Skip to content

D
dragonwell8_hotspot

openanolis / dragonwell8_hotspot

通知 2
Star 2
Fork 0
D
dragonwell8_hotspot
提交 f9bc752b 编写于 作者: A acorn
浏览文件
操作

6988353: refactor contended sync subsystem 

Summary: reduce complexity by factoring synchronizer.cpp
Reviewed-by: dholmes, never, coleenp
上级 044288ff
隐藏空白更改
内联 并排
Showing with 5165 addition and 5352 deletion
src/os/linux/vm/objectMonitor_linux.hpp 已删除 100644 → 0
浏览文件 @ 044288ff
/*
* Copyright (c) 1999, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
private:
src/os/linux/vm/objectMonitor_linux.inline.hpp 已删除 100644 → 0
浏览文件 @ 044288ff
/*
* Copyright (c) 1999, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
src/os/solaris/vm/objectMonitor_solaris.cpp 已删除 100644 → 0
浏览文件 @ 044288ff
/*
* Copyright (c) 1998, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
src/os/solaris/vm/objectMonitor_solaris.hpp 已删除 100644 → 0
浏览文件 @ 044288ff
/*
* Copyright (c) 1998, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
private:
src/os/solaris/vm/objectMonitor_solaris.inline.hpp 已删除 100644 → 0
浏览文件 @ 044288ff
/*
* Copyright (c) 1998, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
src/os/windows/vm/objectMonitor_windows.cpp 已删除 100644 → 0
浏览文件 @ 044288ff
/*
* Copyright (c) 1998, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
#include "incls/_precompiled.incl"
src/os/windows/vm/objectMonitor_windows.inline.hpp 已删除 100644 → 0
浏览文件 @ 044288ff
/*
* Copyright (c) 1998, 2005, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
src/share/vm/includeDB_compiler1
浏览文件 @ f9bc752b
...@@ -301,6 +301,7 @@ c1_MacroAssembler.hpp assembler.hpp ...@@ -301,6 +301,7 @@ c1_MacroAssembler.hpp assembler.hpp
c1_MacroAssembler.hpp assembler_<arch>.inline.hpp c1_MacroAssembler.hpp assembler_<arch>.inline.hpp
c1_MacroAssembler_<arch>.cpp arrayOop.hpp c1_MacroAssembler_<arch>.cpp arrayOop.hpp
c1_MacroAssembler_<arch>.cpp basicLock.hpp
c1_MacroAssembler_<arch>.cpp biasedLocking.hpp c1_MacroAssembler_<arch>.cpp biasedLocking.hpp
c1_MacroAssembler_<arch>.cpp c1_MacroAssembler.hpp c1_MacroAssembler_<arch>.cpp c1_MacroAssembler.hpp
c1_MacroAssembler_<arch>.cpp c1_Runtime1.hpp c1_MacroAssembler_<arch>.cpp c1_Runtime1.hpp
...@@ -309,7 +310,6 @@ c1_MacroAssembler_<arch>.cpp interpreter.hpp ...@@ -309,7 +310,6 @@ c1_MacroAssembler_<arch>.cpp interpreter.hpp
c1_MacroAssembler_<arch>.cpp markOop.hpp c1_MacroAssembler_<arch>.cpp markOop.hpp
c1_MacroAssembler_<arch>.cpp os.hpp c1_MacroAssembler_<arch>.cpp os.hpp
c1_MacroAssembler_<arch>.cpp stubRoutines.hpp c1_MacroAssembler_<arch>.cpp stubRoutines.hpp
c1_MacroAssembler_<arch>.cpp synchronizer.hpp
c1_MacroAssembler_<arch>.cpp systemDictionary.hpp c1_MacroAssembler_<arch>.cpp systemDictionary.hpp
c1_MacroAssembler_<arch>.hpp generate_platform_dependent_include c1_MacroAssembler_<arch>.hpp generate_platform_dependent_include
......
src/share/vm/includeDB_core
浏览文件 @ f9bc752b
...@@ -300,10 +300,17 @@ barrierSet.hpp oopsHierarchy.hpp ...@@ -300,10 +300,17 @@ barrierSet.hpp oopsHierarchy.hpp
barrierSet.inline.hpp barrierSet.hpp barrierSet.inline.hpp barrierSet.hpp
barrierSet.inline.hpp cardTableModRefBS.hpp barrierSet.inline.hpp cardTableModRefBS.hpp
basicLock.cpp basicLock.hpp
basicLock.cpp synchronizer.hpp
basicLock.hpp handles.hpp
basicLock.hpp markOop.hpp
basicLock.hpp top.hpp
biasedLocking.cpp basicLock.hpp
biasedLocking.cpp biasedLocking.hpp biasedLocking.cpp biasedLocking.hpp
biasedLocking.cpp klass.inline.hpp biasedLocking.cpp klass.inline.hpp
biasedLocking.cpp markOop.hpp biasedLocking.cpp markOop.hpp
biasedLocking.cpp synchronizer.hpp
biasedLocking.cpp task.hpp biasedLocking.cpp task.hpp
biasedLocking.cpp vframe.hpp biasedLocking.cpp vframe.hpp
biasedLocking.cpp vmThread.hpp biasedLocking.cpp vmThread.hpp
...@@ -404,13 +411,13 @@ bytecodeInterpreter_<arch>.cpp vframeArray.hpp ...@@ -404,13 +411,13 @@ bytecodeInterpreter_<arch>.cpp vframeArray.hpp
bytecodeInterpreterWithChecks.cpp bytecodeInterpreter.cpp bytecodeInterpreterWithChecks.cpp bytecodeInterpreter.cpp
bytecodeInterpreter.hpp allocation.hpp bytecodeInterpreter.hpp allocation.hpp
bytecodeInterpreter.hpp basicLock.hpp
bytecodeInterpreter.hpp bytes_<arch>.hpp bytecodeInterpreter.hpp bytes_<arch>.hpp
bytecodeInterpreter.hpp frame.hpp bytecodeInterpreter.hpp frame.hpp
bytecodeInterpreter.hpp globalDefinitions.hpp bytecodeInterpreter.hpp globalDefinitions.hpp
bytecodeInterpreter.hpp globals.hpp bytecodeInterpreter.hpp globals.hpp
bytecodeInterpreter.hpp methodDataOop.hpp bytecodeInterpreter.hpp methodDataOop.hpp
bytecodeInterpreter.hpp methodOop.hpp bytecodeInterpreter.hpp methodOop.hpp
bytecodeInterpreter.hpp synchronizer.hpp
bytecodeInterpreter.inline.hpp bytecodeInterpreter.hpp bytecodeInterpreter.inline.hpp bytecodeInterpreter.hpp
bytecodeInterpreter.inline.hpp stubRoutines.hpp bytecodeInterpreter.inline.hpp stubRoutines.hpp
...@@ -1667,10 +1674,10 @@ frame.cpp stubRoutines.hpp ...@@ -1667,10 +1674,10 @@ frame.cpp stubRoutines.hpp
frame.cpp universe.inline.hpp frame.cpp universe.inline.hpp
frame.hpp assembler.hpp frame.hpp assembler.hpp
frame.hpp basicLock.hpp
frame.hpp methodOop.hpp frame.hpp methodOop.hpp
frame.hpp monitorChunk.hpp frame.hpp monitorChunk.hpp
frame.hpp registerMap.hpp frame.hpp registerMap.hpp
frame.hpp synchronizer.hpp
frame.hpp top.hpp frame.hpp top.hpp
frame.inline.hpp bytecodeInterpreter.hpp frame.inline.hpp bytecodeInterpreter.hpp
...@@ -2120,6 +2127,7 @@ interfaceSupport.hpp vmThread.hpp ...@@ -2120,6 +2127,7 @@ interfaceSupport.hpp vmThread.hpp
interfaceSupport_<os_family>.hpp generate_platform_dependent_include interfaceSupport_<os_family>.hpp generate_platform_dependent_include
interp_masm_<arch_model>.cpp arrayOop.hpp interp_masm_<arch_model>.cpp arrayOop.hpp
interp_masm_<arch_model>.cpp basicLock.hpp
interp_masm_<arch_model>.cpp biasedLocking.hpp interp_masm_<arch_model>.cpp biasedLocking.hpp
interp_masm_<arch_model>.cpp interp_masm_<arch_model>.hpp interp_masm_<arch_model>.cpp interp_masm_<arch_model>.hpp
interp_masm_<arch_model>.cpp interpreterRuntime.hpp interp_masm_<arch_model>.cpp interpreterRuntime.hpp
...@@ -2131,7 +2139,6 @@ interp_masm_<arch_model>.cpp markOop.hpp ...@@ -2131,7 +2139,6 @@ interp_masm_<arch_model>.cpp markOop.hpp
interp_masm_<arch_model>.cpp methodDataOop.hpp interp_masm_<arch_model>.cpp methodDataOop.hpp
interp_masm_<arch_model>.cpp methodOop.hpp interp_masm_<arch_model>.cpp methodOop.hpp
interp_masm_<arch_model>.cpp sharedRuntime.hpp interp_masm_<arch_model>.cpp sharedRuntime.hpp
interp_masm_<arch_model>.cpp synchronizer.hpp
interp_masm_<arch_model>.cpp thread_<os_family>.inline.hpp interp_masm_<arch_model>.cpp thread_<os_family>.inline.hpp
interp_masm_<arch_model>.hpp assembler_<arch>.inline.hpp interp_masm_<arch_model>.hpp assembler_<arch>.inline.hpp
...@@ -3094,25 +3101,26 @@ objArrayOop.cpp oop.inline.hpp ...@@ -3094,25 +3101,26 @@ objArrayOop.cpp oop.inline.hpp
objArrayOop.hpp arrayOop.hpp objArrayOop.hpp arrayOop.hpp
objectMonitor.hpp os.hpp objectMonitor.cpp dtrace.hpp
objectMonitor.cpp handles.inline.hpp
objectMonitor_<os_family>.cpp dtrace.hpp objectMonitor.cpp interfaceSupport.hpp
objectMonitor_<os_family>.cpp interfaceSupport.hpp objectMonitor.cpp markOop.hpp
objectMonitor_<os_family>.cpp objectMonitor.hpp objectMonitor.cpp mutexLocker.hpp
objectMonitor_<os_family>.cpp objectMonitor.inline.hpp objectMonitor.cpp objectMonitor.hpp
objectMonitor_<os_family>.cpp oop.inline.hpp objectMonitor.cpp objectMonitor.inline.hpp
objectMonitor_<os_family>.cpp osThread.hpp objectMonitor.cpp oop.inline.hpp
objectMonitor_<os_family>.cpp os_<os_family>.inline.hpp objectMonitor.cpp osThread.hpp
objectMonitor_<os_family>.cpp threadService.hpp objectMonitor.cpp os_<os_family>.inline.hpp
objectMonitor_<os_family>.cpp thread_<os_family>.inline.hpp objectMonitor.cpp preserveException.hpp
objectMonitor_<os_family>.cpp vmSymbols.hpp objectMonitor.cpp resourceArea.hpp
objectMonitor.cpp stubRoutines.hpp
objectMonitor_<os_family>.hpp generate_platform_dependent_include objectMonitor.cpp thread.hpp
objectMonitor_<os_family>.hpp os_<os_family>.inline.hpp objectMonitor.cpp thread_<os_family>.inline.hpp
objectMonitor_<os_family>.hpp thread_<os_family>.inline.hpp objectMonitor.cpp threadService.hpp
objectMonitor_<os_family>.hpp top.hpp objectMonitor.cpp vmSymbols.hpp
objectMonitor_<os_family>.inline.hpp generate_platform_dependent_include objectMonitor.hpp os.hpp
objectMonitor.hpp perfData.hpp
oop.cpp copy.hpp oop.cpp copy.hpp
oop.cpp handles.inline.hpp oop.cpp handles.inline.hpp
...@@ -3329,7 +3337,6 @@ os_<os_family>.cpp mutex_<os_family>.inline.hpp ...@@ -3329,7 +3337,6 @@ os_<os_family>.cpp mutex_<os_family>.inline.hpp
os_<os_family>.cpp nativeInst_<arch>.hpp os_<os_family>.cpp nativeInst_<arch>.hpp
os_<os_family>.cpp no_precompiled_headers os_<os_family>.cpp no_precompiled_headers
os_<os_family>.cpp objectMonitor.hpp os_<os_family>.cpp objectMonitor.hpp
os_<os_family>.cpp objectMonitor.inline.hpp
os_<os_family>.cpp oop.inline.hpp os_<os_family>.cpp oop.inline.hpp
os_<os_family>.cpp osThread.hpp os_<os_family>.cpp osThread.hpp
os_<os_family>.cpp os_share_<os_family>.hpp os_<os_family>.cpp os_share_<os_family>.hpp
...@@ -3389,6 +3396,12 @@ ostream.cpp xmlstream.hpp ...@@ -3389,6 +3396,12 @@ ostream.cpp xmlstream.hpp
ostream.hpp allocation.hpp ostream.hpp allocation.hpp
ostream.hpp timer.hpp ostream.hpp timer.hpp
// include thread.hpp to prevent cyclic includes
park.cpp thread.hpp
park.hpp debug.hpp
park.hpp globalDefinitions.hpp
pcDesc.cpp debugInfoRec.hpp pcDesc.cpp debugInfoRec.hpp
pcDesc.cpp nmethod.hpp pcDesc.cpp nmethod.hpp
pcDesc.cpp pcDesc.hpp pcDesc.cpp pcDesc.hpp
...@@ -4063,10 +4076,10 @@ synchronizer.cpp preserveException.hpp ...@@ -4063,10 +4076,10 @@ synchronizer.cpp preserveException.hpp
synchronizer.cpp resourceArea.hpp synchronizer.cpp resourceArea.hpp
synchronizer.cpp stubRoutines.hpp synchronizer.cpp stubRoutines.hpp
synchronizer.cpp synchronizer.hpp synchronizer.cpp synchronizer.hpp
synchronizer.cpp threadService.hpp
synchronizer.cpp thread_<os_family>.inline.hpp synchronizer.cpp thread_<os_family>.inline.hpp
synchronizer.cpp vmSymbols.hpp synchronizer.cpp vmSymbols.hpp
synchronizer.hpp basicLock.hpp
synchronizer.hpp handles.hpp synchronizer.hpp handles.hpp
synchronizer.hpp markOop.hpp synchronizer.hpp markOop.hpp
synchronizer.hpp perfData.hpp synchronizer.hpp perfData.hpp
...@@ -4238,7 +4251,6 @@ thread.cpp memprofiler.hpp ...@@ -4238,7 +4251,6 @@ thread.cpp memprofiler.hpp
thread.cpp mutexLocker.hpp thread.cpp mutexLocker.hpp
thread.cpp objArrayOop.hpp thread.cpp objArrayOop.hpp
thread.cpp objectMonitor.hpp thread.cpp objectMonitor.hpp
thread.cpp objectMonitor.inline.hpp
thread.cpp oop.inline.hpp thread.cpp oop.inline.hpp
thread.cpp oopFactory.hpp thread.cpp oopFactory.hpp
thread.cpp osThread.hpp thread.cpp osThread.hpp
...@@ -4276,6 +4288,7 @@ thread.hpp mutexLocker.hpp ...@@ -4276,6 +4288,7 @@ thread.hpp mutexLocker.hpp
thread.hpp oop.hpp thread.hpp oop.hpp
thread.hpp os.hpp thread.hpp os.hpp
thread.hpp osThread.hpp thread.hpp osThread.hpp
thread.hpp park.hpp
thread.hpp safepoint.hpp thread.hpp safepoint.hpp
thread.hpp stubRoutines.hpp thread.hpp stubRoutines.hpp
thread.hpp threadLocalAllocBuffer.hpp thread.hpp threadLocalAllocBuffer.hpp
...@@ -4587,6 +4600,7 @@ vframeArray.hpp frame.inline.hpp ...@@ -4587,6 +4600,7 @@ vframeArray.hpp frame.inline.hpp
vframeArray.hpp growableArray.hpp vframeArray.hpp growableArray.hpp
vframeArray.hpp monitorChunk.hpp vframeArray.hpp monitorChunk.hpp
vframe_hp.cpp basicLock.hpp
vframe_hp.cpp codeCache.hpp vframe_hp.cpp codeCache.hpp
vframe_hp.cpp debugInfoRec.hpp vframe_hp.cpp debugInfoRec.hpp
vframe_hp.cpp handles.inline.hpp vframe_hp.cpp handles.inline.hpp
...@@ -4600,7 +4614,6 @@ vframe_hp.cpp pcDesc.hpp ...@@ -4600,7 +4614,6 @@ vframe_hp.cpp pcDesc.hpp
vframe_hp.cpp scopeDesc.hpp vframe_hp.cpp scopeDesc.hpp
vframe_hp.cpp signature.hpp vframe_hp.cpp signature.hpp
vframe_hp.cpp stubRoutines.hpp vframe_hp.cpp stubRoutines.hpp
vframe_hp.cpp synchronizer.hpp
vframe_hp.cpp vframeArray.hpp vframe_hp.cpp vframeArray.hpp
vframe_hp.cpp vframe_hp.hpp vframe_hp.cpp vframe_hp.hpp
...@@ -4752,6 +4765,7 @@ workgroup.cpp os.hpp ...@@ -4752,6 +4765,7 @@ workgroup.cpp os.hpp
workgroup.cpp workgroup.hpp workgroup.cpp workgroup.hpp
workgroup.hpp taskqueue.hpp workgroup.hpp taskqueue.hpp
workgroup.hpp thread_<os_family>.inline.hpp workgroup.hpp thread_<os_family>.inline.hpp
xmlstream.cpp allocation.hpp xmlstream.cpp allocation.hpp
......
src/share/vm/includeDB_features
浏览文件 @ f9bc752b
...@@ -184,6 +184,13 @@ jvmtiImpl.hpp stackValueCollection.hpp ...@@ -184,6 +184,13 @@ jvmtiImpl.hpp stackValueCollection.hpp
jvmtiImpl.hpp systemDictionary.hpp jvmtiImpl.hpp systemDictionary.hpp
jvmtiImpl.hpp vm_operations.hpp jvmtiImpl.hpp vm_operations.hpp
jvmtiRawMonitor.cpp interfaceSupport.hpp
jvmtiRawMonitor.cpp jvmtiRawMonitor.hpp
jvmtiRawMonitor.cpp thread.hpp
jvmtiRawMonitor.hpp growableArray.hpp
jvmtiRawMonitor.hpp objectMonitor.hpp
jvmtiTagMap.cpp biasedLocking.hpp jvmtiTagMap.cpp biasedLocking.hpp
jvmtiTagMap.cpp javaCalls.hpp jvmtiTagMap.cpp javaCalls.hpp
jvmtiTagMap.cpp jniHandles.hpp jvmtiTagMap.cpp jniHandles.hpp
......
src/share/vm/includeDB_jvmti
浏览文件 @ f9bc752b
...@@ -35,6 +35,7 @@ jvmtiClassFileReconstituter.hpp jvmtiEnv.hpp ...@@ -35,6 +35,7 @@ jvmtiClassFileReconstituter.hpp jvmtiEnv.hpp
// jvmtiCodeBlobEvents is jck optional, please put deps in includeDB_features // jvmtiCodeBlobEvents is jck optional, please put deps in includeDB_features
jvmtiEnter.cpp jvmtiEnter.hpp jvmtiEnter.cpp jvmtiEnter.hpp
jvmtiEnter.cpp jvmtiRawMonitor.hpp
jvmtiEnter.cpp jvmtiUtil.hpp jvmtiEnter.cpp jvmtiUtil.hpp
jvmtiEnter.hpp interfaceSupport.hpp jvmtiEnter.hpp interfaceSupport.hpp
...@@ -44,6 +45,7 @@ jvmtiEnter.hpp resourceArea.hpp ...@@ -44,6 +45,7 @@ jvmtiEnter.hpp resourceArea.hpp
jvmtiEnter.hpp systemDictionary.hpp jvmtiEnter.hpp systemDictionary.hpp
jvmtiEnterTrace.cpp jvmtiEnter.hpp jvmtiEnterTrace.cpp jvmtiEnter.hpp
jvmtiEnterTrace.cpp jvmtiRawMonitor.hpp
jvmtiEnterTrace.cpp jvmtiUtil.hpp jvmtiEnterTrace.cpp jvmtiUtil.hpp
jvmtiEnv.cpp arguments.hpp jvmtiEnv.cpp arguments.hpp
...@@ -66,11 +68,11 @@ jvmtiEnv.cpp jvmtiExtensions.hpp ...@@ -66,11 +68,11 @@ jvmtiEnv.cpp jvmtiExtensions.hpp
jvmtiEnv.cpp jvmtiGetLoadedClasses.hpp jvmtiEnv.cpp jvmtiGetLoadedClasses.hpp
jvmtiEnv.cpp jvmtiImpl.hpp jvmtiEnv.cpp jvmtiImpl.hpp
jvmtiEnv.cpp jvmtiManageCapabilities.hpp jvmtiEnv.cpp jvmtiManageCapabilities.hpp
jvmtiEnv.cpp jvmtiRawMonitor.hpp
jvmtiEnv.cpp jvmtiRedefineClasses.hpp jvmtiEnv.cpp jvmtiRedefineClasses.hpp
jvmtiEnv.cpp jvmtiTagMap.hpp jvmtiEnv.cpp jvmtiTagMap.hpp
jvmtiEnv.cpp jvmtiThreadState.inline.hpp jvmtiEnv.cpp jvmtiThreadState.inline.hpp
jvmtiEnv.cpp jvmtiUtil.hpp jvmtiEnv.cpp jvmtiUtil.hpp
jvmtiEnv.cpp objectMonitor.inline.hpp
jvmtiEnv.cpp osThread.hpp jvmtiEnv.cpp osThread.hpp
jvmtiEnv.cpp preserveException.hpp jvmtiEnv.cpp preserveException.hpp
jvmtiEnv.cpp reflectionUtils.hpp jvmtiEnv.cpp reflectionUtils.hpp
...@@ -178,11 +180,13 @@ jvmtiExport.cpp jvmtiEventController.inline.hpp ...@@ -178,11 +180,13 @@ jvmtiExport.cpp jvmtiEventController.inline.hpp
jvmtiExport.cpp jvmtiExport.hpp jvmtiExport.cpp jvmtiExport.hpp
jvmtiExport.cpp jvmtiImpl.hpp jvmtiExport.cpp jvmtiImpl.hpp
jvmtiExport.cpp jvmtiManageCapabilities.hpp jvmtiExport.cpp jvmtiManageCapabilities.hpp
jvmtiExport.cpp jvmtiRawMonitor.hpp
jvmtiExport.cpp jvmtiTagMap.hpp jvmtiExport.cpp jvmtiTagMap.hpp
jvmtiExport.cpp jvmtiThreadState.inline.hpp jvmtiExport.cpp jvmtiThreadState.inline.hpp
jvmtiExport.cpp nmethod.hpp jvmtiExport.cpp nmethod.hpp
jvmtiExport.cpp objArrayKlass.hpp jvmtiExport.cpp objArrayKlass.hpp
jvmtiExport.cpp objArrayOop.hpp jvmtiExport.cpp objArrayOop.hpp
jvmtiExport.cpp objectMonitor.hpp
jvmtiExport.cpp objectMonitor.inline.hpp jvmtiExport.cpp objectMonitor.inline.hpp
jvmtiExport.cpp pcDesc.hpp jvmtiExport.cpp pcDesc.hpp
jvmtiExport.cpp resourceArea.hpp jvmtiExport.cpp resourceArea.hpp
...@@ -210,6 +214,8 @@ jvmtiManageCapabilities.cpp jvmtiManageCapabilities.hpp ...@@ -210,6 +214,8 @@ jvmtiManageCapabilities.cpp jvmtiManageCapabilities.hpp
jvmtiManageCapabilities.hpp allocation.hpp jvmtiManageCapabilities.hpp allocation.hpp
jvmtiManageCapabilities.hpp jvmti.h jvmtiManageCapabilities.hpp jvmti.h
// jvmtiRawMonitor is jck optional, please put deps in includeDB_features
jvmtiRedefineClasses.cpp bitMap.inline.hpp jvmtiRedefineClasses.cpp bitMap.inline.hpp
jvmtiRedefineClasses.cpp codeCache.hpp jvmtiRedefineClasses.cpp codeCache.hpp
jvmtiRedefineClasses.cpp deoptimization.hpp jvmtiRedefineClasses.cpp deoptimization.hpp
......
src/share/vm/prims/jvmtiImpl.cpp
浏览文件 @ f9bc752b
...@@ -25,26 +25,6 @@ ...@@ -25,26 +25,6 @@
# include "incls/_precompiled.incl" # include "incls/_precompiled.incl"
# include "incls/_jvmtiImpl.cpp.incl" # include "incls/_jvmtiImpl.cpp.incl"
GrowableArray<JvmtiRawMonitor*> *JvmtiPendingMonitors::_monitors = new (ResourceObj::C_HEAP) GrowableArray<JvmtiRawMonitor*>(1,true);
void JvmtiPendingMonitors::transition_raw_monitors() {
assert((Threads::number_of_threads()==1),
"Java thread has not created yet or more than one java thread \
is running. Raw monitor transition will not work");
JavaThread *current_java_thread = JavaThread::current();
assert(current_java_thread->thread_state() == _thread_in_vm, "Must be in vm");
{
ThreadBlockInVM __tbivm(current_java_thread);
for(int i=0; i< count(); i++) {
JvmtiRawMonitor *rmonitor = monitors()->at(i);
int r = rmonitor->raw_enter(current_java_thread);
assert(r == ObjectMonitor::OM_OK, "raw_enter should have worked");
}
}
// pending monitors are converted to real monitor so delete them all.
dispose();
}
// //
// class JvmtiAgentThread // class JvmtiAgentThread
// //
...@@ -216,57 +196,6 @@ void GrowableCache::gc_epilogue() { ...@@ -216,57 +196,6 @@ void GrowableCache::gc_epilogue() {
} }
} }
//
// class JvmtiRawMonitor
//
JvmtiRawMonitor::JvmtiRawMonitor(const char *name) {
#ifdef ASSERT
_name = strcpy(NEW_C_HEAP_ARRAY(char, strlen(name) + 1), name);
#else
_name = NULL;
#endif
_magic = JVMTI_RM_MAGIC;
}
JvmtiRawMonitor::~JvmtiRawMonitor() {
#ifdef ASSERT
FreeHeap(_name);
#endif
_magic = 0;
}
bool
JvmtiRawMonitor::is_valid() {
int value = 0;
// This object might not be a JvmtiRawMonitor so we can't assume
// the _magic field is properly aligned. Get the value in a safe
// way and then check against JVMTI_RM_MAGIC.
switch (sizeof(_magic)) {
case 2:
value = Bytes::get_native_u2((address)&_magic);
break;
case 4:
value = Bytes::get_native_u4((address)&_magic);
break;
case 8:
value = Bytes::get_native_u8((address)&_magic);
break;
default:
guarantee(false, "_magic field is an unexpected size");
}
return value == JVMTI_RM_MAGIC;
}
// //
// class JvmtiBreakpoint // class JvmtiBreakpoint
// //
......
src/share/vm/prims/jvmtiImpl.hpp
浏览文件 @ f9bc752b
...@@ -26,7 +26,6 @@ ...@@ -26,7 +26,6 @@
// Forward Declarations // Forward Declarations
// //
class JvmtiRawMonitor;
class JvmtiBreakpoint; class JvmtiBreakpoint;
class JvmtiBreakpoints; class JvmtiBreakpoints;
...@@ -327,76 +326,6 @@ bool JvmtiCurrentBreakpoints::is_breakpoint(address bcp) { ...@@ -327,76 +326,6 @@ bool JvmtiCurrentBreakpoints::is_breakpoint(address bcp) {
return false; return false;
} }
///////////////////////////////////////////////////////////////
//
// class JvmtiRawMonitor
//
// Used by JVMTI methods: All RawMonitor methods (CreateRawMonitor, EnterRawMonitor, etc.)
//
// Wrapper for ObjectMonitor class that saves the Monitor's name
//
class JvmtiRawMonitor : public ObjectMonitor {
private:
int _magic;
char * _name;
// JVMTI_RM_MAGIC is set in contructor and unset in destructor.
enum { JVMTI_RM_MAGIC = (int)(('T' << 24) | ('I' << 16) | ('R' << 8) | 'M') };
public:
JvmtiRawMonitor(const char *name);
~JvmtiRawMonitor();
int magic() { return _magic; }
const char *get_name() { return _name; }
bool is_valid();
};
// Onload pending raw monitors
// Class is used to cache onload or onstart monitor enter
// which will transition into real monitor when
// VM is fully initialized.
class JvmtiPendingMonitors : public AllStatic {
private:
static GrowableArray<JvmtiRawMonitor*> *_monitors; // Cache raw monitor enter
inline static GrowableArray<JvmtiRawMonitor*>* monitors() { return _monitors; }
static void dispose() {
delete monitors();
}
public:
static void enter(JvmtiRawMonitor *monitor) {
monitors()->append(monitor);
}
static int count() {
return monitors()->length();
}
static void destroy(JvmtiRawMonitor *monitor) {
while (monitors()->contains(monitor)) {
monitors()->remove(monitor);
}
}
// Return false if monitor is not found in the list.
static bool exit(JvmtiRawMonitor *monitor) {
if (monitors()->contains(monitor)) {
monitors()->remove(monitor);
return true;
} else {
return false;
}
}
static void transition_raw_monitors();
};
/////////////////////////////////////////////////////////////// ///////////////////////////////////////////////////////////////
// The get/set local operations must only be done by the VM thread // The get/set local operations must only be done by the VM thread
// because the interpreter version needs to access oop maps, which can // because the interpreter version needs to access oop maps, which can
......
src/share/vm/prims/jvmtiRawMonitor.cpp 0 → 100644
浏览文件 @ f9bc752b
/*
* Copyright (c) 2003, 2007, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
# include "incls/_precompiled.incl"
# include "incls/_jvmtiRawMonitor.cpp.incl"
GrowableArray<JvmtiRawMonitor*> *JvmtiPendingMonitors::_monitors = new (ResourceObj::C_HEAP) GrowableArray<JvmtiRawMonitor*>(1,true);
void JvmtiPendingMonitors::transition_raw_monitors() {
assert((Threads::number_of_threads()==1),
"Java thread has not created yet or more than one java thread \
is running. Raw monitor transition will not work");
JavaThread *current_java_thread = JavaThread::current();
assert(current_java_thread->thread_state() == _thread_in_vm, "Must be in vm");
{
ThreadBlockInVM __tbivm(current_java_thread);
for(int i=0; i< count(); i++) {
JvmtiRawMonitor *rmonitor = monitors()->at(i);
int r = rmonitor->raw_enter(current_java_thread);
assert(r == ObjectMonitor::OM_OK, "raw_enter should have worked");
}
}
// pending monitors are converted to real monitor so delete them all.
dispose();
}
//
// class JvmtiRawMonitor
//
JvmtiRawMonitor::JvmtiRawMonitor(const char *name) {
#ifdef ASSERT
_name = strcpy(NEW_C_HEAP_ARRAY(char, strlen(name) + 1), name);
#else
_name = NULL;
#endif
_magic = JVMTI_RM_MAGIC;
}
JvmtiRawMonitor::~JvmtiRawMonitor() {
#ifdef ASSERT
FreeHeap(_name);
#endif
_magic = 0;
}
bool
JvmtiRawMonitor::is_valid() {
int value = 0;
// This object might not be a JvmtiRawMonitor so we can't assume
// the _magic field is properly aligned. Get the value in a safe
// way and then check against JVMTI_RM_MAGIC.
switch (sizeof(_magic)) {
case 2:
value = Bytes::get_native_u2((address)&_magic);
break;
case 4:
value = Bytes::get_native_u4((address)&_magic);
break;
case 8:
value = Bytes::get_native_u8((address)&_magic);
break;
default:
guarantee(false, "_magic field is an unexpected size");
}
return value == JVMTI_RM_MAGIC;
}
// -------------------------------------------------------------------------
// The raw monitor subsystem is entirely distinct from normal
// java-synchronization or jni-synchronization. raw monitors are not
// associated with objects. They can be implemented in any manner
// that makes sense. The original implementors decided to piggy-back
// the raw-monitor implementation on the existing Java objectMonitor mechanism.
// This flaw needs to fixed. We should reimplement raw monitors as sui-generis.
// Specifically, we should not implement raw monitors via java monitors.
// Time permitting, we should disentangle and deconvolve the two implementations
// and move the resulting raw monitor implementation over to the JVMTI directories.
// Ideally, the raw monitor implementation would be built on top of
// park-unpark and nothing else.
//
// raw monitors are used mainly by JVMTI
// The raw monitor implementation borrows the ObjectMonitor structure,
// but the operators are degenerate and extremely simple.
//
// Mixed use of a single objectMonitor instance -- as both a raw monitor
// and a normal java monitor -- is not permissible.
//
// Note that we use the single RawMonitor_lock to protect queue operations for
// _all_ raw monitors. This is a scalability impediment, but since raw monitor usage
// is deprecated and rare, this is not of concern. The RawMonitor_lock can not
// be held indefinitely. The critical sections must be short and bounded.
//
// -------------------------------------------------------------------------
int JvmtiRawMonitor::SimpleEnter (Thread * Self) {
for (;;) {
if (Atomic::cmpxchg_ptr (Self, &_owner, NULL) == NULL) {
return OS_OK ;
}
ObjectWaiter Node (Self) ;
Self->_ParkEvent->reset() ; // strictly optional
Node.TState = ObjectWaiter::TS_ENTER ;
RawMonitor_lock->lock_without_safepoint_check() ;
Node._next = _EntryList ;
_EntryList = &Node ;
OrderAccess::fence() ;
if (_owner == NULL && Atomic::cmpxchg_ptr (Self, &_owner, NULL) == NULL) {
_EntryList = Node._next ;
RawMonitor_lock->unlock() ;
return OS_OK ;
}
RawMonitor_lock->unlock() ;
while (Node.TState == ObjectWaiter::TS_ENTER) {
Self->_ParkEvent->park() ;
}
}
}
int JvmtiRawMonitor::SimpleExit (Thread * Self) {
guarantee (_owner == Self, "invariant") ;
OrderAccess::release_store_ptr (&_owner, NULL) ;
OrderAccess::fence() ;
if (_EntryList == NULL) return OS_OK ;
ObjectWaiter * w ;
RawMonitor_lock->lock_without_safepoint_check() ;
w = _EntryList ;
if (w != NULL) {
_EntryList = w->_next ;
}
RawMonitor_lock->unlock() ;
if (w != NULL) {
guarantee (w ->TState == ObjectWaiter::TS_ENTER, "invariant") ;
ParkEvent * ev = w->_event ;
w->TState = ObjectWaiter::TS_RUN ;
OrderAccess::fence() ;
ev->unpark() ;
}
return OS_OK ;
}
int JvmtiRawMonitor::SimpleWait (Thread * Self, jlong millis) {
guarantee (_owner == Self , "invariant") ;
guarantee (_recursions == 0, "invariant") ;
ObjectWaiter Node (Self) ;
Node._notified = 0 ;
Node.TState = ObjectWaiter::TS_WAIT ;
RawMonitor_lock->lock_without_safepoint_check() ;
Node._next = _WaitSet ;
_WaitSet = &Node ;
RawMonitor_lock->unlock() ;
SimpleExit (Self) ;
guarantee (_owner != Self, "invariant") ;
int ret = OS_OK ;
if (millis <= 0) {
Self->_ParkEvent->park();
} else {
ret = Self->_ParkEvent->park(millis);
}
// If thread still resides on the waitset then unlink it.
// Double-checked locking -- the usage is safe in this context
// as we TState is volatile and the lock-unlock operators are
// serializing (barrier-equivalent).
if (Node.TState == ObjectWaiter::TS_WAIT) {
RawMonitor_lock->lock_without_safepoint_check() ;
if (Node.TState == ObjectWaiter::TS_WAIT) {
// Simple O(n) unlink, but performance isn't critical here.
ObjectWaiter * p ;
ObjectWaiter * q = NULL ;
for (p = _WaitSet ; p != &Node; p = p->_next) {
q = p ;
}
guarantee (p == &Node, "invariant") ;
if (q == NULL) {
guarantee (p == _WaitSet, "invariant") ;
_WaitSet = p->_next ;
} else {
guarantee (p == q->_next, "invariant") ;
q->_next = p->_next ;
}
Node.TState = ObjectWaiter::TS_RUN ;
}
RawMonitor_lock->unlock() ;
}
guarantee (Node.TState == ObjectWaiter::TS_RUN, "invariant") ;
SimpleEnter (Self) ;
guarantee (_owner == Self, "invariant") ;
guarantee (_recursions == 0, "invariant") ;
return ret ;
}
int JvmtiRawMonitor::SimpleNotify (Thread * Self, bool All) {
guarantee (_owner == Self, "invariant") ;
if (_WaitSet == NULL) return OS_OK ;
// We have two options:
// A. Transfer the threads from the WaitSet to the EntryList
// B. Remove the thread from the WaitSet and unpark() it.
//
// We use (B), which is crude and results in lots of futile
// context switching. In particular (B) induces lots of contention.
ParkEvent * ev = NULL ; // consider using a small auto array ...
RawMonitor_lock->lock_without_safepoint_check() ;
for (;;) {
ObjectWaiter * w = _WaitSet ;
if (w == NULL) break ;
_WaitSet = w->_next ;
if (ev != NULL) { ev->unpark(); ev = NULL; }
ev = w->_event ;
OrderAccess::loadstore() ;
w->TState = ObjectWaiter::TS_RUN ;
OrderAccess::storeload();
if (!All) break ;
}
RawMonitor_lock->unlock() ;
if (ev != NULL) ev->unpark();
return OS_OK ;
}
// Any JavaThread will enter here with state _thread_blocked
int JvmtiRawMonitor::raw_enter(TRAPS) {
TEVENT (raw_enter) ;
void * Contended ;
// don't enter raw monitor if thread is being externally suspended, it will
// surprise the suspender if a "suspended" thread can still enter monitor
JavaThread * jt = (JavaThread *)THREAD;
if (THREAD->is_Java_thread()) {
jt->SR_lock()->lock_without_safepoint_check();
while (jt->is_external_suspend()) {
jt->SR_lock()->unlock();
jt->java_suspend_self();
jt->SR_lock()->lock_without_safepoint_check();
}
// guarded by SR_lock to avoid racing with new external suspend requests.
Contended = Atomic::cmpxchg_ptr (THREAD, &_owner, NULL) ;
jt->SR_lock()->unlock();
} else {
Contended = Atomic::cmpxchg_ptr (THREAD, &_owner, NULL) ;
}
if (Contended == THREAD) {
_recursions ++ ;
return OM_OK ;
}
if (Contended == NULL) {
guarantee (_owner == THREAD, "invariant") ;
guarantee (_recursions == 0, "invariant") ;
return OM_OK ;
}
THREAD->set_current_pending_monitor(this);
if (!THREAD->is_Java_thread()) {
// No other non-Java threads besides VM thread would acquire
// a raw monitor.
assert(THREAD->is_VM_thread(), "must be VM thread");
SimpleEnter (THREAD) ;
} else {
guarantee (jt->thread_state() == _thread_blocked, "invariant") ;
for (;;) {
jt->set_suspend_equivalent();
// cleared by handle_special_suspend_equivalent_condition() or
// java_suspend_self()
SimpleEnter (THREAD) ;
// were we externally suspended while we were waiting?
if (!jt->handle_special_suspend_equivalent_condition()) break ;
// This thread was externally suspended
//
// This logic isn't needed for JVMTI raw monitors,
// but doesn't hurt just in case the suspend rules change. This
// logic is needed for the JvmtiRawMonitor.wait() reentry phase.
// We have reentered the contended monitor, but while we were
// waiting another thread suspended us. We don't want to reenter
// the monitor while suspended because that would surprise the
// thread that suspended us.
//
// Drop the lock -
SimpleExit (THREAD) ;
jt->java_suspend_self();
}
assert(_owner == THREAD, "Fatal error with monitor owner!");
assert(_recursions == 0, "Fatal error with monitor recursions!");
}
THREAD->set_current_pending_monitor(NULL);
guarantee (_recursions == 0, "invariant") ;
return OM_OK;
}
// Used mainly for JVMTI raw monitor implementation
// Also used for JvmtiRawMonitor::wait().
int JvmtiRawMonitor::raw_exit(TRAPS) {
TEVENT (raw_exit) ;
if (THREAD != _owner) {
return OM_ILLEGAL_MONITOR_STATE;
}
if (_recursions > 0) {
--_recursions ;
return OM_OK ;
}
void * List = _EntryList ;
SimpleExit (THREAD) ;
return OM_OK;
}
// Used for JVMTI raw monitor implementation.
// All JavaThreads will enter here with state _thread_blocked
int JvmtiRawMonitor::raw_wait(jlong millis, bool interruptible, TRAPS) {
TEVENT (raw_wait) ;
if (THREAD != _owner) {
return OM_ILLEGAL_MONITOR_STATE;
}
// To avoid spurious wakeups we reset the parkevent -- This is strictly optional.
// The caller must be able to tolerate spurious returns from raw_wait().
THREAD->_ParkEvent->reset() ;
OrderAccess::fence() ;
// check interrupt event
if (interruptible && Thread::is_interrupted(THREAD, true)) {
return OM_INTERRUPTED;
}
intptr_t save = _recursions ;
_recursions = 0 ;
_waiters ++ ;
if (THREAD->is_Java_thread()) {
guarantee (((JavaThread *) THREAD)->thread_state() == _thread_blocked, "invariant") ;
((JavaThread *)THREAD)->set_suspend_equivalent();
}
int rv = SimpleWait (THREAD, millis) ;
_recursions = save ;
_waiters -- ;
guarantee (THREAD == _owner, "invariant") ;
if (THREAD->is_Java_thread()) {
JavaThread * jSelf = (JavaThread *) THREAD ;
for (;;) {
if (!jSelf->handle_special_suspend_equivalent_condition()) break ;
SimpleExit (THREAD) ;
jSelf->java_suspend_self();
SimpleEnter (THREAD) ;
jSelf->set_suspend_equivalent() ;
}
}
guarantee (THREAD == _owner, "invariant") ;
if (interruptible && Thread::is_interrupted(THREAD, true)) {
return OM_INTERRUPTED;
}
return OM_OK ;
}
int JvmtiRawMonitor::raw_notify(TRAPS) {
TEVENT (raw_notify) ;
if (THREAD != _owner) {
return OM_ILLEGAL_MONITOR_STATE;
}
SimpleNotify (THREAD, false) ;
return OM_OK;
}
int JvmtiRawMonitor::raw_notifyAll(TRAPS) {
TEVENT (raw_notifyAll) ;
if (THREAD != _owner) {
return OM_ILLEGAL_MONITOR_STATE;
}
SimpleNotify (THREAD, true) ;
return OM_OK;
}
src/os/linux/vm/objectMonitor_linux.cpp → src/share/vm/prims/jvmtiRawMonitor.hpp
浏览文件 @ f9bc752b
/* /*
* Copyright (c) 1999, 2005, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 1999, 2007, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
* *
* This code is free software; you can redistribute it and/or modify it * This code is free software; you can redistribute it and/or modify it
...@@ -22,3 +21,79 @@ ...@@ -22,3 +21,79 @@
* questions. * questions.
* *
*/ */
//
// class JvmtiRawMonitor
//
// Used by JVMTI methods: All RawMonitor methods (CreateRawMonitor, EnterRawMonitor, etc.)
//
// Wrapper for ObjectMonitor class that saves the Monitor's name
//
class JvmtiRawMonitor : public ObjectMonitor {
private:
int _magic;
char * _name;
// JVMTI_RM_MAGIC is set in contructor and unset in destructor.
enum { JVMTI_RM_MAGIC = (int)(('T' << 24) | ('I' << 16) | ('R' << 8) | 'M') };
int SimpleEnter (Thread * Self) ;
int SimpleExit (Thread * Self) ;
int SimpleWait (Thread * Self, jlong millis) ;
int SimpleNotify (Thread * Self, bool All) ;
public:
JvmtiRawMonitor(const char *name);
~JvmtiRawMonitor();
int raw_enter(TRAPS);
int raw_exit(TRAPS);
int raw_wait(jlong millis, bool interruptable, TRAPS);
int raw_notify(TRAPS);
int raw_notifyAll(TRAPS);
int magic() { return _magic; }
const char *get_name() { return _name; }
bool is_valid();
};
// Onload pending raw monitors
// Class is used to cache onload or onstart monitor enter
// which will transition into real monitor when
// VM is fully initialized.
class JvmtiPendingMonitors : public AllStatic {
private:
static GrowableArray<JvmtiRawMonitor*> *_monitors; // Cache raw monitor enter
inline static GrowableArray<JvmtiRawMonitor*>* monitors() { return _monitors; }
static void dispose() {
delete monitors();
}
public:
static void enter(JvmtiRawMonitor *monitor) {
monitors()->append(monitor);
}
static int count() {
return monitors()->length();
}
static void destroy(JvmtiRawMonitor *monitor) {
while (monitors()->contains(monitor)) {
monitors()->remove(monitor);
}
}
// Return false if monitor is not found in the list.
static bool exit(JvmtiRawMonitor *monitor) {
if (monitors()->contains(monitor)) {
monitors()->remove(monitor);
return true;
} else {
return false;
}
}
static void transition_raw_monitors();
};
src/share/vm/runtime/basicLock.cpp 0 → 100644
浏览文件 @ f9bc752b
/*
* Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
# include "incls/_precompiled.incl"
# include "incls/_basicLock.cpp.incl"
void BasicLock::print_on(outputStream* st) const {
st->print("monitor");
}
void BasicLock::move_to(oop obj, BasicLock* dest) {
// Check to see if we need to inflate the lock. This is only needed
// if an object is locked using "this" lightweight monitor. In that
// case, the displaced_header() is unlocked, because the
// displaced_header() contains the header for the originally unlocked
// object. However the object could have already been inflated. But it
// does not matter, the inflation will just a no-op. For other cases,
// the displaced header will be either 0x0 or 0x3, which are location
// independent, therefore the BasicLock is free to move.
//
// During OSR we may need to relocate a BasicLock (which contains a
// displaced word) from a location in an interpreter frame to a
// new location in a compiled frame. "this" refers to the source
// basiclock in the interpreter frame. "dest" refers to the destination
// basiclock in the new compiled frame. We *always* inflate in move_to().
// The always-Inflate policy works properly, but in 1.5.0 it can sometimes
// cause performance problems in code that makes heavy use of a small # of
// uncontended locks. (We'd inflate during OSR, and then sync performance
// would subsequently plummet because the thread would be forced thru the slow-path).
// This problem has been made largely moot on IA32 by inlining the inflated fast-path
// operations in Fast_Lock and Fast_Unlock in i486.ad.
//
// Note that there is a way to safely swing the object's markword from
// one stack location to another. This avoids inflation. Obviously,
// we need to ensure that both locations refer to the current thread's stack.
// There are some subtle concurrency issues, however, and since the benefit is
// is small (given the support for inflated fast-path locking in the fast_lock, etc)
// we'll leave that optimization for another time.
if (displaced_header()->is_neutral()) {
ObjectSynchronizer::inflate_helper(obj);
// WARNING: We can not put check here, because the inflation
// will not update the displaced header. Once BasicLock is inflated,
// no one should ever look at its content.
} else {
// Typically the displaced header will be 0 (recursive stack lock) or
// unused_mark. Naively we'd like to assert that the displaced mark
// value is either 0, neutral, or 3. But with the advent of the
// store-before-CAS avoidance in fast_lock/compiler_lock_object
// we can find any flavor mark in the displaced mark.
}
// [RGV] The next line appears to do nothing!
intptr_t dh = (intptr_t) displaced_header();
dest->set_displaced_header(displaced_header());
}
src/os/windows/vm/objectMonitor_windows.hpp → src/share/vm/runtime/basicLock.hpp
浏览文件 @ f9bc752b
/* /*
* Copyright (c) 1998, 2005, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 1998, 2007, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
* *
* This code is free software; you can redistribute it and/or modify it * This code is free software; you can redistribute it and/or modify it
...@@ -22,4 +22,51 @@ ...@@ -22,4 +22,51 @@
* *
*/ */
class BasicLock VALUE_OBJ_CLASS_SPEC {
friend class VMStructs;
private: private:
volatile markOop _displaced_header;
public:
markOop displaced_header() const { return _displaced_header; }
void set_displaced_header(markOop header) { _displaced_header = header; }
void print_on(outputStream* st) const;
// move a basic lock (used during deoptimization
void move_to(oop obj, BasicLock* dest);
static int displaced_header_offset_in_bytes() { return offset_of(BasicLock, _displaced_header); }
};
// A BasicObjectLock associates a specific Java object with a BasicLock.
// It is currently embedded in an interpreter frame.
// Because some machines have alignment restrictions on the control stack,
// the actual space allocated by the interpreter may include padding words
// after the end of the BasicObjectLock. Also, in order to guarantee
// alignment of the embedded BasicLock objects on such machines, we
// put the embedded BasicLock at the beginning of the struct.
class BasicObjectLock VALUE_OBJ_CLASS_SPEC {
friend class VMStructs;
private:
BasicLock _lock; // the lock, must be double word aligned
oop _obj; // object holds the lock;
public:
// Manipulation
oop obj() const { return _obj; }
void set_obj(oop obj) { _obj = obj; }
BasicLock* lock() { return &_lock; }
// Note: Use frame::interpreter_frame_monitor_size() for the size of BasicObjectLocks
// in interpreter activation frames since it includes machine-specific padding.
static int size() { return sizeof(BasicObjectLock)/wordSize; }
// GC support
void oops_do(OopClosure* f) { f->do_oop(&_obj); }
static int obj_offset_in_bytes() { return offset_of(BasicObjectLock, _obj); }
static int lock_offset_in_bytes() { return offset_of(BasicObjectLock, _lock); }
};
src/share/vm/runtime/mutex.hpp
浏览文件 @ f9bc752b
...@@ -265,48 +265,3 @@ class Mutex : public Monitor { // degenerate Monitor ...@@ -265,48 +265,3 @@ class Mutex : public Monitor { // degenerate Monitor
} }
}; };
/*
* Per-thread blocking support for JSR166. See the Java-level
* Documentation for rationale. Basically, park acts like wait, unpark
* like notify.
*
* 6271289 --
* To avoid errors where an os thread expires but the JavaThread still
* exists, Parkers are immortal (type-stable) and are recycled across
* new threads. This parallels the ParkEvent implementation.
* Because park-unpark allow spurious wakeups it is harmless if an
* unpark call unparks a new thread using the old Parker reference.
*
* In the future we'll want to think about eliminating Parker and using
* ParkEvent instead. There's considerable duplication between the two
* services.
*
*/
class Parker : public os::PlatformParker {
private:
volatile int _counter ;
Parker * FreeNext ;
JavaThread * AssociatedWith ; // Current association
public:
Parker() : PlatformParker() {
_counter = 0 ;
FreeNext = NULL ;
AssociatedWith = NULL ;
}
protected:
~Parker() { ShouldNotReachHere(); }
public:
// For simplicity of interface with Java, all forms of park (indefinite,
// relative, and absolute) are multiplexed into one call.
void park(bool isAbsolute, jlong time);
void unpark();
// Lifecycle operators
static Parker * Allocate (JavaThread * t) ;
static void Release (Parker * e) ;
private:
static Parker * volatile FreeList ;
static volatile int ListLock ;
};
src/share/vm/runtime/objectMonitor.cpp 0 → 100644
浏览文件 @ f9bc752b
/*
* Copyright (c) 1998, 2009, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
# include "incls/_precompiled.incl"
# include "incls/_objectMonitor.cpp.incl"
#if defined(__GNUC__) && !defined(IA64)
// Need to inhibit inlining for older versions of GCC to avoid build-time failures
#define ATTR __attribute__((noinline))
#else
#define ATTR
#endif
#ifdef DTRACE_ENABLED
// Only bother with this argument setup if dtrace is available
// TODO-FIXME: probes should not fire when caller is _blocked. assert() accordingly.
HS_DTRACE_PROBE_DECL4(hotspot, monitor__notify,
jlong, uintptr_t, char*, int);
HS_DTRACE_PROBE_DECL4(hotspot, monitor__notifyAll,
jlong, uintptr_t, char*, int);
HS_DTRACE_PROBE_DECL4(hotspot, monitor__contended__enter,
jlong, uintptr_t, char*, int);
HS_DTRACE_PROBE_DECL4(hotspot, monitor__contended__entered,
jlong, uintptr_t, char*, int);
HS_DTRACE_PROBE_DECL4(hotspot, monitor__contended__exit,
jlong, uintptr_t, char*, int);
#define DTRACE_MONITOR_PROBE_COMMON(klassOop, thread) \
char* bytes = NULL; \
int len = 0; \
jlong jtid = SharedRuntime::get_java_tid(thread); \
symbolOop klassname = ((oop)(klassOop))->klass()->klass_part()->name(); \
if (klassname != NULL) { \
bytes = (char*)klassname->bytes(); \
len = klassname->utf8_length(); \
}
#define DTRACE_MONITOR_WAIT_PROBE(monitor, klassOop, thread, millis) \
{ \
if (DTraceMonitorProbes) { \
DTRACE_MONITOR_PROBE_COMMON(klassOop, thread); \
HS_DTRACE_PROBE5(hotspot, monitor__wait, jtid, \
(monitor), bytes, len, (millis)); \
} \
}
#define DTRACE_MONITOR_PROBE(probe, monitor, klassOop, thread) \
{ \
if (DTraceMonitorProbes) { \
DTRACE_MONITOR_PROBE_COMMON(klassOop, thread); \
HS_DTRACE_PROBE4(hotspot, monitor__##probe, jtid, \
(uintptr_t)(monitor), bytes, len); \
} \
}
#else // ndef DTRACE_ENABLED
#define DTRACE_MONITOR_WAIT_PROBE(klassOop, thread, millis, mon) {;}
#define DTRACE_MONITOR_PROBE(probe, klassOop, thread, mon) {;}
#endif // ndef DTRACE_ENABLED
// Tunables ...
// The knob* variables are effectively final. Once set they should
// never be modified hence. Consider using __read_mostly with GCC.
int ObjectMonitor::Knob_Verbose = 0 ;
int ObjectMonitor::Knob_SpinLimit = 5000 ; // derived by an external tool -
static int Knob_LogSpins = 0 ; // enable jvmstat tally for spins
static int Knob_HandOff = 0 ;
static int Knob_ReportSettings = 0 ;
static int Knob_SpinBase = 0 ; // Floor AKA SpinMin
static int Knob_SpinBackOff = 0 ; // spin-loop backoff
static int Knob_CASPenalty = -1 ; // Penalty for failed CAS
static int Knob_OXPenalty = -1 ; // Penalty for observed _owner change
static int Knob_SpinSetSucc = 1 ; // spinners set the _succ field
static int Knob_SpinEarly = 1 ;
static int Knob_SuccEnabled = 1 ; // futile wake throttling
static int Knob_SuccRestrict = 0 ; // Limit successors + spinners to at-most-one
static int Knob_MaxSpinners = -1 ; // Should be a function of # CPUs
static int Knob_Bonus = 100 ; // spin success bonus
static int Knob_BonusB = 100 ; // spin success bonus
static int Knob_Penalty = 200 ; // spin failure penalty
static int Knob_Poverty = 1000 ;
static int Knob_SpinAfterFutile = 1 ; // Spin after returning from park()
static int Knob_FixedSpin = 0 ;
static int Knob_OState = 3 ; // Spinner checks thread state of _owner
static int Knob_UsePause = 1 ;
static int Knob_ExitPolicy = 0 ;
static int Knob_PreSpin = 10 ; // 20-100 likely better
static int Knob_ResetEvent = 0 ;
static int BackOffMask = 0 ;
static int Knob_FastHSSEC = 0 ;
static int Knob_MoveNotifyee = 2 ; // notify() - disposition of notifyee
static int Knob_QMode = 0 ; // EntryList-cxq policy - queue discipline
static volatile int InitDone = 0 ;
#define TrySpin TrySpin_VaryDuration
// -----------------------------------------------------------------------------
// Theory of operations -- Monitors lists, thread residency, etc:
//
// * A thread acquires ownership of a monitor by successfully
// CAS()ing the _owner field from null to non-null.
//
// * Invariant: A thread appears on at most one monitor list --
// cxq, EntryList or WaitSet -- at any one time.
//
// * Contending threads "push" themselves onto the cxq with CAS
// and then spin/park.
//
// * After a contending thread eventually acquires the lock it must
// dequeue itself from either the EntryList or the cxq.
//
// * The exiting thread identifies and unparks an "heir presumptive"
// tentative successor thread on the EntryList. Critically, the
// exiting thread doesn't unlink the successor thread from the EntryList.
// After having been unparked, the wakee will recontend for ownership of
// the monitor. The successor (wakee) will either acquire the lock or
// re-park itself.
//
// Succession is provided for by a policy of competitive handoff.
// The exiting thread does _not_ grant or pass ownership to the
// successor thread. (This is also referred to as "handoff" succession").
// Instead the exiting thread releases ownership and possibly wakes
// a successor, so the successor can (re)compete for ownership of the lock.
// If the EntryList is empty but the cxq is populated the exiting
// thread will drain the cxq into the EntryList. It does so by
// by detaching the cxq (installing null with CAS) and folding
// the threads from the cxq into the EntryList. The EntryList is
// doubly linked, while the cxq is singly linked because of the
// CAS-based "push" used to enqueue recently arrived threads (RATs).
//
// * Concurrency invariants:
//
// -- only the monitor owner may access or mutate the EntryList.
// The mutex property of the monitor itself protects the EntryList
// from concurrent interference.
// -- Only the monitor owner may detach the cxq.
//
// * The monitor entry list operations avoid locks, but strictly speaking
// they're not lock-free. Enter is lock-free, exit is not.
// See http://j2se.east/~dice/PERSIST/040825-LockFreeQueues.html
//
// * The cxq can have multiple concurrent "pushers" but only one concurrent
// detaching thread. This mechanism is immune from the ABA corruption.
// More precisely, the CAS-based "push" onto cxq is ABA-oblivious.
//
// * Taken together, the cxq and the EntryList constitute or form a
// single logical queue of threads stalled trying to acquire the lock.
// We use two distinct lists to improve the odds of a constant-time
// dequeue operation after acquisition (in the ::enter() epilog) and
// to reduce heat on the list ends. (c.f. Michael Scott's "2Q" algorithm).
// A key desideratum is to minimize queue & monitor metadata manipulation
// that occurs while holding the monitor lock -- that is, we want to
// minimize monitor lock holds times. Note that even a small amount of
// fixed spinning will greatly reduce the # of enqueue-dequeue operations
// on EntryList|cxq. That is, spinning relieves contention on the "inner"
// locks and monitor metadata.
//
// Cxq points to the the set of Recently Arrived Threads attempting entry.
// Because we push threads onto _cxq with CAS, the RATs must take the form of
// a singly-linked LIFO. We drain _cxq into EntryList at unlock-time when
// the unlocking thread notices that EntryList is null but _cxq is != null.
//
// The EntryList is ordered by the prevailing queue discipline and
// can be organized in any convenient fashion, such as a doubly-linked list or
// a circular doubly-linked list. Critically, we want insert and delete operations
// to operate in constant-time. If we need a priority queue then something akin
// to Solaris' sleepq would work nicely. Viz.,
// http://agg.eng/ws/on10_nightly/source/usr/src/uts/common/os/sleepq.c.
// Queue discipline is enforced at ::exit() time, when the unlocking thread
// drains the cxq into the EntryList, and orders or reorders the threads on the
// EntryList accordingly.
//
// Barring "lock barging", this mechanism provides fair cyclic ordering,
// somewhat similar to an elevator-scan.
//
// * The monitor synchronization subsystem avoids the use of native
// synchronization primitives except for the narrow platform-specific
// park-unpark abstraction. See the comments in os_solaris.cpp regarding
// the semantics of park-unpark. Put another way, this monitor implementation
// depends only on atomic operations and park-unpark. The monitor subsystem
// manages all RUNNING->BLOCKED and BLOCKED->READY transitions while the
// underlying OS manages the READY<->RUN transitions.
//
// * Waiting threads reside on the WaitSet list -- wait() puts
// the caller onto the WaitSet.
//
// * notify() or notifyAll() simply transfers threads from the WaitSet to
// either the EntryList or cxq. Subsequent exit() operations will
// unpark the notifyee. Unparking a notifee in notify() is inefficient -
// it's likely the notifyee would simply impale itself on the lock held
// by the notifier.
//
// * An interesting alternative is to encode cxq as (List,LockByte) where
// the LockByte is 0 iff the monitor is owned. _owner is simply an auxiliary
// variable, like _recursions, in the scheme. The threads or Events that form
// the list would have to be aligned in 256-byte addresses. A thread would
// try to acquire the lock or enqueue itself with CAS, but exiting threads
// could use a 1-0 protocol and simply STB to set the LockByte to 0.
// Note that is is *not* word-tearing, but it does presume that full-word
// CAS operations are coherent with intermix with STB operations. That's true
// on most common processors.
//
// * See also http://blogs.sun.com/dave
// -----------------------------------------------------------------------------
// Enter support
bool ObjectMonitor::try_enter(Thread* THREAD) {
if (THREAD != _owner) {
if (THREAD->is_lock_owned ((address)_owner)) {
assert(_recursions == 0, "internal state error");
_owner = THREAD ;
_recursions = 1 ;
OwnerIsThread = 1 ;
return true;
}
if (Atomic::cmpxchg_ptr (THREAD, &_owner, NULL) != NULL) {
return false;
}
return true;
} else {
_recursions++;
return true;
}
}
void ATTR ObjectMonitor::enter(TRAPS) {
// The following code is ordered to check the most common cases first
// and to reduce RTS->RTO cache line upgrades on SPARC and IA32 processors.
Thread * const Self = THREAD ;
void * cur ;
cur = Atomic::cmpxchg_ptr (Self, &_owner, NULL) ;
if (cur == NULL) {
// Either ASSERT _recursions == 0 or explicitly set _recursions = 0.
assert (_recursions == 0 , "invariant") ;
assert (_owner == Self, "invariant") ;
// CONSIDER: set or assert OwnerIsThread == 1
return ;
}
if (cur == Self) {
// TODO-FIXME: check for integer overflow! BUGID 6557169.
_recursions ++ ;
return ;
}
if (Self->is_lock_owned ((address)cur)) {
assert (_recursions == 0, "internal state error");
_recursions = 1 ;
// Commute owner from a thread-specific on-stack BasicLockObject address to
// a full-fledged "Thread *".
_owner = Self ;
OwnerIsThread = 1 ;
return ;
}
// We've encountered genuine contention.
assert (Self->_Stalled == 0, "invariant") ;
Self->_Stalled = intptr_t(this) ;
// Try one round of spinning *before* enqueueing Self
// and before going through the awkward and expensive state
// transitions. The following spin is strictly optional ...
// Note that if we acquire the monitor from an initial spin
// we forgo posting JVMTI events and firing DTRACE probes.
if (Knob_SpinEarly && TrySpin (Self) > 0) {
assert (_owner == Self , "invariant") ;
assert (_recursions == 0 , "invariant") ;
assert (((oop)(object()))->mark() == markOopDesc::encode(this), "invariant") ;
Self->_Stalled = 0 ;
return ;
}
assert (_owner != Self , "invariant") ;
assert (_succ != Self , "invariant") ;
assert (Self->is_Java_thread() , "invariant") ;
JavaThread * jt = (JavaThread *) Self ;
assert (!SafepointSynchronize::is_at_safepoint(), "invariant") ;
assert (jt->thread_state() != _thread_blocked , "invariant") ;
assert (this->object() != NULL , "invariant") ;
assert (_count >= 0, "invariant") ;
// Prevent deflation at STW-time. See deflate_idle_monitors() and is_busy().
// Ensure the object-monitor relationship remains stable while there's contention.
Atomic::inc_ptr(&_count);
{ // Change java thread status to indicate blocked on monitor enter.
JavaThreadBlockedOnMonitorEnterState jtbmes(jt, this);
DTRACE_MONITOR_PROBE(contended__enter, this, object(), jt);
if (JvmtiExport::should_post_monitor_contended_enter()) {
JvmtiExport::post_monitor_contended_enter(jt, this);
}
OSThreadContendState osts(Self->osthread());
ThreadBlockInVM tbivm(jt);
Self->set_current_pending_monitor(this);
// TODO-FIXME: change the following for(;;) loop to straight-line code.
for (;;) {
jt->set_suspend_equivalent();
// cleared by handle_special_suspend_equivalent_condition()
// or java_suspend_self()
EnterI (THREAD) ;
if (!ExitSuspendEquivalent(jt)) break ;
//
// We have acquired the contended monitor, but while we were
// waiting another thread suspended us. We don't want to enter
// the monitor while suspended because that would surprise the
// thread that suspended us.
//
_recursions = 0 ;
_succ = NULL ;
exit (Self) ;
jt->java_suspend_self();
}
Self->set_current_pending_monitor(NULL);
}
Atomic::dec_ptr(&_count);
assert (_count >= 0, "invariant") ;
Self->_Stalled = 0 ;
// Must either set _recursions = 0 or ASSERT _recursions == 0.
assert (_recursions == 0 , "invariant") ;
assert (_owner == Self , "invariant") ;
assert (_succ != Self , "invariant") ;
assert (((oop)(object()))->mark() == markOopDesc::encode(this), "invariant") ;
// The thread -- now the owner -- is back in vm mode.
// Report the glorious news via TI,DTrace and jvmstat.
// The probe effect is non-trivial. All the reportage occurs
// while we hold the monitor, increasing the length of the critical
// section. Amdahl's parallel speedup law comes vividly into play.
//
// Another option might be to aggregate the events (thread local or
// per-monitor aggregation) and defer reporting until a more opportune
// time -- such as next time some thread encounters contention but has
// yet to acquire the lock. While spinning that thread could
// spinning we could increment JVMStat counters, etc.
DTRACE_MONITOR_PROBE(contended__entered, this, object(), jt);
if (JvmtiExport::should_post_monitor_contended_entered()) {
JvmtiExport::post_monitor_contended_entered(jt, this);
}
if (ObjectMonitor::_sync_ContendedLockAttempts != NULL) {
ObjectMonitor::_sync_ContendedLockAttempts->inc() ;
}
}
// Caveat: TryLock() is not necessarily serializing if it returns failure.
// Callers must compensate as needed.
int ObjectMonitor::TryLock (Thread * Self) {
for (;;) {
void * own = _owner ;
if (own != NULL) return 0 ;
if (Atomic::cmpxchg_ptr (Self, &_owner, NULL) == NULL) {
// Either guarantee _recursions == 0 or set _recursions = 0.
assert (_recursions == 0, "invariant") ;
assert (_owner == Self, "invariant") ;
// CONSIDER: set or assert that OwnerIsThread == 1
return 1 ;
}
// The lock had been free momentarily, but we lost the race to the lock.
// Interference -- the CAS failed.
// We can either return -1 or retry.
// Retry doesn't make as much sense because the lock was just acquired.
if (true) return -1 ;
}
}
void ATTR ObjectMonitor::EnterI (TRAPS) {
Thread * Self = THREAD ;
assert (Self->is_Java_thread(), "invariant") ;
assert (((JavaThread *) Self)->thread_state() == _thread_blocked , "invariant") ;
// Try the lock - TATAS
if (TryLock (Self) > 0) {
assert (_succ != Self , "invariant") ;
assert (_owner == Self , "invariant") ;
assert (_Responsible != Self , "invariant") ;
return ;
}
DeferredInitialize () ;
// We try one round of spinning *before* enqueueing Self.
//
// If the _owner is ready but OFFPROC we could use a YieldTo()
// operation to donate the remainder of this thread's quantum
// to the owner. This has subtle but beneficial affinity
// effects.
if (TrySpin (Self) > 0) {
assert (_owner == Self , "invariant") ;
assert (_succ != Self , "invariant") ;
assert (_Responsible != Self , "invariant") ;
return ;
}
// The Spin failed -- Enqueue and park the thread ...
assert (_succ != Self , "invariant") ;
assert (_owner != Self , "invariant") ;
assert (_Responsible != Self , "invariant") ;
// Enqueue "Self" on ObjectMonitor's _cxq.
//
// Node acts as a proxy for Self.
// As an aside, if were to ever rewrite the synchronization code mostly
// in Java, WaitNodes, ObjectMonitors, and Events would become 1st-class
// Java objects. This would avoid awkward lifecycle and liveness issues,
// as well as eliminate a subset of ABA issues.
// TODO: eliminate ObjectWaiter and enqueue either Threads or Events.
//
ObjectWaiter node(Self) ;
Self->_ParkEvent->reset() ;
node._prev = (ObjectWaiter *) 0xBAD ;
node.TState = ObjectWaiter::TS_CXQ ;
// Push "Self" onto the front of the _cxq.
// Once on cxq/EntryList, Self stays on-queue until it acquires the lock.
// Note that spinning tends to reduce the rate at which threads
// enqueue and dequeue on EntryList|cxq.
ObjectWaiter * nxt ;
for (;;) {
node._next = nxt = _cxq ;
if (Atomic::cmpxchg_ptr (&node, &_cxq, nxt) == nxt) break ;
// Interference - the CAS failed because _cxq changed. Just retry.
// As an optional optimization we retry the lock.
if (TryLock (Self) > 0) {
assert (_succ != Self , "invariant") ;
assert (_owner == Self , "invariant") ;
assert (_Responsible != Self , "invariant") ;
return ;
}
}
// Check for cxq|EntryList edge transition to non-null. This indicates
// the onset of contention. While contention persists exiting threads
// will use a ST:MEMBAR:LD 1-1 exit protocol. When contention abates exit
// operations revert to the faster 1-0 mode. This enter operation may interleave
// (race) a concurrent 1-0 exit operation, resulting in stranding, so we
// arrange for one of the contending thread to use a timed park() operations
// to detect and recover from the race. (Stranding is form of progress failure
// where the monitor is unlocked but all the contending threads remain parked).
// That is, at least one of the contended threads will periodically poll _owner.
// One of the contending threads will become the designated "Responsible" thread.
// The Responsible thread uses a timed park instead of a normal indefinite park
// operation -- it periodically wakes and checks for and recovers from potential
// strandings admitted by 1-0 exit operations. We need at most one Responsible
// thread per-monitor at any given moment. Only threads on cxq|EntryList may
// be responsible for a monitor.
//
// Currently, one of the contended threads takes on the added role of "Responsible".
// A viable alternative would be to use a dedicated "stranding checker" thread
// that periodically iterated over all the threads (or active monitors) and unparked
// successors where there was risk of stranding. This would help eliminate the
// timer scalability issues we see on some platforms as we'd only have one thread
// -- the checker -- parked on a timer.
if ((SyncFlags & 16) == 0 && nxt == NULL && _EntryList == NULL) {
// Try to assume the role of responsible thread for the monitor.
// CONSIDER: ST vs CAS vs { if (Responsible==null) Responsible=Self }
Atomic::cmpxchg_ptr (Self, &_Responsible, NULL) ;
}
// The lock have been released while this thread was occupied queueing
// itself onto _cxq. To close the race and avoid "stranding" and
// progress-liveness failure we must resample-retry _owner before parking.
// Note the Dekker/Lamport duality: ST cxq; MEMBAR; LD Owner.
// In this case the ST-MEMBAR is accomplished with CAS().
//
// TODO: Defer all thread state transitions until park-time.
// Since state transitions are heavy and inefficient we'd like
// to defer the state transitions until absolutely necessary,
// and in doing so avoid some transitions ...
TEVENT (Inflated enter - Contention) ;
int nWakeups = 0 ;
int RecheckInterval = 1 ;
for (;;) {
if (TryLock (Self) > 0) break ;
assert (_owner != Self, "invariant") ;
if ((SyncFlags & 2) && _Responsible == NULL) {
Atomic::cmpxchg_ptr (Self, &_Responsible, NULL) ;
}
// park self
if (_Responsible == Self || (SyncFlags & 1)) {
TEVENT (Inflated enter - park TIMED) ;
Self->_ParkEvent->park ((jlong) RecheckInterval) ;
// Increase the RecheckInterval, but clamp the value.
RecheckInterval *= 8 ;
if (RecheckInterval > 1000) RecheckInterval = 1000 ;
} else {
TEVENT (Inflated enter - park UNTIMED) ;
Self->_ParkEvent->park() ;
}
if (TryLock(Self) > 0) break ;
// The lock is still contested.
// Keep a tally of the # of futile wakeups.
// Note that the counter is not protected by a lock or updated by atomics.
// That is by design - we trade "lossy" counters which are exposed to
// races during updates for a lower probe effect.
TEVENT (Inflated enter - Futile wakeup) ;
if (ObjectMonitor::_sync_FutileWakeups != NULL) {
ObjectMonitor::_sync_FutileWakeups->inc() ;
}
++ nWakeups ;
// Assuming this is not a spurious wakeup we'll normally find _succ == Self.
// We can defer clearing _succ until after the spin completes
// TrySpin() must tolerate being called with _succ == Self.
// Try yet another round of adaptive spinning.
if ((Knob_SpinAfterFutile & 1) && TrySpin (Self) > 0) break ;
// We can find that we were unpark()ed and redesignated _succ while
// we were spinning. That's harmless. If we iterate and call park(),
// park() will consume the event and return immediately and we'll
// just spin again. This pattern can repeat, leaving _succ to simply
// spin on a CPU. Enable Knob_ResetEvent to clear pending unparks().
// Alternately, we can sample fired() here, and if set, forgo spinning
// in the next iteration.
if ((Knob_ResetEvent & 1) && Self->_ParkEvent->fired()) {
Self->_ParkEvent->reset() ;
OrderAccess::fence() ;
}
if (_succ == Self) _succ = NULL ;
// Invariant: after clearing _succ a thread *must* retry _owner before parking.
OrderAccess::fence() ;
}
// Egress :
// Self has acquired the lock -- Unlink Self from the cxq or EntryList.
// Normally we'll find Self on the EntryList .
// From the perspective of the lock owner (this thread), the
// EntryList is stable and cxq is prepend-only.
// The head of cxq is volatile but the interior is stable.
// In addition, Self.TState is stable.
assert (_owner == Self , "invariant") ;
assert (object() != NULL , "invariant") ;
// I'd like to write:
// guarantee (((oop)(object()))->mark() == markOopDesc::encode(this), "invariant") ;
// but as we're at a safepoint that's not safe.
UnlinkAfterAcquire (Self, &node) ;
if (_succ == Self) _succ = NULL ;
assert (_succ != Self, "invariant") ;
if (_Responsible == Self) {
_Responsible = NULL ;
// Dekker pivot-point.
// Consider OrderAccess::storeload() here
// We may leave threads on cxq|EntryList without a designated
// "Responsible" thread. This is benign. When this thread subsequently
// exits the monitor it can "see" such preexisting "old" threads --
// threads that arrived on the cxq|EntryList before the fence, above --
// by LDing cxq|EntryList. Newly arrived threads -- that is, threads
// that arrive on cxq after the ST:MEMBAR, above -- will set Responsible
// non-null and elect a new "Responsible" timer thread.
//
// This thread executes:
// ST Responsible=null; MEMBAR (in enter epilog - here)
// LD cxq|EntryList (in subsequent exit)
//
// Entering threads in the slow/contended path execute:
// ST cxq=nonnull; MEMBAR; LD Responsible (in enter prolog)
// The (ST cxq; MEMBAR) is accomplished with CAS().
//
// The MEMBAR, above, prevents the LD of cxq|EntryList in the subsequent
// exit operation from floating above the ST Responsible=null.
//
// In *practice* however, EnterI() is always followed by some atomic
// operation such as the decrement of _count in ::enter(). Those atomics
// obviate the need for the explicit MEMBAR, above.
}
// We've acquired ownership with CAS().
// CAS is serializing -- it has MEMBAR/FENCE-equivalent semantics.
// But since the CAS() this thread may have also stored into _succ,
// EntryList, cxq or Responsible. These meta-data updates must be
// visible __before this thread subsequently drops the lock.
// Consider what could occur if we didn't enforce this constraint --
// STs to monitor meta-data and user-data could reorder with (become
// visible after) the ST in exit that drops ownership of the lock.
// Some other thread could then acquire the lock, but observe inconsistent
// or old monitor meta-data and heap data. That violates the JMM.
// To that end, the 1-0 exit() operation must have at least STST|LDST
// "release" barrier semantics. Specifically, there must be at least a
// STST|LDST barrier in exit() before the ST of null into _owner that drops
// the lock. The barrier ensures that changes to monitor meta-data and data
// protected by the lock will be visible before we release the lock, and
// therefore before some other thread (CPU) has a chance to acquire the lock.
// See also: http://gee.cs.oswego.edu/dl/jmm/cookbook.html.
//
// Critically, any prior STs to _succ or EntryList must be visible before
// the ST of null into _owner in the *subsequent* (following) corresponding
// monitorexit. Recall too, that in 1-0 mode monitorexit does not necessarily
// execute a serializing instruction.
if (SyncFlags & 8) {
OrderAccess::fence() ;
}
return ;
}
// ReenterI() is a specialized inline form of the latter half of the
// contended slow-path from EnterI(). We use ReenterI() only for
// monitor reentry in wait().
//
// In the future we should reconcile EnterI() and ReenterI(), adding
// Knob_Reset and Knob_SpinAfterFutile support and restructuring the
// loop accordingly.
void ATTR ObjectMonitor::ReenterI (Thread * Self, ObjectWaiter * SelfNode) {
assert (Self != NULL , "invariant") ;
assert (SelfNode != NULL , "invariant") ;
assert (SelfNode->_thread == Self , "invariant") ;
assert (_waiters > 0 , "invariant") ;
assert (((oop)(object()))->mark() == markOopDesc::encode(this) , "invariant") ;
assert (((JavaThread *)Self)->thread_state() != _thread_blocked, "invariant") ;
JavaThread * jt = (JavaThread *) Self ;
int nWakeups = 0 ;
for (;;) {
ObjectWaiter::TStates v = SelfNode->TState ;
guarantee (v == ObjectWaiter::TS_ENTER || v == ObjectWaiter::TS_CXQ, "invariant") ;
assert (_owner != Self, "invariant") ;
if (TryLock (Self) > 0) break ;
if (TrySpin (Self) > 0) break ;
TEVENT (Wait Reentry - parking) ;
// State transition wrappers around park() ...
// ReenterI() wisely defers state transitions until
// it's clear we must park the thread.
{
OSThreadContendState osts(Self->osthread());
ThreadBlockInVM tbivm(jt);
// cleared by handle_special_suspend_equivalent_condition()
// or java_suspend_self()
jt->set_suspend_equivalent();
if (SyncFlags & 1) {
Self->_ParkEvent->park ((jlong)1000) ;
} else {
Self->_ParkEvent->park () ;
}
// were we externally suspended while we were waiting?
for (;;) {
if (!ExitSuspendEquivalent (jt)) break ;
if (_succ == Self) { _succ = NULL; OrderAccess::fence(); }
jt->java_suspend_self();
jt->set_suspend_equivalent();
}
}
// Try again, but just so we distinguish between futile wakeups and
// successful wakeups. The following test isn't algorithmically
// necessary, but it helps us maintain sensible statistics.
if (TryLock(Self) > 0) break ;
// The lock is still contested.
// Keep a tally of the # of futile wakeups.
// Note that the counter is not protected by a lock or updated by atomics.
// That is by design - we trade "lossy" counters which are exposed to
// races during updates for a lower probe effect.
TEVENT (Wait Reentry - futile wakeup) ;
++ nWakeups ;
// Assuming this is not a spurious wakeup we'll normally
// find that _succ == Self.
if (_succ == Self) _succ = NULL ;
// Invariant: after clearing _succ a contending thread
// *must* retry _owner before parking.
OrderAccess::fence() ;
if (ObjectMonitor::_sync_FutileWakeups != NULL) {
ObjectMonitor::_sync_FutileWakeups->inc() ;
}
}
// Self has acquired the lock -- Unlink Self from the cxq or EntryList .
// Normally we'll find Self on the EntryList.
// Unlinking from the EntryList is constant-time and atomic-free.
// From the perspective of the lock owner (this thread), the
// EntryList is stable and cxq is prepend-only.
// The head of cxq is volatile but the interior is stable.
// In addition, Self.TState is stable.
assert (_owner == Self, "invariant") ;
assert (((oop)(object()))->mark() == markOopDesc::encode(this), "invariant") ;
UnlinkAfterAcquire (Self, SelfNode) ;
if (_succ == Self) _succ = NULL ;
assert (_succ != Self, "invariant") ;
SelfNode->TState = ObjectWaiter::TS_RUN ;
OrderAccess::fence() ; // see comments at the end of EnterI()
}
// after the thread acquires the lock in ::enter(). Equally, we could defer
// unlinking the thread until ::exit()-time.
void ObjectMonitor::UnlinkAfterAcquire (Thread * Self, ObjectWaiter * SelfNode)
{
assert (_owner == Self, "invariant") ;
assert (SelfNode->_thread == Self, "invariant") ;
if (SelfNode->TState == ObjectWaiter::TS_ENTER) {
// Normal case: remove Self from the DLL EntryList .
// This is a constant-time operation.
ObjectWaiter * nxt = SelfNode->_next ;
ObjectWaiter * prv = SelfNode->_prev ;
if (nxt != NULL) nxt->_prev = prv ;
if (prv != NULL) prv->_next = nxt ;
if (SelfNode == _EntryList ) _EntryList = nxt ;
assert (nxt == NULL || nxt->TState == ObjectWaiter::TS_ENTER, "invariant") ;
assert (prv == NULL || prv->TState == ObjectWaiter::TS_ENTER, "invariant") ;
TEVENT (Unlink from EntryList) ;
} else {
guarantee (SelfNode->TState == ObjectWaiter::TS_CXQ, "invariant") ;
// Inopportune interleaving -- Self is still on the cxq.
// This usually means the enqueue of self raced an exiting thread.
// Normally we'll find Self near the front of the cxq, so
// dequeueing is typically fast. If needbe we can accelerate
// this with some MCS/CHL-like bidirectional list hints and advisory
// back-links so dequeueing from the interior will normally operate
// in constant-time.
// Dequeue Self from either the head (with CAS) or from the interior
// with a linear-time scan and normal non-atomic memory operations.
// CONSIDER: if Self is on the cxq then simply drain cxq into EntryList
// and then unlink Self from EntryList. We have to drain eventually,
// so it might as well be now.
ObjectWaiter * v = _cxq ;
assert (v != NULL, "invariant") ;
if (v != SelfNode || Atomic::cmpxchg_ptr (SelfNode->_next, &_cxq, v) != v) {
// The CAS above can fail from interference IFF a "RAT" arrived.
// In that case Self must be in the interior and can no longer be
// at the head of cxq.
if (v == SelfNode) {
assert (_cxq != v, "invariant") ;
v = _cxq ; // CAS above failed - start scan at head of list
}
ObjectWaiter * p ;
ObjectWaiter * q = NULL ;
for (p = v ; p != NULL && p != SelfNode; p = p->_next) {
q = p ;
assert (p->TState == ObjectWaiter::TS_CXQ, "invariant") ;
}
assert (v != SelfNode, "invariant") ;
assert (p == SelfNode, "Node not found on cxq") ;
assert (p != _cxq, "invariant") ;
assert (q != NULL, "invariant") ;
assert (q->_next == p, "invariant") ;
q->_next = p->_next ;
}
TEVENT (Unlink from cxq) ;
}
// Diagnostic hygiene ...
SelfNode->_prev = (ObjectWaiter *) 0xBAD ;
SelfNode->_next = (ObjectWaiter *) 0xBAD ;
SelfNode->TState = ObjectWaiter::TS_RUN ;
}
// -----------------------------------------------------------------------------
// Exit support
//
// exit()
// ~~~~~~
// Note that the collector can't reclaim the objectMonitor or deflate
// the object out from underneath the thread calling ::exit() as the
// thread calling ::exit() never transitions to a stable state.
// This inhibits GC, which in turn inhibits asynchronous (and
// inopportune) reclamation of "this".
//
// We'd like to assert that: (THREAD->thread_state() != _thread_blocked) ;
// There's one exception to the claim above, however. EnterI() can call
// exit() to drop a lock if the acquirer has been externally suspended.
// In that case exit() is called with _thread_state as _thread_blocked,
// but the monitor's _count field is > 0, which inhibits reclamation.
//
// 1-0 exit
// ~~~~~~~~
// ::exit() uses a canonical 1-1 idiom with a MEMBAR although some of
// the fast-path operators have been optimized so the common ::exit()
// operation is 1-0. See i486.ad fast_unlock(), for instance.
// The code emitted by fast_unlock() elides the usual MEMBAR. This
// greatly improves latency -- MEMBAR and CAS having considerable local
// latency on modern processors -- but at the cost of "stranding". Absent the
// MEMBAR, a thread in fast_unlock() can race a thread in the slow
// ::enter() path, resulting in the entering thread being stranding
// and a progress-liveness failure. Stranding is extremely rare.
// We use timers (timed park operations) & periodic polling to detect
// and recover from stranding. Potentially stranded threads periodically
// wake up and poll the lock. See the usage of the _Responsible variable.
//
// The CAS() in enter provides for safety and exclusion, while the CAS or
// MEMBAR in exit provides for progress and avoids stranding. 1-0 locking
// eliminates the CAS/MEMBAR from the exist path, but it admits stranding.
// We detect and recover from stranding with timers.
//
// If a thread transiently strands it'll park until (a) another
// thread acquires the lock and then drops the lock, at which time the
// exiting thread will notice and unpark the stranded thread, or, (b)
// the timer expires. If the lock is high traffic then the stranding latency
// will be low due to (a). If the lock is low traffic then the odds of
// stranding are lower, although the worst-case stranding latency
// is longer. Critically, we don't want to put excessive load in the
// platform's timer subsystem. We want to minimize both the timer injection
// rate (timers created/sec) as well as the number of timers active at
// any one time. (more precisely, we want to minimize timer-seconds, which is
// the integral of the # of active timers at any instant over time).
// Both impinge on OS scalability. Given that, at most one thread parked on
// a monitor will use a timer.
void ATTR ObjectMonitor::exit(TRAPS) {
Thread * Self = THREAD ;
if (THREAD != _owner) {
if (THREAD->is_lock_owned((address) _owner)) {
// Transmute _owner from a BasicLock pointer to a Thread address.
// We don't need to hold _mutex for this transition.
// Non-null to Non-null is safe as long as all readers can
// tolerate either flavor.
assert (_recursions == 0, "invariant") ;
_owner = THREAD ;
_recursions = 0 ;
OwnerIsThread = 1 ;
} else {
// NOTE: we need to handle unbalanced monitor enter/exit
// in native code by throwing an exception.
// TODO: Throw an IllegalMonitorStateException ?
TEVENT (Exit - Throw IMSX) ;
assert(false, "Non-balanced monitor enter/exit!");
if (false) {
THROW(vmSymbols::java_lang_IllegalMonitorStateException());
}
return;
}
}
if (_recursions != 0) {
_recursions--; // this is simple recursive enter
TEVENT (Inflated exit - recursive) ;
return ;
}
// Invariant: after setting Responsible=null an thread must execute
// a MEMBAR or other serializing instruction before fetching EntryList|cxq.
if ((SyncFlags & 4) == 0) {
_Responsible = NULL ;
}
for (;;) {
assert (THREAD == _owner, "invariant") ;
if (Knob_ExitPolicy == 0) {
// release semantics: prior loads and stores from within the critical section
// must not float (reorder) past the following store that drops the lock.
// On SPARC that requires MEMBAR #loadstore|#storestore.
// But of course in TSO #loadstore|#storestore is not required.
// I'd like to write one of the following:
// A. OrderAccess::release() ; _owner = NULL
// B. OrderAccess::loadstore(); OrderAccess::storestore(); _owner = NULL;
// Unfortunately OrderAccess::release() and OrderAccess::loadstore() both
// store into a _dummy variable. That store is not needed, but can result
// in massive wasteful coherency traffic on classic SMP systems.
// Instead, I use release_store(), which is implemented as just a simple
// ST on x64, x86 and SPARC.
OrderAccess::release_store_ptr (&_owner, NULL) ; // drop the lock
OrderAccess::storeload() ; // See if we need to wake a successor
if ((intptr_t(_EntryList)|intptr_t(_cxq)) == 0 || _succ != NULL) {
TEVENT (Inflated exit - simple egress) ;
return ;
}
TEVENT (Inflated exit - complex egress) ;
// Normally the exiting thread is responsible for ensuring succession,
// but if other successors are ready or other entering threads are spinning
// then this thread can simply store NULL into _owner and exit without
// waking a successor. The existence of spinners or ready successors
// guarantees proper succession (liveness). Responsibility passes to the
// ready or running successors. The exiting thread delegates the duty.
// More precisely, if a successor already exists this thread is absolved
// of the responsibility of waking (unparking) one.
//
// The _succ variable is critical to reducing futile wakeup frequency.
// _succ identifies the "heir presumptive" thread that has been made
// ready (unparked) but that has not yet run. We need only one such
// successor thread to guarantee progress.
// See http://www.usenix.org/events/jvm01/full_papers/dice/dice.pdf
// section 3.3 "Futile Wakeup Throttling" for details.
//
// Note that spinners in Enter() also set _succ non-null.
// In the current implementation spinners opportunistically set
// _succ so that exiting threads might avoid waking a successor.
// Another less appealing alternative would be for the exiting thread
// to drop the lock and then spin briefly to see if a spinner managed
// to acquire the lock. If so, the exiting thread could exit
// immediately without waking a successor, otherwise the exiting
// thread would need to dequeue and wake a successor.
// (Note that we'd need to make the post-drop spin short, but no
// shorter than the worst-case round-trip cache-line migration time.
// The dropped lock needs to become visible to the spinner, and then
// the acquisition of the lock by the spinner must become visible to
// the exiting thread).
//
// It appears that an heir-presumptive (successor) must be made ready.
// Only the current lock owner can manipulate the EntryList or
// drain _cxq, so we need to reacquire the lock. If we fail
// to reacquire the lock the responsibility for ensuring succession
// falls to the new owner.
//
if (Atomic::cmpxchg_ptr (THREAD, &_owner, NULL) != NULL) {
return ;
}
TEVENT (Exit - Reacquired) ;
} else {
if ((intptr_t(_EntryList)|intptr_t(_cxq)) == 0 || _succ != NULL) {
OrderAccess::release_store_ptr (&_owner, NULL) ; // drop the lock
OrderAccess::storeload() ;
// Ratify the previously observed values.
if (_cxq == NULL || _succ != NULL) {
TEVENT (Inflated exit - simple egress) ;
return ;
}
// inopportune interleaving -- the exiting thread (this thread)
// in the fast-exit path raced an entering thread in the slow-enter
// path.
// We have two choices:
// A. Try to reacquire the lock.
// If the CAS() fails return immediately, otherwise
// we either restart/rerun the exit operation, or simply
// fall-through into the code below which wakes a successor.
// B. If the elements forming the EntryList|cxq are TSM
// we could simply unpark() the lead thread and return
// without having set _succ.
if (Atomic::cmpxchg_ptr (THREAD, &_owner, NULL) != NULL) {
TEVENT (Inflated exit - reacquired succeeded) ;
return ;
}
TEVENT (Inflated exit - reacquired failed) ;
} else {
TEVENT (Inflated exit - complex egress) ;
}
}
guarantee (_owner == THREAD, "invariant") ;
ObjectWaiter * w = NULL ;
int QMode = Knob_QMode ;
if (QMode == 2 && _cxq != NULL) {
// QMode == 2 : cxq has precedence over EntryList.
// Try to directly wake a successor from the cxq.
// If successful, the successor will need to unlink itself from cxq.
w = _cxq ;
assert (w != NULL, "invariant") ;
assert (w->TState == ObjectWaiter::TS_CXQ, "Invariant") ;
ExitEpilog (Self, w) ;
return ;
}
if (QMode == 3 && _cxq != NULL) {
// Aggressively drain cxq into EntryList at the first opportunity.
// This policy ensure that recently-run threads live at the head of EntryList.
// Drain _cxq into EntryList - bulk transfer.
// First, detach _cxq.
// The following loop is tantamount to: w = swap (&cxq, NULL)
w = _cxq ;
for (;;) {
assert (w != NULL, "Invariant") ;
ObjectWaiter * u = (ObjectWaiter *) Atomic::cmpxchg_ptr (NULL, &_cxq, w) ;
if (u == w) break ;
w = u ;
}
assert (w != NULL , "invariant") ;
ObjectWaiter * q = NULL ;
ObjectWaiter * p ;
for (p = w ; p != NULL ; p = p->_next) {
guarantee (p->TState == ObjectWaiter::TS_CXQ, "Invariant") ;
p->TState = ObjectWaiter::TS_ENTER ;
p->_prev = q ;
q = p ;
}
// Append the RATs to the EntryList
// TODO: organize EntryList as a CDLL so we can locate the tail in constant-time.
ObjectWaiter * Tail ;
for (Tail = _EntryList ; Tail != NULL && Tail->_next != NULL ; Tail = Tail->_next) ;
if (Tail == NULL) {
_EntryList = w ;
} else {
Tail->_next = w ;
w->_prev = Tail ;
}
// Fall thru into code that tries to wake a successor from EntryList
}
if (QMode == 4 && _cxq != NULL) {
// Aggressively drain cxq into EntryList at the first opportunity.
// This policy ensure that recently-run threads live at the head of EntryList.
// Drain _cxq into EntryList - bulk transfer.
// First, detach _cxq.
// The following loop is tantamount to: w = swap (&cxq, NULL)
w = _cxq ;
for (;;) {
assert (w != NULL, "Invariant") ;
ObjectWaiter * u = (ObjectWaiter *) Atomic::cmpxchg_ptr (NULL, &_cxq, w) ;
if (u == w) break ;
w = u ;
}
assert (w != NULL , "invariant") ;
ObjectWaiter * q = NULL ;
ObjectWaiter * p ;
for (p = w ; p != NULL ; p = p->_next) {
guarantee (p->TState == ObjectWaiter::TS_CXQ, "Invariant") ;
p->TState = ObjectWaiter::TS_ENTER ;
p->_prev = q ;
q = p ;
}
// Prepend the RATs to the EntryList
if (_EntryList != NULL) {
q->_next = _EntryList ;
_EntryList->_prev = q ;
}
_EntryList = w ;
// Fall thru into code that tries to wake a successor from EntryList
}
w = _EntryList ;
if (w != NULL) {
// I'd like to write: guarantee (w->_thread != Self).
// But in practice an exiting thread may find itself on the EntryList.
// Lets say thread T1 calls O.wait(). Wait() enqueues T1 on O's waitset and
// then calls exit(). Exit release the lock by setting O._owner to NULL.
// Lets say T1 then stalls. T2 acquires O and calls O.notify(). The
// notify() operation moves T1 from O's waitset to O's EntryList. T2 then
// release the lock "O". T2 resumes immediately after the ST of null into
// _owner, above. T2 notices that the EntryList is populated, so it
// reacquires the lock and then finds itself on the EntryList.
// Given all that, we have to tolerate the circumstance where "w" is
// associated with Self.
assert (w->TState == ObjectWaiter::TS_ENTER, "invariant") ;
ExitEpilog (Self, w) ;
return ;
}
// If we find that both _cxq and EntryList are null then just
// re-run the exit protocol from the top.
w = _cxq ;
if (w == NULL) continue ;
// Drain _cxq into EntryList - bulk transfer.
// First, detach _cxq.
// The following loop is tantamount to: w = swap (&cxq, NULL)
for (;;) {
assert (w != NULL, "Invariant") ;
ObjectWaiter * u = (ObjectWaiter *) Atomic::cmpxchg_ptr (NULL, &_cxq, w) ;
if (u == w) break ;
w = u ;
}
TEVENT (Inflated exit - drain cxq into EntryList) ;
assert (w != NULL , "invariant") ;
assert (_EntryList == NULL , "invariant") ;
// Convert the LIFO SLL anchored by _cxq into a DLL.
// The list reorganization step operates in O(LENGTH(w)) time.
// It's critical that this step operate quickly as
// "Self" still holds the outer-lock, restricting parallelism
// and effectively lengthening the critical section.
// Invariant: s chases t chases u.
// TODO-FIXME: consider changing EntryList from a DLL to a CDLL so
// we have faster access to the tail.
if (QMode == 1) {
// QMode == 1 : drain cxq to EntryList, reversing order
// We also reverse the order of the list.
ObjectWaiter * s = NULL ;
ObjectWaiter * t = w ;
ObjectWaiter * u = NULL ;
while (t != NULL) {
guarantee (t->TState == ObjectWaiter::TS_CXQ, "invariant") ;
t->TState = ObjectWaiter::TS_ENTER ;
u = t->_next ;
t->_prev = u ;
t->_next = s ;
s = t;
t = u ;
}
_EntryList = s ;
assert (s != NULL, "invariant") ;
} else {
// QMode == 0 or QMode == 2
_EntryList = w ;
ObjectWaiter * q = NULL ;
ObjectWaiter * p ;
for (p = w ; p != NULL ; p = p->_next) {
guarantee (p->TState == ObjectWaiter::TS_CXQ, "Invariant") ;
p->TState = ObjectWaiter::TS_ENTER ;
p->_prev = q ;
q = p ;
}
}
// In 1-0 mode we need: ST EntryList; MEMBAR #storestore; ST _owner = NULL
// The MEMBAR is satisfied by the release_store() operation in ExitEpilog().
// See if we can abdicate to a spinner instead of waking a thread.
// A primary goal of the implementation is to reduce the
// context-switch rate.
if (_succ != NULL) continue;
w = _EntryList ;
if (w != NULL) {
guarantee (w->TState == ObjectWaiter::TS_ENTER, "invariant") ;
ExitEpilog (Self, w) ;
return ;
}
}
}
// ExitSuspendEquivalent:
// A faster alternate to handle_special_suspend_equivalent_condition()
//
// handle_special_suspend_equivalent_condition() unconditionally
// acquires the SR_lock. On some platforms uncontended MutexLocker()
// operations have high latency. Note that in ::enter() we call HSSEC
// while holding the monitor, so we effectively lengthen the critical sections.
//
// There are a number of possible solutions:
//
// A. To ameliorate the problem we might also defer state transitions
// to as late as possible -- just prior to parking.
// Given that, we'd call HSSEC after having returned from park(),
// but before attempting to acquire the monitor. This is only a
// partial solution. It avoids calling HSSEC while holding the
// monitor (good), but it still increases successor reacquisition latency --
// the interval between unparking a successor and the time the successor
// resumes and retries the lock. See ReenterI(), which defers state transitions.
// If we use this technique we can also avoid EnterI()-exit() loop
// in ::enter() where we iteratively drop the lock and then attempt
// to reacquire it after suspending.
//
// B. In the future we might fold all the suspend bits into a
// composite per-thread suspend flag and then update it with CAS().
// Alternately, a Dekker-like mechanism with multiple variables
// would suffice:
// ST Self->_suspend_equivalent = false
// MEMBAR
// LD Self_>_suspend_flags
//
bool ObjectMonitor::ExitSuspendEquivalent (JavaThread * jSelf) {
int Mode = Knob_FastHSSEC ;
if (Mode && !jSelf->is_external_suspend()) {
assert (jSelf->is_suspend_equivalent(), "invariant") ;
jSelf->clear_suspend_equivalent() ;
if (2 == Mode) OrderAccess::storeload() ;
if (!jSelf->is_external_suspend()) return false ;
// We raced a suspension -- fall thru into the slow path
TEVENT (ExitSuspendEquivalent - raced) ;
jSelf->set_suspend_equivalent() ;
}
return jSelf->handle_special_suspend_equivalent_condition() ;
}
void ObjectMonitor::ExitEpilog (Thread * Self, ObjectWaiter * Wakee) {
assert (_owner == Self, "invariant") ;
// Exit protocol:
// 1. ST _succ = wakee
// 2. membar #loadstore|#storestore;
// 2. ST _owner = NULL
// 3. unpark(wakee)
_succ = Knob_SuccEnabled ? Wakee->_thread : NULL ;
ParkEvent * Trigger = Wakee->_event ;
// Hygiene -- once we've set _owner = NULL we can't safely dereference Wakee again.
// The thread associated with Wakee may have grabbed the lock and "Wakee" may be
// out-of-scope (non-extant).
Wakee = NULL ;
// Drop the lock
OrderAccess::release_store_ptr (&_owner, NULL) ;
OrderAccess::fence() ; // ST _owner vs LD in unpark()
if (SafepointSynchronize::do_call_back()) {
TEVENT (unpark before SAFEPOINT) ;
}
DTRACE_MONITOR_PROBE(contended__exit, this, object(), Self);
Trigger->unpark() ;
// Maintain stats and report events to JVMTI
if (ObjectMonitor::_sync_Parks != NULL) {
ObjectMonitor::_sync_Parks->inc() ;
}
}
// -----------------------------------------------------------------------------
// Class Loader deadlock handling.
//
// complete_exit exits a lock returning recursion count
// complete_exit/reenter operate as a wait without waiting
// complete_exit requires an inflated monitor
// The _owner field is not always the Thread addr even with an
// inflated monitor, e.g. the monitor can be inflated by a non-owning
// thread due to contention.
intptr_t ObjectMonitor::complete_exit(TRAPS) {
Thread * const Self = THREAD;
assert(Self->is_Java_thread(), "Must be Java thread!");
JavaThread *jt = (JavaThread *)THREAD;
DeferredInitialize();
if (THREAD != _owner) {
if (THREAD->is_lock_owned ((address)_owner)) {
assert(_recursions == 0, "internal state error");
_owner = THREAD ; /* Convert from basiclock addr to Thread addr */
_recursions = 0 ;
OwnerIsThread = 1 ;
}
}
guarantee(Self == _owner, "complete_exit not owner");
intptr_t save = _recursions; // record the old recursion count
_recursions = 0; // set the recursion level to be 0
exit (Self) ; // exit the monitor
guarantee (_owner != Self, "invariant");
return save;
}
// reenter() enters a lock and sets recursion count
// complete_exit/reenter operate as a wait without waiting
void ObjectMonitor::reenter(intptr_t recursions, TRAPS) {
Thread * const Self = THREAD;
assert(Self->is_Java_thread(), "Must be Java thread!");
JavaThread *jt = (JavaThread *)THREAD;
guarantee(_owner != Self, "reenter already owner");
enter (THREAD); // enter the monitor
guarantee (_recursions == 0, "reenter recursion");
_recursions = recursions;
return;
}
// -----------------------------------------------------------------------------
// A macro is used below because there may already be a pending
// exception which should not abort the execution of the routines
// which use this (which is why we don't put this into check_slow and
// call it with a CHECK argument).
#define CHECK_OWNER() \
do { \
if (THREAD != _owner) { \
if (THREAD->is_lock_owned((address) _owner)) { \
_owner = THREAD ; /* Convert from basiclock addr to Thread addr */ \
_recursions = 0; \
OwnerIsThread = 1 ; \
} else { \
TEVENT (Throw IMSX) ; \
THROW(vmSymbols::java_lang_IllegalMonitorStateException()); \
} \
} \
} while (false)
// check_slow() is a misnomer. It's called to simply to throw an IMSX exception.
// TODO-FIXME: remove check_slow() -- it's likely dead.
void ObjectMonitor::check_slow(TRAPS) {
TEVENT (check_slow - throw IMSX) ;
assert(THREAD != _owner && !THREAD->is_lock_owned((address) _owner), "must not be owner");
THROW_MSG(vmSymbols::java_lang_IllegalMonitorStateException(), "current thread not owner");
}
static int Adjust (volatile int * adr, int dx) {
int v ;
for (v = *adr ; Atomic::cmpxchg (v + dx, adr, v) != v; v = *adr) ;
return v ;
}
// -----------------------------------------------------------------------------
// Wait/Notify/NotifyAll
//
// Note: a subset of changes to ObjectMonitor::wait()
// will need to be replicated in complete_exit above
void ObjectMonitor::wait(jlong millis, bool interruptible, TRAPS) {
Thread * const Self = THREAD ;
assert(Self->is_Java_thread(), "Must be Java thread!");
JavaThread *jt = (JavaThread *)THREAD;
DeferredInitialize () ;
// Throw IMSX or IEX.
CHECK_OWNER();
// check for a pending interrupt
if (interruptible && Thread::is_interrupted(Self, true) && !HAS_PENDING_EXCEPTION) {
// post monitor waited event. Note that this is past-tense, we are done waiting.
if (JvmtiExport::should_post_monitor_waited()) {
// Note: 'false' parameter is passed here because the
// wait was not timed out due to thread interrupt.
JvmtiExport::post_monitor_waited(jt, this, false);
}
TEVENT (Wait - Throw IEX) ;
THROW(vmSymbols::java_lang_InterruptedException());
return ;
}
TEVENT (Wait) ;
assert (Self->_Stalled == 0, "invariant") ;
Self->_Stalled = intptr_t(this) ;
jt->set_current_waiting_monitor(this);
// create a node to be put into the queue
// Critically, after we reset() the event but prior to park(), we must check
// for a pending interrupt.
ObjectWaiter node(Self);
node.TState = ObjectWaiter::TS_WAIT ;
Self->_ParkEvent->reset() ;
OrderAccess::fence(); // ST into Event; membar ; LD interrupted-flag
// Enter the waiting queue, which is a circular doubly linked list in this case
// but it could be a priority queue or any data structure.
// _WaitSetLock protects the wait queue. Normally the wait queue is accessed only
// by the the owner of the monitor *except* in the case where park()
// returns because of a timeout of interrupt. Contention is exceptionally rare
// so we use a simple spin-lock instead of a heavier-weight blocking lock.
Thread::SpinAcquire (&_WaitSetLock, "WaitSet - add") ;
AddWaiter (&node) ;
Thread::SpinRelease (&_WaitSetLock) ;
if ((SyncFlags & 4) == 0) {
_Responsible = NULL ;
}
intptr_t save = _recursions; // record the old recursion count
_waiters++; // increment the number of waiters
_recursions = 0; // set the recursion level to be 1
exit (Self) ; // exit the monitor
guarantee (_owner != Self, "invariant") ;
// As soon as the ObjectMonitor's ownership is dropped in the exit()
// call above, another thread can enter() the ObjectMonitor, do the
// notify(), and exit() the ObjectMonitor. If the other thread's
// exit() call chooses this thread as the successor and the unpark()
// call happens to occur while this thread is posting a
// MONITOR_CONTENDED_EXIT event, then we run the risk of the event
// handler using RawMonitors and consuming the unpark().
//
// To avoid the problem, we re-post the event. This does no harm
// even if the original unpark() was not consumed because we are the
// chosen successor for this monitor.
if (node._notified != 0 && _succ == Self) {
node._event->unpark();
}
// The thread is on the WaitSet list - now park() it.
// On MP systems it's conceivable that a brief spin before we park
// could be profitable.
//
// TODO-FIXME: change the following logic to a loop of the form
// while (!timeout && !interrupted && _notified == 0) park()
int ret = OS_OK ;
int WasNotified = 0 ;
{ // State transition wrappers
OSThread* osthread = Self->osthread();
OSThreadWaitState osts(osthread, true);
{
ThreadBlockInVM tbivm(jt);
// Thread is in thread_blocked state and oop access is unsafe.
jt->set_suspend_equivalent();
if (interruptible && (Thread::is_interrupted(THREAD, false) || HAS_PENDING_EXCEPTION)) {
// Intentionally empty
} else
if (node._notified == 0) {
if (millis <= 0) {
Self->_ParkEvent->park () ;
} else {
ret = Self->_ParkEvent->park (millis) ;
}
}
// were we externally suspended while we were waiting?
if (ExitSuspendEquivalent (jt)) {
// TODO-FIXME: add -- if succ == Self then succ = null.
jt->java_suspend_self();
}
} // Exit thread safepoint: transition _thread_blocked -> _thread_in_vm
// Node may be on the WaitSet, the EntryList (or cxq), or in transition
// from the WaitSet to the EntryList.
// See if we need to remove Node from the WaitSet.
// We use double-checked locking to avoid grabbing _WaitSetLock
// if the thread is not on the wait queue.
//
// Note that we don't need a fence before the fetch of TState.
// In the worst case we'll fetch a old-stale value of TS_WAIT previously
// written by the is thread. (perhaps the fetch might even be satisfied
// by a look-aside into the processor's own store buffer, although given
// the length of the code path between the prior ST and this load that's
// highly unlikely). If the following LD fetches a stale TS_WAIT value
// then we'll acquire the lock and then re-fetch a fresh TState value.
// That is, we fail toward safety.
if (node.TState == ObjectWaiter::TS_WAIT) {
Thread::SpinAcquire (&_WaitSetLock, "WaitSet - unlink") ;
if (node.TState == ObjectWaiter::TS_WAIT) {
DequeueSpecificWaiter (&node) ; // unlink from WaitSet
assert(node._notified == 0, "invariant");
node.TState = ObjectWaiter::TS_RUN ;
}
Thread::SpinRelease (&_WaitSetLock) ;
}
// The thread is now either on off-list (TS_RUN),
// on the EntryList (TS_ENTER), or on the cxq (TS_CXQ).
// The Node's TState variable is stable from the perspective of this thread.
// No other threads will asynchronously modify TState.
guarantee (node.TState != ObjectWaiter::TS_WAIT, "invariant") ;
OrderAccess::loadload() ;
if (_succ == Self) _succ = NULL ;
WasNotified = node._notified ;
// Reentry phase -- reacquire the monitor.
// re-enter contended monitor after object.wait().
// retain OBJECT_WAIT state until re-enter successfully completes
// Thread state is thread_in_vm and oop access is again safe,
// although the raw address of the object may have changed.
// (Don't cache naked oops over safepoints, of course).
// post monitor waited event. Note that this is past-tense, we are done waiting.
if (JvmtiExport::should_post_monitor_waited()) {
JvmtiExport::post_monitor_waited(jt, this, ret == OS_TIMEOUT);
}
OrderAccess::fence() ;
assert (Self->_Stalled != 0, "invariant") ;
Self->_Stalled = 0 ;
assert (_owner != Self, "invariant") ;
ObjectWaiter::TStates v = node.TState ;
if (v == ObjectWaiter::TS_RUN) {
enter (Self) ;
} else {
guarantee (v == ObjectWaiter::TS_ENTER || v == ObjectWaiter::TS_CXQ, "invariant") ;
ReenterI (Self, &node) ;
node.wait_reenter_end(this);
}
// Self has reacquired the lock.
// Lifecycle - the node representing Self must not appear on any queues.
// Node is about to go out-of-scope, but even if it were immortal we wouldn't
// want residual elements associated with this thread left on any lists.
guarantee (node.TState == ObjectWaiter::TS_RUN, "invariant") ;
assert (_owner == Self, "invariant") ;
assert (_succ != Self , "invariant") ;
} // OSThreadWaitState()
jt->set_current_waiting_monitor(NULL);
guarantee (_recursions == 0, "invariant") ;
_recursions = save; // restore the old recursion count
_waiters--; // decrement the number of waiters
// Verify a few postconditions
assert (_owner == Self , "invariant") ;
assert (_succ != Self , "invariant") ;
assert (((oop)(object()))->mark() == markOopDesc::encode(this), "invariant") ;
if (SyncFlags & 32) {
OrderAccess::fence() ;
}
// check if the notification happened
if (!WasNotified) {
// no, it could be timeout or Thread.interrupt() or both
// check for interrupt event, otherwise it is timeout
if (interruptible && Thread::is_interrupted(Self, true) && !HAS_PENDING_EXCEPTION) {
TEVENT (Wait - throw IEX from epilog) ;
THROW(vmSymbols::java_lang_InterruptedException());
}
}
// NOTE: Spurious wake up will be consider as timeout.
// Monitor notify has precedence over thread interrupt.
}
// Consider:
// If the lock is cool (cxq == null && succ == null) and we're on an MP system
// then instead of transferring a thread from the WaitSet to the EntryList
// we might just dequeue a thread from the WaitSet and directly unpark() it.
void ObjectMonitor::notify(TRAPS) {
CHECK_OWNER();
if (_WaitSet == NULL) {
TEVENT (Empty-Notify) ;
return ;
}
DTRACE_MONITOR_PROBE(notify, this, object(), THREAD);
int Policy = Knob_MoveNotifyee ;
Thread::SpinAcquire (&_WaitSetLock, "WaitSet - notify") ;
ObjectWaiter * iterator = DequeueWaiter() ;
if (iterator != NULL) {
TEVENT (Notify1 - Transfer) ;
guarantee (iterator->TState == ObjectWaiter::TS_WAIT, "invariant") ;
guarantee (iterator->_notified == 0, "invariant") ;
if (Policy != 4) {
iterator->TState = ObjectWaiter::TS_ENTER ;
}
iterator->_notified = 1 ;
ObjectWaiter * List = _EntryList ;
if (List != NULL) {
assert (List->_prev == NULL, "invariant") ;
assert (List->TState == ObjectWaiter::TS_ENTER, "invariant") ;
assert (List != iterator, "invariant") ;
}
if (Policy == 0) { // prepend to EntryList
if (List == NULL) {
iterator->_next = iterator->_prev = NULL ;
_EntryList = iterator ;
} else {
List->_prev = iterator ;
iterator->_next = List ;
iterator->_prev = NULL ;
_EntryList = iterator ;
}
} else
if (Policy == 1) { // append to EntryList
if (List == NULL) {
iterator->_next = iterator->_prev = NULL ;
_EntryList = iterator ;
} else {
// CONSIDER: finding the tail currently requires a linear-time walk of
// the EntryList. We can make tail access constant-time by converting to
// a CDLL instead of using our current DLL.
ObjectWaiter * Tail ;
for (Tail = List ; Tail->_next != NULL ; Tail = Tail->_next) ;
assert (Tail != NULL && Tail->_next == NULL, "invariant") ;
Tail->_next = iterator ;
iterator->_prev = Tail ;
iterator->_next = NULL ;
}
} else
if (Policy == 2) { // prepend to cxq
// prepend to cxq
if (List == NULL) {
iterator->_next = iterator->_prev = NULL ;
_EntryList = iterator ;
} else {
iterator->TState = ObjectWaiter::TS_CXQ ;
for (;;) {
ObjectWaiter * Front = _cxq ;
iterator->_next = Front ;
if (Atomic::cmpxchg_ptr (iterator, &_cxq, Front) == Front) {
break ;
}
}
}
} else
if (Policy == 3) { // append to cxq
iterator->TState = ObjectWaiter::TS_CXQ ;
for (;;) {
ObjectWaiter * Tail ;
Tail = _cxq ;
if (Tail == NULL) {
iterator->_next = NULL ;
if (Atomic::cmpxchg_ptr (iterator, &_cxq, NULL) == NULL) {
break ;
}
} else {
while (Tail->_next != NULL) Tail = Tail->_next ;
Tail->_next = iterator ;
iterator->_prev = Tail ;
iterator->_next = NULL ;
break ;
}
}
} else {
ParkEvent * ev = iterator->_event ;
iterator->TState = ObjectWaiter::TS_RUN ;
OrderAccess::fence() ;
ev->unpark() ;
}
if (Policy < 4) {
iterator->wait_reenter_begin(this);
}
// _WaitSetLock protects the wait queue, not the EntryList. We could
// move the add-to-EntryList operation, above, outside the critical section
// protected by _WaitSetLock. In practice that's not useful. With the
// exception of wait() timeouts and interrupts the monitor owner
// is the only thread that grabs _WaitSetLock. There's almost no contention
// on _WaitSetLock so it's not profitable to reduce the length of the
// critical section.
}
Thread::SpinRelease (&_WaitSetLock) ;
if (iterator != NULL && ObjectMonitor::_sync_Notifications != NULL) {
ObjectMonitor::_sync_Notifications->inc() ;
}
}
void ObjectMonitor::notifyAll(TRAPS) {
CHECK_OWNER();
ObjectWaiter* iterator;
if (_WaitSet == NULL) {
TEVENT (Empty-NotifyAll) ;
return ;
}
DTRACE_MONITOR_PROBE(notifyAll, this, object(), THREAD);
int Policy = Knob_MoveNotifyee ;
int Tally = 0 ;
Thread::SpinAcquire (&_WaitSetLock, "WaitSet - notifyall") ;
for (;;) {
iterator = DequeueWaiter () ;
if (iterator == NULL) break ;
TEVENT (NotifyAll - Transfer1) ;
++Tally ;
// Disposition - what might we do with iterator ?
// a. add it directly to the EntryList - either tail or head.
// b. push it onto the front of the _cxq.
// For now we use (a).
guarantee (iterator->TState == ObjectWaiter::TS_WAIT, "invariant") ;
guarantee (iterator->_notified == 0, "invariant") ;
iterator->_notified = 1 ;
if (Policy != 4) {
iterator->TState = ObjectWaiter::TS_ENTER ;
}
ObjectWaiter * List = _EntryList ;
if (List != NULL) {
assert (List->_prev == NULL, "invariant") ;
assert (List->TState == ObjectWaiter::TS_ENTER, "invariant") ;
assert (List != iterator, "invariant") ;
}
if (Policy == 0) { // prepend to EntryList
if (List == NULL) {
iterator->_next = iterator->_prev = NULL ;
_EntryList = iterator ;
} else {
List->_prev = iterator ;
iterator->_next = List ;
iterator->_prev = NULL ;
_EntryList = iterator ;
}
} else
if (Policy == 1) { // append to EntryList
if (List == NULL) {
iterator->_next = iterator->_prev = NULL ;
_EntryList = iterator ;
} else {
// CONSIDER: finding the tail currently requires a linear-time walk of
// the EntryList. We can make tail access constant-time by converting to
// a CDLL instead of using our current DLL.
ObjectWaiter * Tail ;
for (Tail = List ; Tail->_next != NULL ; Tail = Tail->_next) ;
assert (Tail != NULL && Tail->_next == NULL, "invariant") ;
Tail->_next = iterator ;
iterator->_prev = Tail ;
iterator->_next = NULL ;
}
} else
if (Policy == 2) { // prepend to cxq
// prepend to cxq
iterator->TState = ObjectWaiter::TS_CXQ ;
for (;;) {
ObjectWaiter * Front = _cxq ;
iterator->_next = Front ;
if (Atomic::cmpxchg_ptr (iterator, &_cxq, Front) == Front) {
break ;
}
}
} else
if (Policy == 3) { // append to cxq
iterator->TState = ObjectWaiter::TS_CXQ ;
for (;;) {
ObjectWaiter * Tail ;
Tail = _cxq ;
if (Tail == NULL) {
iterator->_next = NULL ;
if (Atomic::cmpxchg_ptr (iterator, &_cxq, NULL) == NULL) {
break ;
}
} else {
while (Tail->_next != NULL) Tail = Tail->_next ;
Tail->_next = iterator ;
iterator->_prev = Tail ;
iterator->_next = NULL ;
break ;
}
}
} else {
ParkEvent * ev = iterator->_event ;
iterator->TState = ObjectWaiter::TS_RUN ;
OrderAccess::fence() ;
ev->unpark() ;
}
if (Policy < 4) {
iterator->wait_reenter_begin(this);
}
// _WaitSetLock protects the wait queue, not the EntryList. We could
// move the add-to-EntryList operation, above, outside the critical section
// protected by _WaitSetLock. In practice that's not useful. With the
// exception of wait() timeouts and interrupts the monitor owner
// is the only thread that grabs _WaitSetLock. There's almost no contention
// on _WaitSetLock so it's not profitable to reduce the length of the
// critical section.
}
Thread::SpinRelease (&_WaitSetLock) ;
if (Tally != 0 && ObjectMonitor::_sync_Notifications != NULL) {
ObjectMonitor::_sync_Notifications->inc(Tally) ;
}
}
// -----------------------------------------------------------------------------
// Adaptive Spinning Support
//
// Adaptive spin-then-block - rational spinning
//
// Note that we spin "globally" on _owner with a classic SMP-polite TATAS
// algorithm. On high order SMP systems it would be better to start with
// a brief global spin and then revert to spinning locally. In the spirit of MCS/CLH,
// a contending thread could enqueue itself on the cxq and then spin locally
// on a thread-specific variable such as its ParkEvent._Event flag.
// That's left as an exercise for the reader. Note that global spinning is
// not problematic on Niagara, as the L2$ serves the interconnect and has both
// low latency and massive bandwidth.
//
// Broadly, we can fix the spin frequency -- that is, the % of contended lock
// acquisition attempts where we opt to spin -- at 100% and vary the spin count
// (duration) or we can fix the count at approximately the duration of
// a context switch and vary the frequency. Of course we could also
// vary both satisfying K == Frequency * Duration, where K is adaptive by monitor.
// See http://j2se.east/~dice/PERSIST/040824-AdaptiveSpinning.html.
//
// This implementation varies the duration "D", where D varies with
// the success rate of recent spin attempts. (D is capped at approximately
// length of a round-trip context switch). The success rate for recent
// spin attempts is a good predictor of the success rate of future spin
// attempts. The mechanism adapts automatically to varying critical
// section length (lock modality), system load and degree of parallelism.
// D is maintained per-monitor in _SpinDuration and is initialized
// optimistically. Spin frequency is fixed at 100%.
//
// Note that _SpinDuration is volatile, but we update it without locks
// or atomics. The code is designed so that _SpinDuration stays within
// a reasonable range even in the presence of races. The arithmetic
// operations on _SpinDuration are closed over the domain of legal values,
// so at worst a race will install and older but still legal value.
// At the very worst this introduces some apparent non-determinism.
// We might spin when we shouldn't or vice-versa, but since the spin
// count are relatively short, even in the worst case, the effect is harmless.
//
// Care must be taken that a low "D" value does not become an
// an absorbing state. Transient spinning failures -- when spinning
// is overall profitable -- should not cause the system to converge
// on low "D" values. We want spinning to be stable and predictable
// and fairly responsive to change and at the same time we don't want
// it to oscillate, become metastable, be "too" non-deterministic,
// or converge on or enter undesirable stable absorbing states.
//
// We implement a feedback-based control system -- using past behavior
// to predict future behavior. We face two issues: (a) if the
// input signal is random then the spin predictor won't provide optimal
// results, and (b) if the signal frequency is too high then the control
// system, which has some natural response lag, will "chase" the signal.
// (b) can arise from multimodal lock hold times. Transient preemption
// can also result in apparent bimodal lock hold times.
// Although sub-optimal, neither condition is particularly harmful, as
// in the worst-case we'll spin when we shouldn't or vice-versa.
// The maximum spin duration is rather short so the failure modes aren't bad.
// To be conservative, I've tuned the gain in system to bias toward
// _not spinning. Relatedly, the system can sometimes enter a mode where it
// "rings" or oscillates between spinning and not spinning. This happens
// when spinning is just on the cusp of profitability, however, so the
// situation is not dire. The state is benign -- there's no need to add
// hysteresis control to damp the transition rate between spinning and
// not spinning.
//
intptr_t ObjectMonitor::SpinCallbackArgument = 0 ;
int (*ObjectMonitor::SpinCallbackFunction)(intptr_t, int) = NULL ;
// Spinning: Fixed frequency (100%), vary duration
int ObjectMonitor::TrySpin_VaryDuration (Thread * Self) {
// Dumb, brutal spin. Good for comparative measurements against adaptive spinning.
int ctr = Knob_FixedSpin ;
if (ctr != 0) {
while (--ctr >= 0) {
if (TryLock (Self) > 0) return 1 ;
SpinPause () ;
}
return 0 ;
}
for (ctr = Knob_PreSpin + 1; --ctr >= 0 ; ) {
if (TryLock(Self) > 0) {
// Increase _SpinDuration ...
// Note that we don't clamp SpinDuration precisely at SpinLimit.
// Raising _SpurDuration to the poverty line is key.
int x = _SpinDuration ;
if (x < Knob_SpinLimit) {
if (x < Knob_Poverty) x = Knob_Poverty ;
_SpinDuration = x + Knob_BonusB ;
}
return 1 ;
}
SpinPause () ;
}
// Admission control - verify preconditions for spinning
//
// We always spin a little bit, just to prevent _SpinDuration == 0 from
// becoming an absorbing state. Put another way, we spin briefly to
// sample, just in case the system load, parallelism, contention, or lock
// modality changed.
//
// Consider the following alternative:
// Periodically set _SpinDuration = _SpinLimit and try a long/full
// spin attempt. "Periodically" might mean after a tally of
// the # of failed spin attempts (or iterations) reaches some threshold.
// This takes us into the realm of 1-out-of-N spinning, where we
// hold the duration constant but vary the frequency.
ctr = _SpinDuration ;
if (ctr < Knob_SpinBase) ctr = Knob_SpinBase ;
if (ctr <= 0) return 0 ;
if (Knob_SuccRestrict && _succ != NULL) return 0 ;
if (Knob_OState && NotRunnable (Self, (Thread *) _owner)) {
TEVENT (Spin abort - notrunnable [TOP]);
return 0 ;
}
int MaxSpin = Knob_MaxSpinners ;
if (MaxSpin >= 0) {
if (_Spinner > MaxSpin) {
TEVENT (Spin abort -- too many spinners) ;
return 0 ;
}
// Slighty racy, but benign ...
Adjust (&_Spinner, 1) ;
}
// We're good to spin ... spin ingress.
// CONSIDER: use Prefetch::write() to avoid RTS->RTO upgrades
// when preparing to LD...CAS _owner, etc and the CAS is likely
// to succeed.
int hits = 0 ;
int msk = 0 ;
int caspty = Knob_CASPenalty ;
int oxpty = Knob_OXPenalty ;
int sss = Knob_SpinSetSucc ;
if (sss && _succ == NULL ) _succ = Self ;
Thread * prv = NULL ;
// There are three ways to exit the following loop:
// 1. A successful spin where this thread has acquired the lock.
// 2. Spin failure with prejudice
// 3. Spin failure without prejudice
while (--ctr >= 0) {
// Periodic polling -- Check for pending GC
// Threads may spin while they're unsafe.
// We don't want spinning threads to delay the JVM from reaching
// a stop-the-world safepoint or to steal cycles from GC.
// If we detect a pending safepoint we abort in order that
// (a) this thread, if unsafe, doesn't delay the safepoint, and (b)
// this thread, if safe, doesn't steal cycles from GC.
// This is in keeping with the "no loitering in runtime" rule.
// We periodically check to see if there's a safepoint pending.
if ((ctr & 0xFF) == 0) {
if (SafepointSynchronize::do_call_back()) {
TEVENT (Spin: safepoint) ;
goto Abort ; // abrupt spin egress
}
if (Knob_UsePause & 1) SpinPause () ;
int (*scb)(intptr_t,int) = SpinCallbackFunction ;
if (hits > 50 && scb != NULL) {
int abend = (*scb)(SpinCallbackArgument, 0) ;
}
}
if (Knob_UsePause & 2) SpinPause() ;
// Exponential back-off ... Stay off the bus to reduce coherency traffic.
// This is useful on classic SMP systems, but is of less utility on
// N1-style CMT platforms.
//
// Trade-off: lock acquisition latency vs coherency bandwidth.
// Lock hold times are typically short. A histogram
// of successful spin attempts shows that we usually acquire
// the lock early in the spin. That suggests we want to
// sample _owner frequently in the early phase of the spin,
// but then back-off and sample less frequently as the spin
// progresses. The back-off makes a good citizen on SMP big
// SMP systems. Oversampling _owner can consume excessive
// coherency bandwidth. Relatedly, if we _oversample _owner we
// can inadvertently interfere with the the ST m->owner=null.
// executed by the lock owner.
if (ctr & msk) continue ;
++hits ;
if ((hits & 0xF) == 0) {
// The 0xF, above, corresponds to the exponent.
// Consider: (msk+1)|msk
msk = ((msk << 2)|3) & BackOffMask ;
}
// Probe _owner with TATAS
// If this thread observes the monitor transition or flicker
// from locked to unlocked to locked, then the odds that this
// thread will acquire the lock in this spin attempt go down
// considerably. The same argument applies if the CAS fails
// or if we observe _owner change from one non-null value to
// another non-null value. In such cases we might abort
// the spin without prejudice or apply a "penalty" to the
// spin count-down variable "ctr", reducing it by 100, say.
Thread * ox = (Thread *) _owner ;
if (ox == NULL) {
ox = (Thread *) Atomic::cmpxchg_ptr (Self, &_owner, NULL) ;
if (ox == NULL) {
// The CAS succeeded -- this thread acquired ownership
// Take care of some bookkeeping to exit spin state.
if (sss && _succ == Self) {
_succ = NULL ;
}
if (MaxSpin > 0) Adjust (&_Spinner, -1) ;
// Increase _SpinDuration :
// The spin was successful (profitable) so we tend toward
// longer spin attempts in the future.
// CONSIDER: factor "ctr" into the _SpinDuration adjustment.
// If we acquired the lock early in the spin cycle it
// makes sense to increase _SpinDuration proportionally.
// Note that we don't clamp SpinDuration precisely at SpinLimit.
int x = _SpinDuration ;
if (x < Knob_SpinLimit) {
if (x < Knob_Poverty) x = Knob_Poverty ;
_SpinDuration = x + Knob_Bonus ;
}
return 1 ;
}
// The CAS failed ... we can take any of the following actions:
// * penalize: ctr -= Knob_CASPenalty
// * exit spin with prejudice -- goto Abort;
// * exit spin without prejudice.
// * Since CAS is high-latency, retry again immediately.
prv = ox ;
TEVENT (Spin: cas failed) ;
if (caspty == -2) break ;
if (caspty == -1) goto Abort ;
ctr -= caspty ;
continue ;
}
// Did lock ownership change hands ?
if (ox != prv && prv != NULL ) {
TEVENT (spin: Owner changed)
if (oxpty == -2) break ;
if (oxpty == -1) goto Abort ;
ctr -= oxpty ;
}
prv = ox ;
// Abort the spin if the owner is not executing.
// The owner must be executing in order to drop the lock.
// Spinning while the owner is OFFPROC is idiocy.
// Consider: ctr -= RunnablePenalty ;
if (Knob_OState && NotRunnable (Self, ox)) {
TEVENT (Spin abort - notrunnable);
goto Abort ;
}
if (sss && _succ == NULL ) _succ = Self ;
}
// Spin failed with prejudice -- reduce _SpinDuration.
// TODO: Use an AIMD-like policy to adjust _SpinDuration.
// AIMD is globally stable.
TEVENT (Spin failure) ;
{
int x = _SpinDuration ;
if (x > 0) {
// Consider an AIMD scheme like: x -= (x >> 3) + 100
// This is globally sample and tends to damp the response.
x -= Knob_Penalty ;
if (x < 0) x = 0 ;
_SpinDuration = x ;
}
}
Abort:
if (MaxSpin >= 0) Adjust (&_Spinner, -1) ;
if (sss && _succ == Self) {
_succ = NULL ;
// Invariant: after setting succ=null a contending thread
// must recheck-retry _owner before parking. This usually happens
// in the normal usage of TrySpin(), but it's safest
// to make TrySpin() as foolproof as possible.
OrderAccess::fence() ;
if (TryLock(Self) > 0) return 1 ;
}
return 0 ;
}
// NotRunnable() -- informed spinning
//
// Don't bother spinning if the owner is not eligible to drop the lock.
// Peek at the owner's schedctl.sc_state and Thread._thread_values and
// spin only if the owner thread is _thread_in_Java or _thread_in_vm.
// The thread must be runnable in order to drop the lock in timely fashion.
// If the _owner is not runnable then spinning will not likely be
// successful (profitable).
//
// Beware -- the thread referenced by _owner could have died
// so a simply fetch from _owner->_thread_state might trap.
// Instead, we use SafeFetchXX() to safely LD _owner->_thread_state.
// Because of the lifecycle issues the schedctl and _thread_state values
// observed by NotRunnable() might be garbage. NotRunnable must
// tolerate this and consider the observed _thread_state value
// as advisory.
//
// Beware too, that _owner is sometimes a BasicLock address and sometimes
// a thread pointer. We differentiate the two cases with OwnerIsThread.
// Alternately, we might tag the type (thread pointer vs basiclock pointer)
// with the LSB of _owner. Another option would be to probablistically probe
// the putative _owner->TypeTag value.
//
// Checking _thread_state isn't perfect. Even if the thread is
// in_java it might be blocked on a page-fault or have been preempted
// and sitting on a ready/dispatch queue. _thread state in conjunction
// with schedctl.sc_state gives us a good picture of what the
// thread is doing, however.
//
// TODO: check schedctl.sc_state.
// We'll need to use SafeFetch32() to read from the schedctl block.
// See RFE #5004247 and http://sac.sfbay.sun.com/Archives/CaseLog/arc/PSARC/2005/351/
//
// The return value from NotRunnable() is *advisory* -- the
// result is based on sampling and is not necessarily coherent.
// The caller must tolerate false-negative and false-positive errors.
// Spinning, in general, is probabilistic anyway.
int ObjectMonitor::NotRunnable (Thread * Self, Thread * ox) {
// Check either OwnerIsThread or ox->TypeTag == 2BAD.
if (!OwnerIsThread) return 0 ;
if (ox == NULL) return 0 ;
// Avoid transitive spinning ...
// Say T1 spins or blocks trying to acquire L. T1._Stalled is set to L.
// Immediately after T1 acquires L it's possible that T2, also
// spinning on L, will see L.Owner=T1 and T1._Stalled=L.
// This occurs transiently after T1 acquired L but before
// T1 managed to clear T1.Stalled. T2 does not need to abort
// its spin in this circumstance.
intptr_t BlockedOn = SafeFetchN ((intptr_t *) &ox->_Stalled, intptr_t(1)) ;
if (BlockedOn == 1) return 1 ;
if (BlockedOn != 0) {
return BlockedOn != intptr_t(this) && _owner == ox ;
}
assert (sizeof(((JavaThread *)ox)->_thread_state == sizeof(int)), "invariant") ;
int jst = SafeFetch32 ((int *) &((JavaThread *) ox)->_thread_state, -1) ; ;
// consider also: jst != _thread_in_Java -- but that's overspecific.
return jst == _thread_blocked || jst == _thread_in_native ;
}
// -----------------------------------------------------------------------------
// WaitSet management ...
ObjectWaiter::ObjectWaiter(Thread* thread) {
_next = NULL;
_prev = NULL;
_notified = 0;
TState = TS_RUN ;
_thread = thread;
_event = thread->_ParkEvent ;
_active = false;
assert (_event != NULL, "invariant") ;
}
void ObjectWaiter::wait_reenter_begin(ObjectMonitor *mon) {
JavaThread *jt = (JavaThread *)this->_thread;
_active = JavaThreadBlockedOnMonitorEnterState::wait_reenter_begin(jt, mon);
}
void ObjectWaiter::wait_reenter_end(ObjectMonitor *mon) {
JavaThread *jt = (JavaThread *)this->_thread;
JavaThreadBlockedOnMonitorEnterState::wait_reenter_end(jt, _active);
}
inline void ObjectMonitor::AddWaiter(ObjectWaiter* node) {
assert(node != NULL, "should not dequeue NULL node");
assert(node->_prev == NULL, "node already in list");
assert(node->_next == NULL, "node already in list");
// put node at end of queue (circular doubly linked list)
if (_WaitSet == NULL) {
_WaitSet = node;
node->_prev = node;
node->_next = node;
} else {
ObjectWaiter* head = _WaitSet ;
ObjectWaiter* tail = head->_prev;
assert(tail->_next == head, "invariant check");
tail->_next = node;
head->_prev = node;
node->_next = head;
node->_prev = tail;
}
}
inline ObjectWaiter* ObjectMonitor::DequeueWaiter() {
// dequeue the very first waiter
ObjectWaiter* waiter = _WaitSet;
if (waiter) {
DequeueSpecificWaiter(waiter);
}
return waiter;
}
inline void ObjectMonitor::DequeueSpecificWaiter(ObjectWaiter* node) {
assert(node != NULL, "should not dequeue NULL node");
assert(node->_prev != NULL, "node already removed from list");
assert(node->_next != NULL, "node already removed from list");
// when the waiter has woken up because of interrupt,
// timeout or other spurious wake-up, dequeue the
// waiter from waiting list
ObjectWaiter* next = node->_next;
if (next == node) {
assert(node->_prev == node, "invariant check");
_WaitSet = NULL;
} else {
ObjectWaiter* prev = node->_prev;
assert(prev->_next == node, "invariant check");
assert(next->_prev == node, "invariant check");
next->_prev = prev;
prev->_next = next;
if (_WaitSet == node) {
_WaitSet = next;
}
}
node->_next = NULL;
node->_prev = NULL;
}
// -----------------------------------------------------------------------------
// PerfData support
PerfCounter * ObjectMonitor::_sync_ContendedLockAttempts = NULL ;
PerfCounter * ObjectMonitor::_sync_FutileWakeups = NULL ;
PerfCounter * ObjectMonitor::_sync_Parks = NULL ;
PerfCounter * ObjectMonitor::_sync_EmptyNotifications = NULL ;
PerfCounter * ObjectMonitor::_sync_Notifications = NULL ;
PerfCounter * ObjectMonitor::_sync_PrivateA = NULL ;
PerfCounter * ObjectMonitor::_sync_PrivateB = NULL ;
PerfCounter * ObjectMonitor::_sync_SlowExit = NULL ;
PerfCounter * ObjectMonitor::_sync_SlowEnter = NULL ;
PerfCounter * ObjectMonitor::_sync_SlowNotify = NULL ;
PerfCounter * ObjectMonitor::_sync_SlowNotifyAll = NULL ;
PerfCounter * ObjectMonitor::_sync_FailedSpins = NULL ;
PerfCounter * ObjectMonitor::_sync_SuccessfulSpins = NULL ;
PerfCounter * ObjectMonitor::_sync_MonInCirculation = NULL ;
PerfCounter * ObjectMonitor::_sync_MonScavenged = NULL ;
PerfCounter * ObjectMonitor::_sync_Inflations = NULL ;
PerfCounter * ObjectMonitor::_sync_Deflations = NULL ;
PerfLongVariable * ObjectMonitor::_sync_MonExtant = NULL ;
// One-shot global initialization for the sync subsystem.
// We could also defer initialization and initialize on-demand
// the first time we call inflate(). Initialization would
// be protected - like so many things - by the MonitorCache_lock.
void ObjectMonitor::Initialize () {
static int InitializationCompleted = 0 ;
assert (InitializationCompleted == 0, "invariant") ;
InitializationCompleted = 1 ;
if (UsePerfData) {
EXCEPTION_MARK ;
#define NEWPERFCOUNTER(n) {n = PerfDataManager::create_counter(SUN_RT, #n, PerfData::U_Events,CHECK); }
#define NEWPERFVARIABLE(n) {n = PerfDataManager::create_variable(SUN_RT, #n, PerfData::U_Events,CHECK); }
NEWPERFCOUNTER(_sync_Inflations) ;
NEWPERFCOUNTER(_sync_Deflations) ;
NEWPERFCOUNTER(_sync_ContendedLockAttempts) ;
NEWPERFCOUNTER(_sync_FutileWakeups) ;
NEWPERFCOUNTER(_sync_Parks) ;
NEWPERFCOUNTER(_sync_EmptyNotifications) ;
NEWPERFCOUNTER(_sync_Notifications) ;
NEWPERFCOUNTER(_sync_SlowEnter) ;
NEWPERFCOUNTER(_sync_SlowExit) ;
NEWPERFCOUNTER(_sync_SlowNotify) ;
NEWPERFCOUNTER(_sync_SlowNotifyAll) ;
NEWPERFCOUNTER(_sync_FailedSpins) ;
NEWPERFCOUNTER(_sync_SuccessfulSpins) ;
NEWPERFCOUNTER(_sync_PrivateA) ;
NEWPERFCOUNTER(_sync_PrivateB) ;
NEWPERFCOUNTER(_sync_MonInCirculation) ;
NEWPERFCOUNTER(_sync_MonScavenged) ;
NEWPERFVARIABLE(_sync_MonExtant) ;
#undef NEWPERFCOUNTER
}
}
// Compile-time asserts
// When possible, it's better to catch errors deterministically at
// compile-time than at runtime. The down-side to using compile-time
// asserts is that error message -- often something about negative array
// indices -- is opaque.
#define CTASSERT(x) { int tag[1-(2*!(x))]; printf ("Tag @" INTPTR_FORMAT "\n", (intptr_t)tag); }
void ObjectMonitor::ctAsserts() {
CTASSERT(offset_of (ObjectMonitor, _header) == 0);
}
static char * kvGet (char * kvList, const char * Key) {
if (kvList == NULL) return NULL ;
size_t n = strlen (Key) ;
char * Search ;
for (Search = kvList ; *Search ; Search += strlen(Search) + 1) {
if (strncmp (Search, Key, n) == 0) {
if (Search[n] == '=') return Search + n + 1 ;
if (Search[n] == 0) return (char *) "1" ;
}
}
return NULL ;
}
static int kvGetInt (char * kvList, const char * Key, int Default) {
char * v = kvGet (kvList, Key) ;
int rslt = v ? ::strtol (v, NULL, 0) : Default ;
if (Knob_ReportSettings && v != NULL) {
::printf (" SyncKnob: %s %d(%d)\n", Key, rslt, Default) ;
::fflush (stdout) ;
}
return rslt ;
}
void ObjectMonitor::DeferredInitialize () {
if (InitDone > 0) return ;
if (Atomic::cmpxchg (-1, &InitDone, 0) != 0) {
while (InitDone != 1) ;
return ;
}
// One-shot global initialization ...
// The initialization is idempotent, so we don't need locks.
// In the future consider doing this via os::init_2().
// SyncKnobs consist of <Key>=<Value> pairs in the style
// of environment variables. Start by converting ':' to NUL.
if (SyncKnobs == NULL) SyncKnobs = "" ;
size_t sz = strlen (SyncKnobs) ;
char * knobs = (char *) malloc (sz + 2) ;
if (knobs == NULL) {
vm_exit_out_of_memory (sz + 2, "Parse SyncKnobs") ;
guarantee (0, "invariant") ;
}
strcpy (knobs, SyncKnobs) ;
knobs[sz+1] = 0 ;
for (char * p = knobs ; *p ; p++) {
if (*p == ':') *p = 0 ;
}
#define SETKNOB(x) { Knob_##x = kvGetInt (knobs, #x, Knob_##x); }
SETKNOB(ReportSettings) ;
SETKNOB(Verbose) ;
SETKNOB(FixedSpin) ;
SETKNOB(SpinLimit) ;
SETKNOB(SpinBase) ;
SETKNOB(SpinBackOff);
SETKNOB(CASPenalty) ;
SETKNOB(OXPenalty) ;
SETKNOB(LogSpins) ;
SETKNOB(SpinSetSucc) ;
SETKNOB(SuccEnabled) ;
SETKNOB(SuccRestrict) ;
SETKNOB(Penalty) ;
SETKNOB(Bonus) ;
SETKNOB(BonusB) ;
SETKNOB(Poverty) ;
SETKNOB(SpinAfterFutile) ;
SETKNOB(UsePause) ;
SETKNOB(SpinEarly) ;
SETKNOB(OState) ;
SETKNOB(MaxSpinners) ;
SETKNOB(PreSpin) ;
SETKNOB(ExitPolicy) ;
SETKNOB(QMode);
SETKNOB(ResetEvent) ;
SETKNOB(MoveNotifyee) ;
SETKNOB(FastHSSEC) ;
#undef SETKNOB
if (os::is_MP()) {
BackOffMask = (1 << Knob_SpinBackOff) - 1 ;
if (Knob_ReportSettings) ::printf ("BackOffMask=%X\n", BackOffMask) ;
// CONSIDER: BackOffMask = ROUNDUP_NEXT_POWER2 (ncpus-1)
} else {
Knob_SpinLimit = 0 ;
Knob_SpinBase = 0 ;
Knob_PreSpin = 0 ;
Knob_FixedSpin = -1 ;
}
if (Knob_LogSpins == 0) {
ObjectMonitor::_sync_FailedSpins = NULL ;
}
free (knobs) ;
OrderAccess::fence() ;
InitDone = 1 ;
}
#ifndef PRODUCT
void ObjectMonitor::verify() {
}
void ObjectMonitor::print() {
}
#endif
src/share/vm/runtime/objectMonitor.hpp
浏览文件 @ f9bc752b
...@@ -22,6 +22,32 @@ ...@@ -22,6 +22,32 @@
* *
*/ */
// ObjectWaiter serves as a "proxy" or surrogate thread.
// TODO-FIXME: Eliminate ObjectWaiter and use the thread-specific
// ParkEvent instead. Beware, however, that the JVMTI code
// knows about ObjectWaiters, so we'll have to reconcile that code.
// See next_waiter(), first_waiter(), etc.
class ObjectWaiter : public StackObj {
public:
enum TStates { TS_UNDEF, TS_READY, TS_RUN, TS_WAIT, TS_ENTER, TS_CXQ } ;
enum Sorted { PREPEND, APPEND, SORTED } ;
ObjectWaiter * volatile _next;
ObjectWaiter * volatile _prev;
Thread* _thread;
ParkEvent * _event;
volatile int _notified ;
volatile TStates TState ;
Sorted _Sorted ; // List placement disposition
bool _active ; // Contention monitoring is enabled
public:
ObjectWaiter(Thread* thread);
void wait_reenter_begin(ObjectMonitor *mon);
void wait_reenter_end(ObjectMonitor *mon);
};
// WARNING: // WARNING:
// This is a very sensitive and fragile class. DO NOT make any // This is a very sensitive and fragile class. DO NOT make any
// change unless you are fully aware of the underlying semantics. // change unless you are fully aware of the underlying semantics.
...@@ -38,8 +64,6 @@ ...@@ -38,8 +64,6 @@
// It is also used as RawMonitor by the JVMTI // It is also used as RawMonitor by the JVMTI
class ObjectWaiter;
class ObjectMonitor { class ObjectMonitor {
public: public:
enum { enum {
...@@ -74,13 +98,16 @@ class ObjectMonitor { ...@@ -74,13 +98,16 @@ class ObjectMonitor {
public: public:
ObjectMonitor();
~ObjectMonitor();
markOop header() const; markOop header() const;
void set_header(markOop hdr); void set_header(markOop hdr);
intptr_t is_busy() const; intptr_t is_busy() const {
// TODO-FIXME: merge _count and _waiters.
// TODO-FIXME: assert _owner == null implies _recursions = 0
// TODO-FIXME: assert _WaitSet != null implies _count > 0
return _count|_waiters|intptr_t(_owner)|intptr_t(_cxq)|intptr_t(_EntryList ) ;
}
intptr_t is_entered(Thread* current) const; intptr_t is_entered(Thread* current) const;
void* owner() const; void* owner() const;
...@@ -91,13 +118,58 @@ class ObjectMonitor { ...@@ -91,13 +118,58 @@ class ObjectMonitor {
intptr_t count() const; intptr_t count() const;
void set_count(intptr_t count); void set_count(intptr_t count);
intptr_t contentions() const ; intptr_t contentions() const ;
intptr_t recursions() const { return _recursions; }
// JVM/DI GetMonitorInfo() needs this // JVM/DI GetMonitorInfo() needs this
Thread * thread_of_waiter (ObjectWaiter *) ; ObjectWaiter* first_waiter() { return _WaitSet; }
ObjectWaiter * first_waiter () ; ObjectWaiter* next_waiter(ObjectWaiter* o) { return o->_next; }
ObjectWaiter * next_waiter(ObjectWaiter* o); Thread* thread_of_waiter(ObjectWaiter* o) { return o->_thread; }
intptr_t recursions() const { return _recursions; } // initialize the monitor, exception the semaphore, all other fields
// are simple integers or pointers
ObjectMonitor() {
_header = NULL;
_count = 0;
_waiters = 0,
_recursions = 0;
_object = NULL;
_owner = NULL;
_WaitSet = NULL;
_WaitSetLock = 0 ;
_Responsible = NULL ;
_succ = NULL ;
_cxq = NULL ;
FreeNext = NULL ;
_EntryList = NULL ;
_SpinFreq = 0 ;
_SpinClock = 0 ;
OwnerIsThread = 0 ;
}
~ObjectMonitor() {
// TODO: Add asserts ...
// _cxq == 0 _succ == NULL _owner == NULL _waiters == 0
// _count == 0 _EntryList == NULL etc
}
private:
void Recycle () {
// TODO: add stronger asserts ...
// _cxq == 0 _succ == NULL _owner == NULL _waiters == 0
// _count == 0 EntryList == NULL
// _recursions == 0 _WaitSet == NULL
// TODO: assert (is_busy()|_recursions) == 0
_succ = NULL ;
_EntryList = NULL ;
_cxq = NULL ;
_WaitSet = NULL ;
_recursions = 0 ;
_SpinFreq = 0 ;
_SpinClock = 0 ;
OwnerIsThread = 0 ;
}
public:
void* object() const; void* object() const;
void* object_addr(); void* object_addr();
...@@ -122,22 +194,9 @@ class ObjectMonitor { ...@@ -122,22 +194,9 @@ class ObjectMonitor {
intptr_t complete_exit(TRAPS); intptr_t complete_exit(TRAPS);
void reenter(intptr_t recursions, TRAPS); void reenter(intptr_t recursions, TRAPS);
int raw_enter(TRAPS);
int raw_exit(TRAPS);
int raw_wait(jlong millis, bool interruptable, TRAPS);
int raw_notify(TRAPS);
int raw_notifyAll(TRAPS);
private: private:
// JVMTI support -- remove ASAP
int SimpleEnter (Thread * Self) ;
int SimpleExit (Thread * Self) ;
int SimpleWait (Thread * Self, jlong millis) ;
int SimpleNotify (Thread * Self, bool All) ;
private:
void Recycle () ;
void AddWaiter (ObjectWaiter * waiter) ; void AddWaiter (ObjectWaiter * waiter) ;
static void DeferredInitialize();
ObjectWaiter * DequeueWaiter () ; ObjectWaiter * DequeueWaiter () ;
void DequeueSpecificWaiter (ObjectWaiter * waiter) ; void DequeueSpecificWaiter (ObjectWaiter * waiter) ;
...@@ -172,13 +231,17 @@ class ObjectMonitor { ...@@ -172,13 +231,17 @@ class ObjectMonitor {
// The VM assumes write ordering wrt these fields, which can be // The VM assumes write ordering wrt these fields, which can be
// read from other threads. // read from other threads.
protected: // protected for jvmtiRawMonitor
void * volatile _owner; // pointer to owning thread OR BasicLock void * volatile _owner; // pointer to owning thread OR BasicLock
volatile intptr_t _recursions; // recursion count, 0 for first entry volatile intptr_t _recursions; // recursion count, 0 for first entry
private:
int OwnerIsThread ; // _owner is (Thread *) vs SP/BasicLock int OwnerIsThread ; // _owner is (Thread *) vs SP/BasicLock
ObjectWaiter * volatile _cxq ; // LL of recently-arrived threads blocked on entry. ObjectWaiter * volatile _cxq ; // LL of recently-arrived threads blocked on entry.
// The list is actually composed of WaitNodes, acting // The list is actually composed of WaitNodes, acting
// as proxies for Threads. // as proxies for Threads.
protected:
ObjectWaiter * volatile _EntryList ; // Threads blocked on entry or reentry. ObjectWaiter * volatile _EntryList ; // Threads blocked on entry or reentry.
private:
Thread * volatile _succ ; // Heir presumptive thread - used for futile wakeup throttling Thread * volatile _succ ; // Heir presumptive thread - used for futile wakeup throttling
Thread * volatile _Responsible ; Thread * volatile _Responsible ;
int _PromptDrain ; // rqst to drain cxq into EntryList ASAP int _PromptDrain ; // rqst to drain cxq into EntryList ASAP
...@@ -196,8 +259,12 @@ class ObjectMonitor { ...@@ -196,8 +259,12 @@ class ObjectMonitor {
volatile intptr_t _count; // reference count to prevent reclaimation/deflation volatile intptr_t _count; // reference count to prevent reclaimation/deflation
// at stop-the-world time. See deflate_idle_monitors(). // at stop-the-world time. See deflate_idle_monitors().
// _count is approximately |_WaitSet| + |_EntryList| // _count is approximately |_WaitSet| + |_EntryList|
protected:
volatile intptr_t _waiters; // number of waiting threads volatile intptr_t _waiters; // number of waiting threads
private:
protected:
ObjectWaiter * volatile _WaitSet; // LL of threads wait()ing on the monitor ObjectWaiter * volatile _WaitSet; // LL of threads wait()ing on the monitor
private:
volatile int _WaitSetLock; // protects Wait Queue - simple spinlock volatile int _WaitSetLock; // protects Wait Queue - simple spinlock
public: public:
...@@ -205,4 +272,37 @@ class ObjectMonitor { ...@@ -205,4 +272,37 @@ class ObjectMonitor {
ObjectMonitor * FreeNext ; // Free list linkage ObjectMonitor * FreeNext ; // Free list linkage
intptr_t StatA, StatsB ; intptr_t StatA, StatsB ;
public:
static void Initialize () ;
static PerfCounter * _sync_ContendedLockAttempts ;
static PerfCounter * _sync_FutileWakeups ;
static PerfCounter * _sync_Parks ;
static PerfCounter * _sync_EmptyNotifications ;
static PerfCounter * _sync_Notifications ;
static PerfCounter * _sync_SlowEnter ;
static PerfCounter * _sync_SlowExit ;
static PerfCounter * _sync_SlowNotify ;
static PerfCounter * _sync_SlowNotifyAll ;
static PerfCounter * _sync_FailedSpins ;
static PerfCounter * _sync_SuccessfulSpins ;
static PerfCounter * _sync_PrivateA ;
static PerfCounter * _sync_PrivateB ;
static PerfCounter * _sync_MonInCirculation ;
static PerfCounter * _sync_MonScavenged ;
static PerfCounter * _sync_Inflations ;
static PerfCounter * _sync_Deflations ;
static PerfLongVariable * _sync_MonExtant ;
public:
static int Knob_Verbose;
static int Knob_SpinLimit;
}; };
#undef TEVENT
#define TEVENT(nom) {if (SyncVerbose) FEVENT(nom); }
#define FEVENT(nom) { static volatile int ctr = 0 ; int v = ++ctr ; if ((v & (v-1)) == 0) { ::printf (#nom " : %d \n", v); ::fflush(stdout); }}
#undef TEVENT
#define TEVENT(nom) {;}
src/share/vm/runtime/objectMonitor.inline.hpp
浏览文件 @ f9bc752b
...@@ -104,7 +104,3 @@ inline void ObjectMonitor::set_owner(void* owner) { ...@@ -104,7 +104,3 @@ inline void ObjectMonitor::set_owner(void* owner) {
_count = 0; _count = 0;
} }
// here are the platform-dependent bodies:
# include "incls/_objectMonitor_pd.inline.hpp.incl"
src/share/vm/runtime/park.cpp 0 → 100644
浏览文件 @ f9bc752b
/*
* Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
# include "incls/_precompiled.incl"
# include "incls/_park.cpp.incl"
// Lifecycle management for TSM ParkEvents.
// ParkEvents are type-stable (TSM).
// In our particular implementation they happen to be immortal.
//
// We manage concurrency on the FreeList with a CAS-based
// detach-modify-reattach idiom that avoids the ABA problems
// that would otherwise be present in a simple CAS-based
// push-pop implementation. (push-one and pop-all)
//
// Caveat: Allocate() and Release() may be called from threads
// other than the thread associated with the Event!
// If we need to call Allocate() when running as the thread in
// question then look for the PD calls to initialize native TLS.
// Native TLS (Win32/Linux/Solaris) can only be initialized or
// accessed by the associated thread.
// See also pd_initialize().
//
// Note that we could defer associating a ParkEvent with a thread
// until the 1st time the thread calls park(). unpark() calls to
// an unprovisioned thread would be ignored. The first park() call
// for a thread would allocate and associate a ParkEvent and return
// immediately.
volatile int ParkEvent::ListLock = 0 ;
ParkEvent * volatile ParkEvent::FreeList = NULL ;
ParkEvent * ParkEvent::Allocate (Thread * t) {
// In rare cases -- JVM_RawMonitor* operations -- we can find t == null.
ParkEvent * ev ;
// Start by trying to recycle an existing but unassociated
// ParkEvent from the global free list.
for (;;) {
ev = FreeList ;
if (ev == NULL) break ;
// 1: Detach - sequester or privatize the list
// Tantamount to ev = Swap (&FreeList, NULL)
if (Atomic::cmpxchg_ptr (NULL, &FreeList, ev) != ev) {
continue ;
}
// We've detached the list. The list in-hand is now
// local to this thread. This thread can operate on the
// list without risk of interference from other threads.
// 2: Extract -- pop the 1st element from the list.
ParkEvent * List = ev->FreeNext ;
if (List == NULL) break ;
for (;;) {
// 3: Try to reattach the residual list
guarantee (List != NULL, "invariant") ;
ParkEvent * Arv = (ParkEvent *) Atomic::cmpxchg_ptr (List, &FreeList, NULL) ;
if (Arv == NULL) break ;
// New nodes arrived. Try to detach the recent arrivals.
if (Atomic::cmpxchg_ptr (NULL, &FreeList, Arv) != Arv) {
continue ;
}
guarantee (Arv != NULL, "invariant") ;
// 4: Merge Arv into List
ParkEvent * Tail = List ;
while (Tail->FreeNext != NULL) Tail = Tail->FreeNext ;
Tail->FreeNext = Arv ;
}
break ;
}
if (ev != NULL) {
guarantee (ev->AssociatedWith == NULL, "invariant") ;
} else {
// Do this the hard way -- materialize a new ParkEvent.
// In rare cases an allocating thread might detach a long list --
// installing null into FreeList -- and then stall or be obstructed.
// A 2nd thread calling Allocate() would see FreeList == null.
// The list held privately by the 1st thread is unavailable to the 2nd thread.
// In that case the 2nd thread would have to materialize a new ParkEvent,
// even though free ParkEvents existed in the system. In this case we end up
// with more ParkEvents in circulation than we need, but the race is
// rare and the outcome is benign. Ideally, the # of extant ParkEvents
// is equal to the maximum # of threads that existed at any one time.
// Because of the race mentioned above, segments of the freelist
// can be transiently inaccessible. At worst we may end up with the
// # of ParkEvents in circulation slightly above the ideal.
// Note that if we didn't have the TSM/immortal constraint, then
// when reattaching, above, we could trim the list.
ev = new ParkEvent () ;
guarantee ((intptr_t(ev) & 0xFF) == 0, "invariant") ;
}
ev->reset() ; // courtesy to caller
ev->AssociatedWith = t ; // Associate ev with t
ev->FreeNext = NULL ;
return ev ;
}
void ParkEvent::Release (ParkEvent * ev) {
if (ev == NULL) return ;
guarantee (ev->FreeNext == NULL , "invariant") ;
ev->AssociatedWith = NULL ;
for (;;) {
// Push ev onto FreeList
// The mechanism is "half" lock-free.
ParkEvent * List = FreeList ;
ev->FreeNext = List ;
if (Atomic::cmpxchg_ptr (ev, &FreeList, List) == List) break ;
}
}
// Override operator new and delete so we can ensure that the
// least significant byte of ParkEvent addresses is 0.
// Beware that excessive address alignment is undesirable
// as it can result in D$ index usage imbalance as
// well as bank access imbalance on Niagara-like platforms,
// although Niagara's hash function should help.
void * ParkEvent::operator new (size_t sz) {
return (void *) ((intptr_t (CHeapObj::operator new (sz + 256)) + 256) & -256) ;
}
void ParkEvent::operator delete (void * a) {
// ParkEvents are type-stable and immortal ...
ShouldNotReachHere();
}
// 6399321 As a temporary measure we copied & modified the ParkEvent::
// allocate() and release() code for use by Parkers. The Parker:: forms
// will eventually be removed as we consolide and shift over to ParkEvents
// for both builtin synchronization and JSR166 operations.
volatile int Parker::ListLock = 0 ;
Parker * volatile Parker::FreeList = NULL ;
Parker * Parker::Allocate (JavaThread * t) {
guarantee (t != NULL, "invariant") ;
Parker * p ;
// Start by trying to recycle an existing but unassociated
// Parker from the global free list.
for (;;) {
p = FreeList ;
if (p == NULL) break ;
// 1: Detach
// Tantamount to p = Swap (&FreeList, NULL)
if (Atomic::cmpxchg_ptr (NULL, &FreeList, p) != p) {
continue ;
}
// We've detached the list. The list in-hand is now
// local to this thread. This thread can operate on the
// list without risk of interference from other threads.
// 2: Extract -- pop the 1st element from the list.
Parker * List = p->FreeNext ;
if (List == NULL) break ;
for (;;) {
// 3: Try to reattach the residual list
guarantee (List != NULL, "invariant") ;
Parker * Arv = (Parker *) Atomic::cmpxchg_ptr (List, &FreeList, NULL) ;
if (Arv == NULL) break ;
// New nodes arrived. Try to detach the recent arrivals.
if (Atomic::cmpxchg_ptr (NULL, &FreeList, Arv) != Arv) {
continue ;
}
guarantee (Arv != NULL, "invariant") ;
// 4: Merge Arv into List
Parker * Tail = List ;
while (Tail->FreeNext != NULL) Tail = Tail->FreeNext ;
Tail->FreeNext = Arv ;
}
break ;
}
if (p != NULL) {
guarantee (p->AssociatedWith == NULL, "invariant") ;
} else {
// Do this the hard way -- materialize a new Parker..
// In rare cases an allocating thread might detach
// a long list -- installing null into FreeList --and
// then stall. Another thread calling Allocate() would see
// FreeList == null and then invoke the ctor. In this case we
// end up with more Parkers in circulation than we need, but
// the race is rare and the outcome is benign.
// Ideally, the # of extant Parkers is equal to the
// maximum # of threads that existed at any one time.
// Because of the race mentioned above, segments of the
// freelist can be transiently inaccessible. At worst
// we may end up with the # of Parkers in circulation
// slightly above the ideal.
p = new Parker() ;
}
p->AssociatedWith = t ; // Associate p with t
p->FreeNext = NULL ;
return p ;
}
void Parker::Release (Parker * p) {
if (p == NULL) return ;
guarantee (p->AssociatedWith != NULL, "invariant") ;
guarantee (p->FreeNext == NULL , "invariant") ;
p->AssociatedWith = NULL ;
for (;;) {
// Push p onto FreeList
Parker * List = FreeList ;
p->FreeNext = List ;
if (Atomic::cmpxchg_ptr (p, &FreeList, List) == List) break ;
}
}
src/share/vm/runtime/park.hpp 0 → 100644
浏览文件 @ f9bc752b
/*
* Copyright (c) 1997, 2010, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
/*
* Per-thread blocking support for JSR166. See the Java-level
* Documentation for rationale. Basically, park acts like wait, unpark
* like notify.
*
* 6271289 --
* To avoid errors where an os thread expires but the JavaThread still
* exists, Parkers are immortal (type-stable) and are recycled across
* new threads. This parallels the ParkEvent implementation.
* Because park-unpark allow spurious wakeups it is harmless if an
* unpark call unparks a new thread using the old Parker reference.
*
* In the future we'll want to think about eliminating Parker and using
* ParkEvent instead. There's considerable duplication between the two
* services.
*
*/
class Parker : public os::PlatformParker {
private:
volatile int _counter ;
Parker * FreeNext ;
JavaThread * AssociatedWith ; // Current association
public:
Parker() : PlatformParker() {
_counter = 0 ;
FreeNext = NULL ;
AssociatedWith = NULL ;
}
protected:
~Parker() { ShouldNotReachHere(); }
public:
// For simplicity of interface with Java, all forms of park (indefinite,
// relative, and absolute) are multiplexed into one call.
void park(bool isAbsolute, jlong time);
void unpark();
// Lifecycle operators
static Parker * Allocate (JavaThread * t) ;
static void Release (Parker * e) ;
private:
static Parker * volatile FreeList ;
static volatile int ListLock ;
};
/////////////////////////////////////////////////////////////
//
// ParkEvents are type-stable and immortal.
//
// Lifecycle: Once a ParkEvent is associated with a thread that ParkEvent remains
// associated with the thread for the thread's entire lifetime - the relationship is
// stable. A thread will be associated at most one ParkEvent. When the thread
// expires, the ParkEvent moves to the EventFreeList. New threads attempt to allocate from
// the EventFreeList before creating a new Event. Type-stability frees us from
// worrying about stale Event or Thread references in the objectMonitor subsystem.
// (A reference to ParkEvent is always valid, even though the event may no longer be associated
// with the desired or expected thread. A key aspect of this design is that the callers of
// park, unpark, etc must tolerate stale references and spurious wakeups).
//
// Only the "associated" thread can block (park) on the ParkEvent, although
// any other thread can unpark a reachable parkevent. Park() is allowed to
// return spuriously. In fact park-unpark a really just an optimization to
// avoid unbounded spinning and surrender the CPU to be a polite system citizen.
// A degenerate albeit "impolite" park-unpark implementation could simply return.
// See http://blogs.sun.com/dave for more details.
//
// Eventually I'd like to eliminate Events and ObjectWaiters, both of which serve as
// thread proxies, and simply make the THREAD structure type-stable and persistent.
// Currently, we unpark events associated with threads, but ideally we'd just
// unpark threads.
//
// The base-class, PlatformEvent, is platform-specific while the ParkEvent is
// platform-independent. PlatformEvent provides park(), unpark(), etc., and
// is abstract -- that is, a PlatformEvent should never be instantiated except
// as part of a ParkEvent.
// Equivalently we could have defined a platform-independent base-class that
// exported Allocate(), Release(), etc. The platform-specific class would extend
// that base-class, adding park(), unpark(), etc.
//
// A word of caution: The JVM uses 2 very similar constructs:
// 1. ParkEvent are used for Java-level "monitor" synchronization.
// 2. Parkers are used by JSR166-JUC park-unpark.
//
// We'll want to eventually merge these redundant facilities and use ParkEvent.
class ParkEvent : public os::PlatformEvent {
private:
ParkEvent * FreeNext ;
// Current association
Thread * AssociatedWith ;
intptr_t RawThreadIdentity ; // LWPID etc
volatile int Incarnation ;
// diagnostic : keep track of last thread to wake this thread.
// this is useful for construction of dependency graphs.
void * LastWaker ;
public:
// MCS-CLH list linkage and Native Mutex/Monitor
ParkEvent * volatile ListNext ;
ParkEvent * volatile ListPrev ;
volatile intptr_t OnList ;
volatile int TState ;
volatile int Notified ; // for native monitor construct
volatile int IsWaiting ; // Enqueued on WaitSet
private:
static ParkEvent * volatile FreeList ;
static volatile int ListLock ;
// It's prudent to mark the dtor as "private"
// ensuring that it's not visible outside the package.
// Unfortunately gcc warns about such usage, so
// we revert to the less desirable "protected" visibility.
// The other compilers accept private dtors.
protected: // Ensure dtor is never invoked
~ParkEvent() { guarantee (0, "invariant") ; }
ParkEvent() : PlatformEvent() {
AssociatedWith = NULL ;
FreeNext = NULL ;
ListNext = NULL ;
ListPrev = NULL ;
OnList = 0 ;
TState = 0 ;
Notified = 0 ;
IsWaiting = 0 ;
}
// We use placement-new to force ParkEvent instances to be
// aligned on 256-byte address boundaries. This ensures that the least
// significant byte of a ParkEvent address is always 0.
void * operator new (size_t sz) ;
void operator delete (void * a) ;
public:
static ParkEvent * Allocate (Thread * t) ;
static void Release (ParkEvent * e) ;
} ;
src/share/vm/runtime/synchronizer.cpp
浏览文件 @ f9bc752b
因为 它太大了无法显示 source diff 。你可以改为 查看blob。
src/share/vm/runtime/synchronizer.hpp
浏览文件 @ f9bc752b
...@@ -22,53 +22,6 @@ ...@@ -22,53 +22,6 @@
* *
*/ */
class BasicLock VALUE_OBJ_CLASS_SPEC {
friend class VMStructs;
private:
volatile markOop _displaced_header;
public:
markOop displaced_header() const { return _displaced_header; }
void set_displaced_header(markOop header) { _displaced_header = header; }
void print_on(outputStream* st) const;
// move a basic lock (used during deoptimization
void move_to(oop obj, BasicLock* dest);
static int displaced_header_offset_in_bytes() { return offset_of(BasicLock, _displaced_header); }
};
// A BasicObjectLock associates a specific Java object with a BasicLock.
// It is currently embedded in an interpreter frame.
// Because some machines have alignment restrictions on the control stack,
// the actual space allocated by the interpreter may include padding words
// after the end of the BasicObjectLock. Also, in order to guarantee
// alignment of the embedded BasicLock objects on such machines, we
// put the embedded BasicLock at the beginning of the struct.
class BasicObjectLock VALUE_OBJ_CLASS_SPEC {
friend class VMStructs;
private:
BasicLock _lock; // the lock, must be double word aligned
oop _obj; // object holds the lock;
public:
// Manipulation
oop obj() const { return _obj; }
void set_obj(oop obj) { _obj = obj; }
BasicLock* lock() { return &_lock; }
// Note: Use frame::interpreter_frame_monitor_size() for the size of BasicObjectLocks
// in interpreter activation frames since it includes machine-specific padding.
static int size() { return sizeof(BasicObjectLock)/wordSize; }
// GC support
void oops_do(OopClosure* f) { f->do_oop(&_obj); }
static int obj_offset_in_bytes() { return offset_of(BasicObjectLock, _obj); }
static int lock_offset_in_bytes() { return offset_of(BasicObjectLock, _lock); }
};
class ObjectMonitor; class ObjectMonitor;
...@@ -163,6 +116,8 @@ class ObjectSynchronizer : AllStatic { ...@@ -163,6 +116,8 @@ class ObjectSynchronizer : AllStatic {
static void verify() PRODUCT_RETURN; static void verify() PRODUCT_RETURN;
static int verify_objmon_isinpool(ObjectMonitor *addr) PRODUCT_RETURN0; static int verify_objmon_isinpool(ObjectMonitor *addr) PRODUCT_RETURN0;
static void RegisterSpinCallback (int (*)(intptr_t, int), intptr_t) ;
private: private:
enum { _BLOCKSIZE = 128 }; enum { _BLOCKSIZE = 128 };
static ObjectMonitor* gBlockList; static ObjectMonitor* gBlockList;
...@@ -170,30 +125,6 @@ class ObjectSynchronizer : AllStatic { ...@@ -170,30 +125,6 @@ class ObjectSynchronizer : AllStatic {
static ObjectMonitor * volatile gOmInUseList; // for moribund thread, so monitors they inflated still get scanned static ObjectMonitor * volatile gOmInUseList; // for moribund thread, so monitors they inflated still get scanned
static int gOmInUseCount; static int gOmInUseCount;
public:
static void Initialize () ;
static PerfCounter * _sync_ContendedLockAttempts ;
static PerfCounter * _sync_FutileWakeups ;
static PerfCounter * _sync_Parks ;
static PerfCounter * _sync_EmptyNotifications ;
static PerfCounter * _sync_Notifications ;
static PerfCounter * _sync_SlowEnter ;
static PerfCounter * _sync_SlowExit ;
static PerfCounter * _sync_SlowNotify ;
static PerfCounter * _sync_SlowNotifyAll ;
static PerfCounter * _sync_FailedSpins ;
static PerfCounter * _sync_SuccessfulSpins ;
static PerfCounter * _sync_PrivateA ;
static PerfCounter * _sync_PrivateB ;
static PerfCounter * _sync_MonInCirculation ;
static PerfCounter * _sync_MonScavenged ;
static PerfCounter * _sync_Inflations ;
static PerfCounter * _sync_Deflations ;
static PerfLongVariable * _sync_MonExtant ;
public:
static void RegisterSpinCallback (int (*)(intptr_t, int), intptr_t) ;
}; };
// ObjectLocker enforced balanced locking and can never thrown an // ObjectLocker enforced balanced locking and can never thrown an
......
src/share/vm/runtime/thread.cpp
浏览文件 @ f9bc752b
...@@ -2995,8 +2995,8 @@ jint Threads::create_vm(JavaVMInitArgs* args, bool* canTryAgain) { ...@@ -2995,8 +2995,8 @@ jint Threads::create_vm(JavaVMInitArgs* args, bool* canTryAgain) {
// crash Linux VM, see notes in os_linux.cpp. // crash Linux VM, see notes in os_linux.cpp.
main_thread->create_stack_guard_pages(); main_thread->create_stack_guard_pages();
// Initialize Java-Leve synchronization subsystem // Initialize Java-Level synchronization subsystem
ObjectSynchronizer::Initialize() ; ObjectMonitor::Initialize() ;
// Initialize global modules // Initialize global modules
jint status = init_globals(); jint status = init_globals();
...@@ -3965,215 +3965,272 @@ void Threads::print_on_error(outputStream* st, Thread* current, char* buf, int b ...@@ -3965,215 +3965,272 @@ void Threads::print_on_error(outputStream* st, Thread* current, char* buf, int b
} }
} }
// Internal SpinLock and Mutex
// Based on ParkEvent
// Lifecycle management for TSM ParkEvents. // Ad-hoc mutual exclusion primitives: SpinLock and Mux
// ParkEvents are type-stable (TSM).
// In our particular implementation they happen to be immortal.
// //
// We manage concurrency on the FreeList with a CAS-based // We employ SpinLocks _only for low-contention, fixed-length
// detach-modify-reattach idiom that avoids the ABA problems // short-duration critical sections where we're concerned
// that would otherwise be present in a simple CAS-based // about native mutex_t or HotSpot Mutex:: latency.
// push-pop implementation. (push-one and pop-all) // The mux construct provides a spin-then-block mutual exclusion
// mechanism.
// //
// Caveat: Allocate() and Release() may be called from threads // Testing has shown that contention on the ListLock guarding gFreeList
// other than the thread associated with the Event! // is common. If we implement ListLock as a simple SpinLock it's common
// If we need to call Allocate() when running as the thread in // for the JVM to devolve to yielding with little progress. This is true
// question then look for the PD calls to initialize native TLS. // despite the fact that the critical sections protected by ListLock are
// Native TLS (Win32/Linux/Solaris) can only be initialized or // extremely short.
// accessed by the associated thread.
// See also pd_initialize().
// //
// Note that we could defer associating a ParkEvent with a thread // TODO-FIXME: ListLock should be of type SpinLock.
// until the 1st time the thread calls park(). unpark() calls to // We should make this a 1st-class type, integrated into the lock
// an unprovisioned thread would be ignored. The first park() call // hierarchy as leaf-locks. Critically, the SpinLock structure
// for a thread would allocate and associate a ParkEvent and return // should have sufficient padding to avoid false-sharing and excessive
// immediately. // cache-coherency traffic.
volatile int ParkEvent::ListLock = 0 ;
ParkEvent * volatile ParkEvent::FreeList = NULL ;
ParkEvent * ParkEvent::Allocate (Thread * t) { typedef volatile int SpinLockT ;
// In rare cases -- JVM_RawMonitor* operations -- we can find t == null.
ParkEvent * ev ;
// Start by trying to recycle an existing but unassociated void Thread::SpinAcquire (volatile int * adr, const char * LockName) {
// ParkEvent from the global free list. if (Atomic::cmpxchg (1, adr, 0) == 0) {
for (;;) { return ; // normal fast-path return
ev = FreeList ;
if (ev == NULL) break ;
// 1: Detach - sequester or privatize the list
// Tantamount to ev = Swap (&FreeList, NULL)
if (Atomic::cmpxchg_ptr (NULL, &FreeList, ev) != ev) {
continue ;
}
// We've detached the list. The list in-hand is now
// local to this thread. This thread can operate on the
// list without risk of interference from other threads.
// 2: Extract -- pop the 1st element from the list.
ParkEvent * List = ev->FreeNext ;
if (List == NULL) break ;
for (;;) {
// 3: Try to reattach the residual list
guarantee (List != NULL, "invariant") ;
ParkEvent * Arv = (ParkEvent *) Atomic::cmpxchg_ptr (List, &FreeList, NULL) ;
if (Arv == NULL) break ;
// New nodes arrived. Try to detach the recent arrivals.
if (Atomic::cmpxchg_ptr (NULL, &FreeList, Arv) != Arv) {
continue ;
}
guarantee (Arv != NULL, "invariant") ;
// 4: Merge Arv into List
ParkEvent * Tail = List ;
while (Tail->FreeNext != NULL) Tail = Tail->FreeNext ;
Tail->FreeNext = Arv ;
}
break ;
} }
if (ev != NULL) { // Slow-path : We've encountered contention -- Spin/Yield/Block strategy.
guarantee (ev->AssociatedWith == NULL, "invariant") ; TEVENT (SpinAcquire - ctx) ;
} else { int ctr = 0 ;
// Do this the hard way -- materialize a new ParkEvent. int Yields = 0 ;
// In rare cases an allocating thread might detach a long list --
// installing null into FreeList -- and then stall or be obstructed.
// A 2nd thread calling Allocate() would see FreeList == null.
// The list held privately by the 1st thread is unavailable to the 2nd thread.
// In that case the 2nd thread would have to materialize a new ParkEvent,
// even though free ParkEvents existed in the system. In this case we end up
// with more ParkEvents in circulation than we need, but the race is
// rare and the outcome is benign. Ideally, the # of extant ParkEvents
// is equal to the maximum # of threads that existed at any one time.
// Because of the race mentioned above, segments of the freelist
// can be transiently inaccessible. At worst we may end up with the
// # of ParkEvents in circulation slightly above the ideal.
// Note that if we didn't have the TSM/immortal constraint, then
// when reattaching, above, we could trim the list.
ev = new ParkEvent () ;
guarantee ((intptr_t(ev) & 0xFF) == 0, "invariant") ;
}
ev->reset() ; // courtesy to caller
ev->AssociatedWith = t ; // Associate ev with t
ev->FreeNext = NULL ;
return ev ;
}
void ParkEvent::Release (ParkEvent * ev) {
if (ev == NULL) return ;
guarantee (ev->FreeNext == NULL , "invariant") ;
ev->AssociatedWith = NULL ;
for (;;) { for (;;) {
// Push ev onto FreeList while (*adr != 0) {
// The mechanism is "half" lock-free. ++ctr ;
ParkEvent * List = FreeList ; if ((ctr & 0xFFF) == 0 || !os::is_MP()) {
ev->FreeNext = List ; if (Yields > 5) {
if (Atomic::cmpxchg_ptr (ev, &FreeList, List) == List) break ; // Consider using a simple NakedSleep() instead.
// Then SpinAcquire could be called by non-JVM threads
Thread::current()->_ParkEvent->park(1) ;
} else {
os::NakedYield() ;
++Yields ;
}
} else {
SpinPause() ;
}
}
if (Atomic::cmpxchg (1, adr, 0) == 0) return ;
} }
} }
// Override operator new and delete so we can ensure that the void Thread::SpinRelease (volatile int * adr) {
// least significant byte of ParkEvent addresses is 0. assert (*adr != 0, "invariant") ;
// Beware that excessive address alignment is undesirable OrderAccess::fence() ; // guarantee at least release consistency.
// as it can result in D$ index usage imbalance as // Roach-motel semantics.
// well as bank access imbalance on Niagara-like platforms, // It's safe if subsequent LDs and STs float "up" into the critical section,
// although Niagara's hash function should help. // but prior LDs and STs within the critical section can't be allowed
// to reorder or float past the ST that releases the lock.
void * ParkEvent::operator new (size_t sz) { *adr = 0 ;
return (void *) ((intptr_t (CHeapObj::operator new (sz + 256)) + 256) & -256) ;
} }
void ParkEvent::operator delete (void * a) { // muxAcquire and muxRelease:
// ParkEvents are type-stable and immortal ... //
ShouldNotReachHere(); // * muxAcquire and muxRelease support a single-word lock-word construct.
} // The LSB of the word is set IFF the lock is held.
// The remainder of the word points to the head of a singly-linked list
// of threads blocked on the lock.
//
// * The current implementation of muxAcquire-muxRelease uses its own
// dedicated Thread._MuxEvent instance. If we're interested in
// minimizing the peak number of extant ParkEvent instances then
// we could eliminate _MuxEvent and "borrow" _ParkEvent as long
// as certain invariants were satisfied. Specifically, care would need
// to be taken with regards to consuming unpark() "permits".
// A safe rule of thumb is that a thread would never call muxAcquire()
// if it's enqueued (cxq, EntryList, WaitList, etc) and will subsequently
// park(). Otherwise the _ParkEvent park() operation in muxAcquire() could
// consume an unpark() permit intended for monitorenter, for instance.
// One way around this would be to widen the restricted-range semaphore
// implemented in park(). Another alternative would be to provide
// multiple instances of the PlatformEvent() for each thread. One
// instance would be dedicated to muxAcquire-muxRelease, for instance.
//
// * Usage:
// -- Only as leaf locks
// -- for short-term locking only as muxAcquire does not perform
// thread state transitions.
//
// Alternatives:
// * We could implement muxAcquire and muxRelease with MCS or CLH locks
// but with parking or spin-then-park instead of pure spinning.
// * Use Taura-Oyama-Yonenzawa locks.
// * It's possible to construct a 1-0 lock if we encode the lockword as
// (List,LockByte). Acquire will CAS the full lockword while Release
// will STB 0 into the LockByte. The 1-0 scheme admits stranding, so
// acquiring threads use timers (ParkTimed) to detect and recover from
// the stranding window. Thread/Node structures must be aligned on 256-byte
// boundaries by using placement-new.
// * Augment MCS with advisory back-link fields maintained with CAS().
// Pictorially: LockWord -> T1 <-> T2 <-> T3 <-> ... <-> Tn <-> Owner.
// The validity of the backlinks must be ratified before we trust the value.
// If the backlinks are invalid the exiting thread must back-track through the
// the forward links, which are always trustworthy.
// * Add a successor indication. The LockWord is currently encoded as
// (List, LOCKBIT:1). We could also add a SUCCBIT or an explicit _succ variable
// to provide the usual futile-wakeup optimization.
// See RTStt for details.
// * Consider schedctl.sc_nopreempt to cover the critical section.
//
typedef volatile intptr_t MutexT ; // Mux Lock-word
enum MuxBits { LOCKBIT = 1 } ;
void Thread::muxAcquire (volatile intptr_t * Lock, const char * LockName) {
intptr_t w = Atomic::cmpxchg_ptr (LOCKBIT, Lock, 0) ;
if (w == 0) return ;
if ((w & LOCKBIT) == 0 && Atomic::cmpxchg_ptr (w|LOCKBIT, Lock, w) == w) {
return ;
}
// 6399321 As a temporary measure we copied & modified the ParkEvent:: TEVENT (muxAcquire - Contention) ;
// allocate() and release() code for use by Parkers. The Parker:: forms ParkEvent * const Self = Thread::current()->_MuxEvent ;
// will eventually be removed as we consolide and shift over to ParkEvents assert ((intptr_t(Self) & LOCKBIT) == 0, "invariant") ;
// for both builtin synchronization and JSR166 operations. for (;;) {
int its = (os::is_MP() ? 100 : 0) + 1 ;
// Optional spin phase: spin-then-park strategy
while (--its >= 0) {
w = *Lock ;
if ((w & LOCKBIT) == 0 && Atomic::cmpxchg_ptr (w|LOCKBIT, Lock, w) == w) {
return ;
}
}
Self->reset() ;
Self->OnList = intptr_t(Lock) ;
// The following fence() isn't _strictly necessary as the subsequent
// CAS() both serializes execution and ratifies the fetched *Lock value.
OrderAccess::fence();
for (;;) {
w = *Lock ;
if ((w & LOCKBIT) == 0) {
if (Atomic::cmpxchg_ptr (w|LOCKBIT, Lock, w) == w) {
Self->OnList = 0 ; // hygiene - allows stronger asserts
return ;
}
continue ; // Interference -- *Lock changed -- Just retry
}
assert (w & LOCKBIT, "invariant") ;
Self->ListNext = (ParkEvent *) (w & ~LOCKBIT );
if (Atomic::cmpxchg_ptr (intptr_t(Self)|LOCKBIT, Lock, w) == w) break ;
}
volatile int Parker::ListLock = 0 ; while (Self->OnList != 0) {
Parker * volatile Parker::FreeList = NULL ; Self->park() ;
}
}
}
Parker * Parker::Allocate (JavaThread * t) { void Thread::muxAcquireW (volatile intptr_t * Lock, ParkEvent * ev) {
guarantee (t != NULL, "invariant") ; intptr_t w = Atomic::cmpxchg_ptr (LOCKBIT, Lock, 0) ;
Parker * p ; if (w == 0) return ;
if ((w & LOCKBIT) == 0 && Atomic::cmpxchg_ptr (w|LOCKBIT, Lock, w) == w) {
return ;
}
// Start by trying to recycle an existing but unassociated TEVENT (muxAcquire - Contention) ;
// Parker from the global free list. ParkEvent * ReleaseAfter = NULL ;
if (ev == NULL) {
ev = ReleaseAfter = ParkEvent::Allocate (NULL) ;
}
assert ((intptr_t(ev) & LOCKBIT) == 0, "invariant") ;
for (;;) { for (;;) {
p = FreeList ; guarantee (ev->OnList == 0, "invariant") ;
if (p == NULL) break ; int its = (os::is_MP() ? 100 : 0) + 1 ;
// 1: Detach
// Tantamount to p = Swap (&FreeList, NULL) // Optional spin phase: spin-then-park strategy
if (Atomic::cmpxchg_ptr (NULL, &FreeList, p) != p) { while (--its >= 0) {
continue ; w = *Lock ;
if ((w & LOCKBIT) == 0 && Atomic::cmpxchg_ptr (w|LOCKBIT, Lock, w) == w) {
if (ReleaseAfter != NULL) {
ParkEvent::Release (ReleaseAfter) ;
}
return ;
}
} }
// We've detached the list. The list in-hand is now ev->reset() ;
// local to this thread. This thread can operate on the ev->OnList = intptr_t(Lock) ;
// list without risk of interference from other threads. // The following fence() isn't _strictly necessary as the subsequent
// 2: Extract -- pop the 1st element from the list. // CAS() both serializes execution and ratifies the fetched *Lock value.
Parker * List = p->FreeNext ; OrderAccess::fence();
if (List == NULL) break ;
for (;;) { for (;;) {
// 3: Try to reattach the residual list w = *Lock ;
guarantee (List != NULL, "invariant") ; if ((w & LOCKBIT) == 0) {
Parker * Arv = (Parker *) Atomic::cmpxchg_ptr (List, &FreeList, NULL) ; if (Atomic::cmpxchg_ptr (w|LOCKBIT, Lock, w) == w) {
if (Arv == NULL) break ; ev->OnList = 0 ;
// We call ::Release while holding the outer lock, thus
// New nodes arrived. Try to detach the recent arrivals. // artificially lengthening the critical section.
if (Atomic::cmpxchg_ptr (NULL, &FreeList, Arv) != Arv) { // Consider deferring the ::Release() until the subsequent unlock(),
continue ; // after we've dropped the outer lock.
if (ReleaseAfter != NULL) {
ParkEvent::Release (ReleaseAfter) ;
}
return ;
} }
guarantee (Arv != NULL, "invariant") ; continue ; // Interference -- *Lock changed -- Just retry
// 4: Merge Arv into List }
Parker * Tail = List ; assert (w & LOCKBIT, "invariant") ;
while (Tail->FreeNext != NULL) Tail = Tail->FreeNext ; ev->ListNext = (ParkEvent *) (w & ~LOCKBIT );
Tail->FreeNext = Arv ; if (Atomic::cmpxchg_ptr (intptr_t(ev)|LOCKBIT, Lock, w) == w) break ;
}
while (ev->OnList != 0) {
ev->park() ;
} }
break ;
} }
}
if (p != NULL) { // Release() must extract a successor from the list and then wake that thread.
guarantee (p->AssociatedWith == NULL, "invariant") ; // It can "pop" the front of the list or use a detach-modify-reattach (DMR) scheme
} else { // similar to that used by ParkEvent::Allocate() and ::Release(). DMR-based
// Do this the hard way -- materialize a new Parker.. // Release() would :
// In rare cases an allocating thread might detach // (A) CAS() or swap() null to *Lock, releasing the lock and detaching the list.
// a long list -- installing null into FreeList --and // (B) Extract a successor from the private list "in-hand"
// then stall. Another thread calling Allocate() would see // (C) attempt to CAS() the residual back into *Lock over null.
// FreeList == null and then invoke the ctor. In this case we // If there were any newly arrived threads and the CAS() would fail.
// end up with more Parkers in circulation than we need, but // In that case Release() would detach the RATs, re-merge the list in-hand
// the race is rare and the outcome is benign. // with the RATs and repeat as needed. Alternately, Release() might
// Ideally, the # of extant Parkers is equal to the // detach and extract a successor, but then pass the residual list to the wakee.
// maximum # of threads that existed at any one time. // The wakee would be responsible for reattaching and remerging before it
// Because of the race mentioned above, segments of the // competed for the lock.
// freelist can be transiently inaccessible. At worst //
// we may end up with the # of Parkers in circulation // Both "pop" and DMR are immune from ABA corruption -- there can be
// slightly above the ideal. // multiple concurrent pushers, but only one popper or detacher.
p = new Parker() ; // This implementation pops from the head of the list. This is unfair,
} // but tends to provide excellent throughput as hot threads remain hot.
p->AssociatedWith = t ; // Associate p with t // (We wake recently run threads first).
p->FreeNext = NULL ;
return p ; void Thread::muxRelease (volatile intptr_t * Lock) {
}
void Parker::Release (Parker * p) {
if (p == NULL) return ;
guarantee (p->AssociatedWith != NULL, "invariant") ;
guarantee (p->FreeNext == NULL , "invariant") ;
p->AssociatedWith = NULL ;
for (;;) { for (;;) {
// Push p onto FreeList const intptr_t w = Atomic::cmpxchg_ptr (0, Lock, LOCKBIT) ;
Parker * List = FreeList ; assert (w & LOCKBIT, "invariant") ;
p->FreeNext = List ; if (w == LOCKBIT) return ;
if (Atomic::cmpxchg_ptr (p, &FreeList, List) == List) break ; ParkEvent * List = (ParkEvent *) (w & ~LOCKBIT) ;
assert (List != NULL, "invariant") ;
assert (List->OnList == intptr_t(Lock), "invariant") ;
ParkEvent * nxt = List->ListNext ;
// The following CAS() releases the lock and pops the head element.
if (Atomic::cmpxchg_ptr (intptr_t(nxt), Lock, w) != w) {
continue ;
}
List->OnList = 0 ;
OrderAccess::fence() ;
List->unpark () ;
return ;
} }
} }
void Threads::verify() { void Threads::verify() {
ALL_JAVA_THREADS(p) { ALL_JAVA_THREADS(p) {
p->verify(); p->verify();
......
src/share/vm/runtime/thread.hpp
浏览文件 @ f9bc752b
...@@ -30,6 +30,7 @@ class JvmtiGetLoadedClassesClosure; ...@@ -30,6 +30,7 @@ class JvmtiGetLoadedClassesClosure;
class ThreadStatistics; class ThreadStatistics;
class ConcurrentLocksDump; class ConcurrentLocksDump;
class ParkEvent ; class ParkEvent ;
class Parker;
class ciEnv; class ciEnv;
class CompileThread; class CompileThread;
...@@ -544,7 +545,6 @@ public: ...@@ -544,7 +545,6 @@ public:
static void muxAcquire (volatile intptr_t * Lock, const char * Name) ; static void muxAcquire (volatile intptr_t * Lock, const char * Name) ;
static void muxAcquireW (volatile intptr_t * Lock, ParkEvent * ev) ; static void muxAcquireW (volatile intptr_t * Lock, ParkEvent * ev) ;
static void muxRelease (volatile intptr_t * Lock) ; static void muxRelease (volatile intptr_t * Lock) ;
}; };
// Inline implementation of Thread::current() // Inline implementation of Thread::current()
...@@ -1769,100 +1769,3 @@ public: ...@@ -1769,100 +1769,3 @@ public:
} }
}; };
// ParkEvents are type-stable and immortal.
//
// Lifecycle: Once a ParkEvent is associated with a thread that ParkEvent remains
// associated with the thread for the thread's entire lifetime - the relationship is
// stable. A thread will be associated at most one ParkEvent. When the thread
// expires, the ParkEvent moves to the EventFreeList. New threads attempt to allocate from
// the EventFreeList before creating a new Event. Type-stability frees us from
// worrying about stale Event or Thread references in the objectMonitor subsystem.
// (A reference to ParkEvent is always valid, even though the event may no longer be associated
// with the desired or expected thread. A key aspect of this design is that the callers of
// park, unpark, etc must tolerate stale references and spurious wakeups).
//
// Only the "associated" thread can block (park) on the ParkEvent, although
// any other thread can unpark a reachable parkevent. Park() is allowed to
// return spuriously. In fact park-unpark a really just an optimization to
// avoid unbounded spinning and surrender the CPU to be a polite system citizen.
// A degenerate albeit "impolite" park-unpark implementation could simply return.
// See http://blogs.sun.com/dave for more details.
//
// Eventually I'd like to eliminate Events and ObjectWaiters, both of which serve as
// thread proxies, and simply make the THREAD structure type-stable and persistent.
// Currently, we unpark events associated with threads, but ideally we'd just
// unpark threads.
//
// The base-class, PlatformEvent, is platform-specific while the ParkEvent is
// platform-independent. PlatformEvent provides park(), unpark(), etc., and
// is abstract -- that is, a PlatformEvent should never be instantiated except
// as part of a ParkEvent.
// Equivalently we could have defined a platform-independent base-class that
// exported Allocate(), Release(), etc. The platform-specific class would extend
// that base-class, adding park(), unpark(), etc.
//
// A word of caution: The JVM uses 2 very similar constructs:
// 1. ParkEvent are used for Java-level "monitor" synchronization.
// 2. Parkers are used by JSR166-JUC park-unpark.
//
// We'll want to eventually merge these redundant facilities and use ParkEvent.
class ParkEvent : public os::PlatformEvent {
private:
ParkEvent * FreeNext ;
// Current association
Thread * AssociatedWith ;
intptr_t RawThreadIdentity ; // LWPID etc
volatile int Incarnation ;
// diagnostic : keep track of last thread to wake this thread.
// this is useful for construction of dependency graphs.
void * LastWaker ;
public:
// MCS-CLH list linkage and Native Mutex/Monitor
ParkEvent * volatile ListNext ;
ParkEvent * volatile ListPrev ;
volatile intptr_t OnList ;
volatile int TState ;
volatile int Notified ; // for native monitor construct
volatile int IsWaiting ; // Enqueued on WaitSet
private:
static ParkEvent * volatile FreeList ;
static volatile int ListLock ;
// It's prudent to mark the dtor as "private"
// ensuring that it's not visible outside the package.
// Unfortunately gcc warns about such usage, so
// we revert to the less desirable "protected" visibility.
// The other compilers accept private dtors.
protected: // Ensure dtor is never invoked
~ParkEvent() { guarantee (0, "invariant") ; }
ParkEvent() : PlatformEvent() {
AssociatedWith = NULL ;
FreeNext = NULL ;
ListNext = NULL ;
ListPrev = NULL ;
OnList = 0 ;
TState = 0 ;
Notified = 0 ;
IsWaiting = 0 ;
}
// We use placement-new to force ParkEvent instances to be
// aligned on 256-byte address boundaries. This ensures that the least
// significant byte of a ParkEvent address is always 0.
void * operator new (size_t sz) ;
void operator delete (void * a) ;
public:
static ParkEvent * Allocate (Thread * t) ;
static void Release (ParkEvent * e) ;
} ;
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或