 // They must get updated if markOop layout get changed.
263 markOop set_unlocked() const {
264 return markOop(value() | unlocked_value);
265 }
266 bool has_locker() const {
267 return ((value() & lock_mask_in_place) == locked_value);
268 }
269 BasicLock* locker() const {
270 assert(has_locker(), "check");
271 return (BasicLock*) value();
272 }
273 bool has_monitor() const {
274 return ((value() & monitor_value) != 0);
275 }
276 ObjectMonitor* monitor() const {
277 assert(has_monitor(), "check");
278 // Use xor instead of &~ to provide one extra tag-bit check.
279 return (ObjectMonitor*) (value() ^ monitor_value);
280 }
281 bool has_displaced_mark_helper() const {
282 return ((value() & unlocked_value) == 0);
283 }
284 markOop displaced_mark_helper() const {
285 assert(has_displaced_mark_helper(), "check");
286 intptr_t ptr = (value() & ~monitor_value);
287 return *(markOop*)ptr;
288 }
289 void set_displaced_mark_helper(markOop m) const {
290 assert(has_displaced_mark_helper(), "check");
291 intptr_t ptr = (value() & ~monitor_value);
292 *(markOop*)ptr = m;
293 }
294 markOop copy_set_hash(intptr_t hash) const {
295 intptr_t tmp = value() & (~hash_mask_in_place);
296 tmp |= ((hash & hash_mask) << hash_shift);
297 return (markOop)tmp;
298 }
299 // it is only used to be stored into BasicLock as the
300 // indicator that the lock is using heavyweight monitor
301 static markOop unused_mark() {
302 return (markOop) marked_value;
303 }
304 // the following two functions create the markOop to be
305 // stored into object header, it encodes monitor info
306 static markOop encode(BasicLock* lock) {
307 return (markOop) lock;
308 }
309 static markOop encode(ObjectMonitor* monitor) {
310 intptr_t tmp = (intptr_t) monitor;
311 return (markOop) (tmp | monitor_value);
312 }
313 static markOop encode(JavaThread* thread, uint age, int bias_epoch) {
314 intptr_t tmp = (intptr_t) thread;
315 assert(UseBiasedLocking && ((tmp & (epoch_mask_in_place | age_mask_in_place | biased_lock_mask_in_place)) == 0), "misaligned JavaThread pointer");
316 assert(age <= max_age, "age too large");
317 assert(bias_epoch <= max_bias_epoch, "bias epoch too large");
318 return (markOop) (tmp | (bias_epoch << epoch_shift) | (age << age_shift) | biased_lock_pattern);
319 }
320
321 // used to encode pointers during GC
322 markOop clear_lock_bits() { return markOop(value() & ~lock_mask_in_place); }
323
324 // age operations
325 markOop set_marked() { return markOop((value() & ~lock_mask_in_place) | marked_value); }
326 markOop set_unmarked() { return markOop((value() & ~lock_mask_in_place) | unlocked_value); }
327
328 uint age() const { return mask_bits(value() >> age_shift, age_mask); }
329 markOop set_age(uint v) const {
330 assert((v & ~age_mask) == 0, "shouldn't overflow age field");
331 return markOop((value() & ~age_mask_in_place) | (((uintptr_t)v & age_mask) << age_shift));
332 }
333 markOop incr_age() const { return age() == max_age ? markOop(this) : set_age(age() + 1); }
334
335 // hash operations
336 intptr_t hash() const {
337 return mask_bits(value() >> hash_shift, hash_mask);
338 }
339
340 bool has_no_hash() const {
341 return hash() == no_hash;
342 }
343
344 // Prototype mark for initialization
345 static markOop prototype() {
346 return markOop( no_hash_in_place | no_lock_in_place );
347 }
348
349 // Helper function for restoration of unmarked mark oops during GC
350 static inline markOop prototype_for_object(oop obj);
351
352 // Debugging
353 void print_on(outputStream* st) const;
354
355 // Prepare address of oop for placement into mark
356 inline static markOop encode_pointer_as_mark(void* p) { return markOop(p)->set_marked(); }
357
358 // Recover address of oop from encoded form used in mark
359 inline void* decode_pointer() { if (UseBiasedLocking && has_bias_pattern()) return NULL; return clear_lock_bits(); }
360
361 // These markOops indicate cms free chunk blocks and not objects.
362 // In 64 bit, the markOop is set to distinguish them from oops.
363 // These are defined in 32 bit mode for vmStructs.
364 const static uintptr_t cms_free_chunk_pattern = 0x1;
365
366 // Constants for the size field.
367 enum { size_shift = cms_shift + cms_bits,
368 size_bits = 35 // need for compressed oops 32G
369 };
370 // These values are too big for Win64
371 const static uintptr_t size_mask = LP64_ONLY(right_n_bits(size_bits))
372 NOT_LP64(0);
373 const static uintptr_t size_mask_in_place =
374 (address_word)size_mask << size_shift;
375
376 #ifdef _LP64
377 static markOop cms_free_prototype() {
378 return markOop(((intptr_t)prototype() & ~cms_mask_in_place) |
379 ((cms_free_chunk_pattern & cms_mask) << cms_shift));
380 }
381 uintptr_t cms_encoding() const {
382 return mask_bits(value() >> cms_shift, cms_mask);
383 }
384 bool is_cms_free_chunk() const {
385 return is_neutral() &&
386 (cms_encoding() & cms_free_chunk_pattern) == cms_free_chunk_pattern;
387 }
388
389 size_t get_size() const { return (size_t)(value() >> size_shift); }
390 static markOop set_size_and_free(size_t size) {
391 assert((size & ~size_mask) == 0, "shouldn't overflow size field");
392 return markOop(((intptr_t)cms_free_prototype() & ~size_mask_in_place) |
393 (((intptr_t)size & size_mask) << size_shift));
394 }
395 #endif // _LP64
396 };
397
398 #endif // SHARE_OOPS_MARKOOP_HPP