1 /* 2 * Copyright (c) 2005, 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_RUNTIME_BIASEDLOCKING_HPP 26 #define SHARE_RUNTIME_BIASEDLOCKING_HPP 27 28 #include "runtime/handles.hpp" 29 #include "utilities/growableArray.hpp" 30 31 // This class describes operations to implement Store-Free Biased 32 // Locking. The high-level properties of the scheme are similar to 33 // IBM's lock reservation, Dice-Moir-Scherer QR locks, and other biased 34 // locking mechanisms. The principal difference is in the handling of 35 // recursive locking which is how this technique achieves a more 36 // efficient fast path than these other schemes. 37 // 38 // The basic observation is that in HotSpot's current fast locking 39 // scheme, recursive locking (in the fast path) causes no update to 40 // the object header. The recursion is described simply by stack 41 // records containing a specific value (NULL). Only the last unlock by 42 // a given thread causes an update to the object header. 43 // 44 // This observation, coupled with the fact that HotSpot only compiles 45 // methods for which monitor matching is obeyed (and which therefore 46 // can not throw IllegalMonitorStateException), implies that we can 47 // completely eliminate modifications to the object header for 48 // recursive locking in compiled code, and perform similar recursion 49 // checks and throwing of IllegalMonitorStateException in the 50 // interpreter with little or no impact on the performance of the fast 51 // path. 52 // 53 // The basic algorithm is as follows (note, see below for more details 54 // and information). A pattern in the low three bits is reserved in 55 // the object header to indicate whether biasing of a given object's 56 // lock is currently being done or is allowed at all. If the bias 57 // pattern is present, the contents of the rest of the header are 58 // either the JavaThread* of the thread to which the lock is biased, 59 // or NULL, indicating that the lock is "anonymously biased". The 60 // first thread which locks an anonymously biased object biases the 61 // lock toward that thread. If another thread subsequently attempts to 62 // lock the same object, the bias is revoked. 63 // 64 // Because there are no updates to the object header at all during 65 // recursive locking while the lock is biased, the biased lock entry 66 // code is simply a test of the object header's value. If this test 67 // succeeds, the lock has been acquired by the thread. If this test 68 // fails, a bit test is done to see whether the bias bit is still 69 // set. If not, we fall back to HotSpot's original CAS-based locking 70 // scheme. If it is set, we attempt to CAS in a bias toward this 71 // thread. The latter operation is expected to be the rarest operation 72 // performed on these locks. We optimistically expect the biased lock 73 // entry to hit most of the time, and want the CAS-based fallthrough 74 // to occur quickly in the situations where the bias has been revoked. 75 // 76 // Revocation of the lock's bias is fairly straightforward. We want to 77 // restore the object's header and stack-based BasicObjectLocks and 78 // BasicLocks to the state they would have been in had the object been 79 // locked by HotSpot's usual fast locking scheme. To do this, we bring 80 // the system to a safepoint and walk the stack of the thread toward 81 // which the lock is biased. We find all of the lock records on the 82 // stack corresponding to this object, in particular the first / 83 // "highest" record. We fill in the highest lock record with the 84 // object's displaced header (which is a well-known value given that 85 // we don't maintain an identity hash nor age bits for the object 86 // while it's in the biased state) and all other lock records with 0, 87 // the value for recursive locks. When the safepoint is released, the 88 // formerly-biased thread and all other threads revert back to 89 // HotSpot's CAS-based locking. 90 // 91 // This scheme can not handle transfers of biases of single objects 92 // from thread to thread efficiently, but it can handle bulk transfers 93 // of such biases, which is a usage pattern showing up in some 94 // applications and benchmarks. We implement "bulk rebias" and "bulk 95 // revoke" operations using a "bias epoch" on a per-data-type basis. 96 // If too many bias revocations are occurring for a particular data 97 // type, the bias epoch for the data type is incremented at a 98 // safepoint, effectively meaning that all previous biases are 99 // invalid. The fast path locking case checks for an invalid epoch in 100 // the object header and attempts to rebias the object with a CAS if 101 // found, avoiding safepoints or bulk heap sweeps (the latter which 102 // was used in a prior version of this algorithm and did not scale 103 // well). If too many bias revocations persist, biasing is completely 104 // disabled for the data type by resetting the prototype header to the 105 // unbiased markOop. The fast-path locking code checks to see whether 106 // the instance's bias pattern differs from the prototype header's and 107 // causes the bias to be revoked without reaching a safepoint or, 108 // again, a bulk heap sweep. 109 110 // Biased locking counters 111 class BiasedLockingCounters { 112 private: 113 int _total_entry_count; 114 int _biased_lock_entry_count; 115 int _anonymously_biased_lock_entry_count; 116 int _rebiased_lock_entry_count; 117 int _revoked_lock_entry_count; 118 int _fast_path_entry_count; 119 int _slow_path_entry_count; 120 121 public: BiasedLockingCounters()122 BiasedLockingCounters() : 123 _total_entry_count(0), 124 _biased_lock_entry_count(0), 125 _anonymously_biased_lock_entry_count(0), 126 _rebiased_lock_entry_count(0), 127 _revoked_lock_entry_count(0), 128 _fast_path_entry_count(0), 129 _slow_path_entry_count(0) {} 130 131 int slow_path_entry_count() const; // Compute this field if necessary 132 total_entry_count_addr()133 int* total_entry_count_addr() { return &_total_entry_count; } biased_lock_entry_count_addr()134 int* biased_lock_entry_count_addr() { return &_biased_lock_entry_count; } anonymously_biased_lock_entry_count_addr()135 int* anonymously_biased_lock_entry_count_addr() { return &_anonymously_biased_lock_entry_count; } rebiased_lock_entry_count_addr()136 int* rebiased_lock_entry_count_addr() { return &_rebiased_lock_entry_count; } revoked_lock_entry_count_addr()137 int* revoked_lock_entry_count_addr() { return &_revoked_lock_entry_count; } fast_path_entry_count_addr()138 int* fast_path_entry_count_addr() { return &_fast_path_entry_count; } slow_path_entry_count_addr()139 int* slow_path_entry_count_addr() { return &_slow_path_entry_count; } 140 nonzero()141 bool nonzero() { return _total_entry_count > 0; } 142 143 void print_on(outputStream* st) const; 144 void print() const; 145 }; 146 147 148 class BiasedLocking : AllStatic { 149 private: 150 static BiasedLockingCounters _counters; 151 152 public: 153 static int* total_entry_count_addr(); 154 static int* biased_lock_entry_count_addr(); 155 static int* anonymously_biased_lock_entry_count_addr(); 156 static int* rebiased_lock_entry_count_addr(); 157 static int* revoked_lock_entry_count_addr(); 158 static int* fast_path_entry_count_addr(); 159 static int* slow_path_entry_count_addr(); 160 161 enum Condition { 162 NOT_BIASED = 1, 163 BIAS_REVOKED = 2, 164 BIAS_REVOKED_AND_REBIASED = 3 165 }; 166 167 // This initialization routine should only be called once and 168 // schedules a PeriodicTask to turn on biased locking a few seconds 169 // into the VM run to avoid startup time regressions 170 static void init(); 171 172 // This provides a global switch for leaving biased locking disabled 173 // for the first part of a run and enabling it later 174 static bool enabled(); 175 176 // This should be called by JavaThreads to revoke the bias of an object 177 static Condition revoke_and_rebias(Handle obj, bool attempt_rebias, TRAPS); 178 179 // These do not allow rebiasing; they are used by deoptimization to 180 // ensure that monitors on the stack can be migrated 181 static void revoke(GrowableArray<Handle>* objs); 182 static void revoke_at_safepoint(Handle obj); 183 static void revoke_at_safepoint(GrowableArray<Handle>* objs); 184 print_counters()185 static void print_counters() { _counters.print(); } counters()186 static BiasedLockingCounters* counters() { return &_counters; } 187 188 // These routines are GC-related and should not be called by end 189 // users. GCs which do not do preservation of mark words do not need 190 // to call these routines. 191 static void preserve_marks(); 192 static void restore_marks(); 193 }; 194 195 #endif // SHARE_RUNTIME_BIASEDLOCKING_HPP 196