vm/thread/doc/ThreadManager.htm

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> <meta http-equiv="Content-Type" content= "text/html; charset=windows-1251"> <link href="drl.css" rel="stylesheet" type="text/css"> <title> Thread Manager Description </title> </head> <body> <a name="top"></a>Thread Manager Component Description <a href="#RevisionHistory">1. Revision History</a> <a href="#Disclaimer_and_Legal_Information">2. Disclaimer and Legal Information</a> <a href="#About_This_Document">3. About this Document</a> <a href="#Purpose">3.1 Purpose</a> <a href="#Intended_Audience">3.2 Intended Audience</a> <a href="#Documentation_Conventions">3.3 Documentation Conventions</a> <a href="#Using_this_document">3.4 Using this Document</a> <a href="#Overview">4. Overview</a> <a href="#Key_Feature">4.1 Key Features</a> <a href="#TM_in_VM">4.2 Thread Manager in VM</a> <a href="#Portability">4.3 Portability</a> <a href="#Achitecture">5. Architecture</a> <a href="#Exported_Interfaces">5.1 Exported interfaces</a> <blockquote> <a href="#Native_Layer">5.1.1 Native Interface</a> <a href="#Java_interface">5.1.2 Java* Interface</a> </blockquote> <a href="#Data_Structures">5.2 Data structures</a> <a href="#Thread_Control_Structure">5.3 Thread control structures</a> <blockquote> <a href="#Native_Structure">5.3.1 Native Thread Structure</a> <a href="#Java_Structure">5.3.2 Java* Thread Structure</a> </blockquote> <a href="#Thread_Group">5.4 Thread Groups</a> <a href="#Synchronizer">5.5 Synchronizers</a> <a href="#Monitors">5.6 Monitors</a> <blockquote> <a href="#MonitorInflation">5.6.1 Inflation Technique</a> <a href="#Thin_monitors">5.6.2 Monitor Structure</a> <a href="#AcquireMonitor">5.6.3 Acquiring a Monitor</a> </blockquote> <a href="#Usage_Scenarios">6. Usage scenarios</a> <a href="#Thread_lifecycle">6.1 Java* Thread Life Cycle</a> <a href="#Thread_suspension">6.2 Thread Suspension</a> <a href="#Stop-the-world_Suspension">6.3 Stop-the-world Thread Suspension</a> <a href="#Thread_locking">6.4 Thread Locking</a> <a href="#Monitor_scenarios">6.5 Monitor Enter and Exit</a> <a href="#References">7. References</a> <h1> <a name="RevisionHistory"></a>1. Revision History </h1> <table border="0" cellpadding="0" width="100%"> <tr> <td class="TableHeading"> Version </td> <td class="TableHeading"> Version Information </td> <td class="TableHeading"> Date </td> </tr> <tr> <td class="TableCell"> Initial version </td> <td class="TableCell"> Nadya Morozova, Andrey Chernyshev: document created. </td> <td class="TableCell"> June 5, 2006 </td> </tr> </table> <h1> <a name="Disclaimer_and_Legal_Information"></a>2. Disclaimer and Legal Information </h1> Copyright 2006 The Apache Software Foundation or its licensors, as applicable. Licensed under the Apache License, Version 2.0 (the License); you may not use this file except in compliance with the License. You may obtain a copy of the License at <a href="http://www.apache.org/licenses/LICENSE-2.0" target= "_blank">http://www.apache.org/licenses/LICENSE-2.0</a>. Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an AS IS BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. <a href="#top">Back to Top</a> <h1> <a name="About_This_Document"></a>3. About This Document </h1> <h2> <a name="Purpose"></a>3.1 Purpose </h2> This document introduces the thread manager component delivered as part of the DRL (Dynamic Runtime Layer) initiative. This document focuses on the specifics of the current implementation showing the thread manager role inside the DRL virtual machine, and the internal organization of the thread management subsystem. <h2> <a name="Intended_Audience"></a>3.2 Intended Audience </h2> The target audience for the document includes a wide community of engineers interested in further work with threading technologies to contribute to their development. The document assumes that readers are familiar with DRLVM architecture basics, threading methodologies and structures. <h2> <a name="Documentation_Conventions"></a>3.3 Documentation Conventions </h2> This document uses the <a href="conventions.htm" target= "_blank">unified conventions</a> for the DRL documentation kit. <h2> <a name="Using_this_document"></a>3.4 Using this Document </h2> Use this document to learn all about implementation specifics of the current version. It describes the thread manager functionality in a variety of aspects, including internal data structures, architecture specifics, and the key usage scenarios involving the thread manager. The document has the following major parts: <ul> <li> <a href="#Overview">Overview</a> gives the general definition of the thread manager component and its role in the VM architecture. <li> <a href="#Achitecture">Architecture</a> describes the internal structure of the thread manager, its data structures and the interfaces it exports. <li> <a href="#Usage_Scenarios">Usage scenarios</a> demonstrate major thread-related operations, such as the thread life cycle and thread suspension. <li> <a href="#References">References</a> are links to materials relevant to this description. </ul> <a href="#top">Back to Top</a> <h1> <a name="Overview"></a>4. Overview </h1> The thread manager (TM) is a library aimed to provide threading capabilities for Java<a href="#*">*</a> virtual machines. The main purpose of TM is to build a bridge between the POSIX-like threading model [<a href= "#posix_ref">5</a>] provided by the operating system, and the Java<a href="#*">*</a>-like threading model implied by the J2SE specification [<a href="#J2SESpec">1</a>]. In the current implementation, the JVM threading subsystem consists of three different layers: <ul> <li> The <a href="#Portability">porting layer</a> interacting with the operating system <li> The <a href="#Native_Layer">native layer</a> providing basic threading functionality <li> The <a href="#Java_interface">Java* layer</a> interacting with the Java<a href="#*">*</a> objects of the user application </ul> Note that the thread manager consists of the native and Java<a href="#*">*</a> layers of the subsystem, whereas as the porting layer is external. Each layer adds certain functionality to the threading provided by the underlying layer. That is, the porting layer adds portability to the threading provided by OS, the native layer adds Java<a href="#*">*</a>-specific enhancements to the porting layer, and the Java<a href="#*">*</a> layer adds a connection to Java<a href="#*">*</a> threads and objects to the native layer, as shown in Figure 1 below. These interfaces are grouped in a set of headers described in the <a href="#Exported_Interfaces">Exported Interfaces</a> section below. <img src="images/ThreadingSystem.gif" alt= "Basic layers in the thread manager"> Figure 1: Threading Subsystem <a href="#top">Back to Top</a> <h2> <a name="Key_Feature"></a>4.1 Key features </h2> The supplied thread manager has the following characteristics: <ul> <li> Support for the threading functionality required by J2SE API [<a href="#J2SESpec">1</a>], JVMTI [<a href= "#JVMTI_ref">2</a>] and JNI [<a href="#JNI_ref">3</a>] specifications <li> Portable implementation mostly based on the Apache Porting Layer [<a href="#APR_ref">4</a>] <li> Compliance with the Harmony�<code>hythread</code> interface [<a href="#hythread_ref">8</a>] <li> Support for the garbage collector <li> Monitor optimizations specific to the just-in-time (JIT) compiler supplied with DRLVM </ul> <h2> <a name="TM_in_VM"></a>4.2 Thread Manager in VM </h2> Figure 2 below demonstrates the interaction of the thread manager with the following components of the virtual machine: <ul> <li> The VM core to access information on object layout and for binding between�<code>java.lang.Thread</code> objects and appropriate native threads. For that, the thread manager queries the <code>thread_externals.h</code> interface of the VM core. <li> The garbage collector to serve thread manipulation requests for root set enumeration and garbage collection activities. GC works with the native layer of the thread manager. <li> The porting layer to interact with the underlying system and enable <a href="#Portability">portability</a> for threading. The TM native layer queries functions of the APR interfaces and the portlib interfaces. <li> The just-in-time compiler to provide optimized threading functions, called VM helpers, for JIT-compiled code. The thread manager exports this functionality via the <code>thread_helpers</code> interface of the Java<a href="#*">*</a> layer. </ul> <img src="images/tm_in_vm.gif" alt= "Thread Manager and other VM components"> Figure 2: Thread Manager in VM Architecture <h2> <a name="Portability"></a>4.3 Portability </h2> The thread manager code is mostly platform-independent and relies on the underlying porting layer to adjust to platform specifics. The current TM implementation is written on top of the Apache Porting Layer (APR) with certain extensions added to it. The platform-dependent TM parts are the VM helpers package, which is tied to the specific architecture, and the APR extensions package, which is partially tied with the OS API. APR-based porting enables compilation of the thread manager code on every platform where APR is available. The current version of the thread manager supports the Linux<a href= "#*">*</a> and Windows<a href="#*">*</a> IA-32 platforms. <a href="#top">Back to Top</a> <h1> <a name="Achitecture"></a>5. Architecture </h1> Subsequent sections describe the functional interfaces that the thread manager exports to interact with other VM components and its internal data structures. <h2> <a name="Exported_Interfaces"></a>5.1 Exported interfaces </h2> As indicated in the overview, the thread manager exports the native and the Java<a href="#*">*</a> interfaces. These interfaces are represented as groups of functions providing specific functionality upon external requests, as described in the subsequent sections. <h3> <a name="Native_Layer"></a>5.1.1 Native Interface </h3> The native interface is inspired by the Harmony <code>hythread</code> module. This is a low-level layer that provides Java<a href="#*">*</a>-like threading functionality, such as interruption support for waiting operations (for example,�<code>wait</code>,� <code>park</code>,�<code>join</code> and�<code>sleep</code>) and helps establish correct interaction of threads with the garbage collector. This layer does not deal with Java<a href="#*">*</a> objects. The native interface consists of the following function sets: �<code>hythread.h</code> Functions of the�<code>hythread</code> set [<a href= "#hythread_ref">8</a>] responsible for: <ul> <li> Basic manipulation <li> Parking <li> Thread local storage support <li> Read-write mutex support <li> Monitors support </ul> �<code>hythread_ext.h</code> Set of functions extending the�<code>hythread</code> set responsible for: <ul> <li> Thread manager initialization and shutdown <li> Thread groups support <li> Conditional variable <li> Safe suspension support <li> Latch <li> Thread iterator support <li> Attributes access <li> Querying state of the thread <li> Semaphore <li> Mutex <li> Thin monitors support <li> Querying the thread state <li> Thread attributes access <li> Interruption support <li> Task management support </ul> <a href="#top">Back to Top</a> <h3> <a name="Java_interface"></a>5.1.2 Java<a href="#*">*</a> Interface </h3> The Java<a href="#*">*</a> interface connects the threading functionality provided by the native layer to Java<a href= "#*">*</a> threads and objects. The functions of the Java<a href="#*">*</a> interface take Java<a href="#*">*</a> objects as parameters and can be easily used to implement kernel classes, JNI or JVMTI function sets. The Java<a href="#*">*</a> interface consists of 3 parts: �<code>jthread.h</code> Functions supporting <code>java.lang.Object</code> and <code>java.lang.Thread</code> API responsible for: <ul> <li> Basic manipulation <li> Identification <li> Pointer conversion <li> Attributes access <li> Interruption <li> Monitors <li> Parking <li> Suspension </ul> �<code>ti_thread.h</code> Functions supporting various JVMTI functions and the <code>java.lang.management</code> API responsible for: <ul> <li> State query <li> Instrumentation <li> Local storage <li> Monitor info <li> CPU timing <li> Peak count <li> Raw monitors </ul> �<code>thread_helpers.h</code> Functions providing the stubs that help to optimize the performance due to tighter TM and JIT integration. <a href="#Top">Back to Top</a> <h2> <a name="Data_Structures"></a>5.2 Data structures </h2> The thread manager data structures are typically not exposed: external VM components access these structures via opaque pointers instead. The pointers are defined in the public header files <code>hythread.h</code>, <code>hythread_ext.h</code>, <code>jthread.h</code> and <code>ti_thread.h</code>. Structures themselves are described in the <code>thread_private.h</code> file. <h2> <a name="Thread_Control_Structure"></a>5.3 Thread control structures </h2> The thread manager requires each thread to be registered before threading functions can be called. Thread registration is called attaching a thread and can be done by using one of the following: <ul> <li> Function <code>hythread_attach()</code> registers the current thread in the thread manager, so that threading operations can be performed over this thread via the native layer. <li> Function <code>jthread_attach()</code> associates the current thread with the appropriate <code>java.lang.Thread</code> object, so that threading operations can be performed over this thread via the Java<a href="#*">*</a> layer. </ul> Depending on the attaching function, the thread manager operates with two types of threads: <ul> <li> Native thread attached to the native layer of the thread manager <li> Java<a href="#*">*</a> thread attached to the Java<a href="#*">*</a> layer of the thread manager and associated with a�<code>java.lang.Thread</code> object </ul> Each thread type has a structure assigned to it that holds thread-specific data, as described below. Other VM components work with opaque handles to those structures and have no information about their contents. This way, to work with a thread, a component calls one of the attaching functions, receives an opaque handle to the thread control structure for the thread, and performs whatever operations with this thread using this opaque handle. <h3> <a name="Native_Structure"></a>5.3.1 Native Thread Structure </h3> When registered with the thread manager’s native layer, each thread obtains a control structure with all thread-specific data required for operations with the thread, such as state, attributes, references to OS-specific thread structures, and synchronization aids. The control structure is subsequently used for miscellaneous threading operations. Note The actual content of a thread control structure is implementation-specific and is not exposed to other components. � The following figure shows the thread control structure of a native attached thread described by the type <code>HyThread</code>: <img border="0" src="images/NativeUnattachedThread.gif" alt= "Structure of a native thread not registered in TM"> Figure 3:� Native Thread The following table lists the data fields making up the thread control structure of a native thread: <table width="100%"> <tr> <td colspan="2" class="TableHeading"> Data Field </td> <td class="TableHeading"> Description </td> </tr> <tr> <td colspan="3" class="TableCell"> Suspension </td> </tr> <tr> <td class="TableCell"> <code>IDATA</code> </td> <td class="TableCell"> <code>suspend_request</code> </td> <td class="TableCell"> Number of suspend requests made for this thread. </td> </tr> <tr> <td class="TableCell"> <code>int16�</code> </td> <td class="TableCell"> <code>suspend_disable_count</code> </td> <td class="TableCell"> Flag indicating that the thread can safely be suspended. </td> </tr> <tr> <td class="TableCell"> <code>hylatch_t�</code> </td> <td class="TableCell"> <code>safe_region_event</code> </td> <td class="TableCell"> Event used to notify interested threads when a thread enters a safe region. </td> </tr> <tr> <td class="TableCell"> <code>hysem_t�</code> </td> <td class="TableCell"> <code>resume_event</code> </td> <td class="TableCell"> Event used to notify a suspended thread that it needs to wake up. </td> </tr> <tr> <td class="TableCell"> <code>hythread_event_callback_proc�</code> </td> <td class="TableCell"> <code>safepoint_callback</code> </td> <td class="TableCell"> Function to be executed at a safe point on resuming a thread. </td> </tr> <tr> <td colspan="3" class="TableCell"> Basic manipulation fields </td> </tr> <tr> <td class="TableCell"> <code>hythread_group_t�</code> </td> <td class="TableCell"> <code>group</code> </td> <td class="TableCell"> Group for this thread equal to the address of the head of the list of threads for this group. Groups are used to quickly iterate over lists of threads. Examples: Java<a href="#*">*</a> threads, GC private threads. </td> </tr> <tr> <td class="TableCell"> <code>hythread_t�</code> </td> <td class="TableCell"> <code>next</code> </td> <td class="TableCell"> Pointer to the next thread within the group. </td> </tr> <tr> <td class="TableCell"> <code>hythread_t�</code> </td> <td class="TableCell"> <code>prev</code> </td> <td class="TableCell"> Pointer to the last thread within the group. </td> </tr> <tr> <td class="TableCell"> <code>apr_thread_t *�</code> </td> <td class="TableCell"> <code>os_handle</code> </td> <td class="TableCell"> Handle to the OS thread. </td> </tr> <tr> <td class="TableCell"> <code>void *�</code> </td> <td class="TableCell"> <code>private_data</code> </td> <td class="TableCell"> Placeholder for any data to be associated with this thread. The Java<a href="#*">*</a> layer uses this field to store Java-specific context. </td> </tr> <tr> <td class="TableCell"> <code>Boolean�</code> </td> <td class="TableCell"> <code>exit_request</code> </td> <td class="TableCell"> Flag indicating that a request to exit has been received. </td> </tr> <tr> <td class="TableCell"> <code>IDATA�</code> </td> <td class="TableCell"> <code>exit_value</code> </td> <td class="TableCell"> Exit value of this thread. </td> </tr> <tr> <td colspan="3" class="TableCell"> Synchronization support </td> </tr> <tr> <td class="TableCell"> <code>hysem_t�</code> </td> <td class="TableCell"> <code>park_event</code> </td> <td class="TableCell"> Event for parking threads. </td> </tr> <tr> <td class="TableCell"> <code>hysem_t�</code> </td> <td class="TableCell"> <code>sleep_event</code> </td> <td class="TableCell"> Event for sleeping threads. </td> </tr> <tr> <td class="TableCell"> <code>hylatch_t�</code> </td> <td class="TableCell"> <code>join_event</code> </td> <td class="TableCell"> Event reserved for threads that invoke join. </td> </tr> <tr> <td class="TableCell"> <code>hycond_t�</code> </td> <td class="TableCell"> <code>current_condition</code> </td> <td class="TableCell"> Current conditional variable that the thread is waiting on. Used for interrupting. </td> </tr> <tr> <td colspan="3" class="TableCell"> State </td> </tr> <tr> <td class="TableCell"> <code>IDATA�</code> </td> <td class="TableCell"> <code>state</code> </td> <td class="TableCell"> Thread state. Holds thread state flags as defined in the JVMTI specification [<a href="#JVMTI_ref">2</a>] plus additional flags. </td> </tr> <tr> <td colspan="3" class="TableCell"> Attributes </td> </tr> <tr> <td class="TableCell"> <code>char *�</code> </td> <td class="TableCell"> <code>name</code> </td> <td class="TableCell"> Name of the thread, can be used for debugging. </td> </tr> <tr> <td class="TableCell"> <code>IDATA�</code> </td> <td class="TableCell"> <code>priority</code> </td> <td class="TableCell"> Hint to the scheduler about thread priority. </td> </tr> <tr> <td class="TableCell"> <code>IDATA�</code> </td> <td class="TableCell"> <code>daemon</code> </td> <td class="TableCell"> Checks whether this thread is a daemon. </td> </tr> <tr> <td colspan="3" class="TableCell"> Other </td> </tr> <tr> <td class="TableCell"> <code>IDATA�</code> </td> <td class="TableCell"> <code>thread_id</code> </td> <td class="TableCell"> ID for this thread. The maximum number of threads is governed by the size of the lock word record. See section <a href="#Thin_monitors">Thin Monitors</a> for details on the lock word structure. </td> </tr> <tr> <td class="TableCell"> <code>apr_pool_t *�</code> </td> <td class="TableCell"> <code>pool</code> </td> <td class="TableCell"> Memory pool, where this thread is allocated. </td> </tr> <tr> <td class="TableCell"> <code>apr_threadattr_t *�</code> </td> <td class="TableCell"> <code>apr_attrs</code> </td> <td class="TableCell"> APR thread attributes. </td> </tr> <tr> <td class="TableCell"> <code>hythread_entrypoint_t�</code> </td> <td class="TableCell"> <code>start_proc</code> </td> <td class="TableCell"> Procedure that describes thread body to be executed. </td> </tr> <tr> <td class="TableCell"> <code>void *�</code> </td> <td class="TableCell"> <code>start_proc_args</code> </td> <td class="TableCell"> Arguments to be passed to the thread body. </td> </tr> <tr> <td class="TableCell"> <code>void *�</code> </td> <td class="TableCell"> <code>thread_local_storage</code> </td> <td class="TableCell"> Array representing the thread local storage. </td> </tr> <tr> <td class="TableCell"> <code>void *�</code> </td> <td class="TableCell"> <code>big_local_storage</code> </td> <td class="TableCell"> Extension to the standard local storage slot. </td> </tr> </table> For details on thread control structures, see the <code>thread_private.h</code> header file supplied with the source bundle. <a href="#Top">Back to Top</a> <h3> <a name="Java_Structure"></a>5.3.2 Java<a href="#*">*</a> Thread Structure </h3> A thread control structure of a Java<a href="#*">*</a> thread is defined by the <code>JVMTIThread</code> type and holds mostly JVMTI information specific to that thread, as shown in Figure 4 below. <img src="images/JavaAttached.gif" alt= "Structure of the Java* Attached thread"> Figure 4: Java<a href="#*">*</a> Attached Thread The following table lists the data fields making up the thread control structure of a Java<a href="#*">*</a> thread: <table width="100%"> <tr> <td colspan="2" class="TableHeading"> Data Field </td> <td class="TableHeading"> Description </td> </tr> <tr> <td class="TableCell"> <code>JNIEnv *�</code> </td> <td class="TableCell"> <code>jenv</code> </td> <td class="TableCell"> JNI environment variable associated with this Java<a href="#*">*</a> thread. </td> </tr> <tr> <td class="TableCell"> <code>jthread�</code> </td> <td class="TableCell"> <code>thread_object</code> </td> <td class="TableCell"> The object <code>jthread</code> associated with <code>hythread_t</code>. </td> </tr> <tr> <td class="TableCell"> <code>hycond_t�</code> </td> <td class="TableCell"> <code>monitor_condition</code> </td> <td class="TableCell"> Conditional variable used to wait/notify Java<a href= "#*">*</a> monitors. </td> </tr> <tr> <td class="TableCell"> <code>jthrowable�</code> </td> <td class="TableCell"> <code>stop_exception</code> </td> <td class="TableCell"> Exception to be thrown in a stopped thread. </td> </tr> <tr> <td class="TableCell"> <code>jlong�</code> </td> <td class="TableCell"> <code>blocked_time</code> </td> <td class="TableCell"> Time duration of the thread blocked on a monitor, in nanoseconds. </td> </tr> <tr> <td class="TableCell"> <code>jlong�</code> </td> <td class="TableCell"> <code>waited_time</code> </td> <td class="TableCell"> Time duration of the thread waiting on a monitor, in nanoseconds. </td> </tr> <tr> <td class="TableCell"> <code>JVMTILocalStorage�</code> </td> <td class="TableCell"> <code>jvmti_local_storage</code> </td> <td class="TableCell"> JVMTI local storage. </td> </tr> <tr> <td class="TableCell"> <code>jobject�</code> </td> <td class="TableCell"> <code>contended_monitor</code> </td> <td class="TableCell"> Monitor that this thread is blocked on. </td> </tr> <tr> <td class="TableCell"> <code>jobject�</code> </td> <td class="TableCell"> <code>wait_monitor</code> </td> <td class="TableCell"> Monitor that this thread is waiting on. </td> </tr> <tr> <td class="TableCell"> <code>jobject *�</code> </td> <td class="TableCell"> <code>owned_monitors</code> </td> <td class="TableCell"> Monitors owned by this thread. </td> </tr> <tr> <td class="TableCell"> <code>int�</code> </td> <td class="TableCell"> <code>owned_monitors_nmb</code> </td> <td class="TableCell"> The counter for the monitors owned by the thread. </td> </tr> <tr> <td class="TableCell"> <code>apr_pool_t *�</code> </td> <td class="TableCell"> <code>pool</code> </td> <td class="TableCell"> APR pool for this structure. </td> </tr> <tr> <td class="TableCell"> <code>jobject�</code> </td> <td class="TableCell"> <code>thread_ref</code> </td> <td class="TableCell"> Weak reference to the corresponding <code>java.lang.Thread</code> instance. </td> </tr> </table> For details on thread control structures, see the <code>thread_private.h</code> header file supplied with the source bundle. The data structures associated with a thread attached to a <code>java.lang.Thread</code> object and for a standalone (unattached) thread are shown in Figures 3 and 4 respectively. <a href="#Top">Back to Top</a> <h2> <a name="Thread_Group"></a>5.4 Thread Groups </h2> The thread manager enables co-existence of multiple groups of threads, for example, groups of Java<a href="#*">*</a> threads and GC threads not visible for Java<a href= "#*">*</a> applications. Each thread maintained by the thread manager belongs to a specific thread group, as shown in Figure 5. <img src="images/thread_groups.gif" alt= "Threads distributed into thread groups, for example a group of Java* threads and a group of GC threads"> Figure 5: Thread Groups The thread manager provides a set of functions for iterating over the list of threads within a specific group. All threads are organized in a group array and a specific system-wide lock is used to prevent concurrent modifications of the groups array and the thread list inside the group. This lock is acquired internally during thread creation, deletion and iteration over the thread list. <a href="#Top">Back to Top</a> <h2> <a name="Synchronizer"></a>5.5 Synchronizers </h2> The thread manager synchronizers are functional modules used for thread synchronization. Certain synchronizers have internal data structures associated with them, others can only delegate function calls to the appropriate synchronizers provided by APR. The current implementation of synchronizers within the thread manager is based on two fundamental primitives: the conditional variable and the mutex, as shown in Figure 6. <img src="images/Synchronizer_mutex.gif" alt= "Implementing the thread synchronizer in TM" border="0"> Figure 6: Components of the TM Synchronizer� The elements in the figure have the following meaning: <ul> <li> The APR conditional variable and APR mutex are basic primitives provided by the Apache Portable Runtime. <li> The TM conditional variable and TM mutex wrap appropriate APR primitives by adding the�<code>wait</code> interruption support. These synchronizers also ensure that a thread enters the safe suspension mode when it is put into a�<code>wait</code> state using the conditional variable or when the thread is blocked while acquiring a mutex. <li> The <a href="#Thin_monitors">thin monitor</a> is an inflatable lock coupled with the condition variable. This combination serves as a base for building Java<a href= "#*">*</a> monitors. <li> The semaphore is the same as the POSIX semaphore, and also enables specifying the count limit. <li> The Java<a href="#*">*</a> monitor is the same as <code>java.lang.Object</code>. <li> The JVMTI raw monitor is the monitor defined in the JVMTI specification. <li> The <code>park</code> and <code>unpark</code> lock support primitives are used in the <code>java.util.concurrent</code> package. </ul> The above hierarchy is optimized for APR code re-use. Other implementations of the Thread Manager component are also possible and can utilize a different set of APR synchronizers. Note The thread manager does not expose the internal structures of synchronizers to the external components. All synchronizers are referenced by means of opaque handles similarly to thread control structures. <a href="#Top">Back to Top</a> <h2> <a name="Monitors"></a>5.6 Monitors </h2> The current version of the thread manager implements Java<a href="#*">*</a> monitors in a specific way to address the common problem of space comsumption. The DRL thread manager provides has a special type of monitor, thin monitor, holding the lock optimized for space consumption and single-threaded usage. <h3> <a name="MonitorInflation"></a>5.6.1 Inflation technique </h3> Monitor inflation is implemented using a thin-fat lock technique [<a href="#thin_lock_ref">6</a>], which works as follows: <ul> <li> In the absence of thread contention, lock data are stored in a few bytes, so that the lock can be allocated directly within the Java<a href="#*">*</a> object. <li> Whenever contention takes place, the bytes allocated for lock data hold a reference to the fat lock, which can be conventional mutex. </ul> Different implementations of thin monitors are free to choose any space compaction or other optimization techniques (or none at all). However, the general recommendation is to use thin monitors when memory needs to be saved and a thread contention is not expected to be high. It is also recommended that the conventional mutex and conditional variables be used to achieve the better scalability in case of high contention. Java<a href="#*">*</a> monitors in the thread manager are built on top of thin monitors. This enables the thread manager to allocate the lock structure for thin monitors directly in the Java<a href="#*">*</a> objects and thus makes Java<a href="#*">*</a> monitors space usage more efficient. <h3> <a name="Thin_monitors"></a>5.6.2 Monitor Structure </h3> The thin monitor is a synchronizer primitive that implements the lock compression technique and serves as a base for building Java<a href="#*">*</a> monitors [<a href= "#thin_lock_ref">6</a>]. In other words, the thin monitor resides in the native layer of the TM subsystem and has no data on Java<a href="#*">*</a> objects. Java<a href= "#*">*</a> monitors are tightly coupled with Java<a href= "#*">*</a> objects and reside on the higher Java<a href= "#*">*</a> level of the TM subsystem. The central point of the synchronizer is the lock word, which holds the thin lock value or a reference to the fat lock depending on the contention. In the absence of contention, the lock type is zero, and the lock word has the following structure: <img src="images/uninflated_lockword.gif" alt= "uninflated loack" border="0"> Figure 7: Lock Word Structure: Contention Bit is 0 <ul> <li> Contention bit : 0 indicating that absence of contention <li> Thread ID (15 bits): the ID of the owning thread, or 0 if the lock is free <li> Recursion count: the number of times that the lock has been acquired by the same thread minus 1 <li> Reservation bit: the flag indicating whether the lock is reserved by a thread [<a href="#lock_overhead_ref">7</a>] <li> Rightmost 10 bits unused in TM and reserved for storing the hash codes of Java<a href="#*">*</a> objects </ul> In the presence of contention, the contention bit is set to 1, and a thin compressed lock becomes a fat inflated lock with the following layout: <img alt="inflated lock" border="0" src= "images/inflated_lockword.gif"> Figure 8: Lock Word Structure: Contention Bit is 1 <ul> <li> Contention bit: 1 indicating presence of contention <li> Fat Lock ID (20 bits): the ID of the corresponding fat lock <li> Reservation bit: the flag indicating whether the lock is reserved by a thread [<a href="#lock_overhead_ref">7</a>] <li> Rightmost 10 bits unused in TM and reserved for storing the hash codes of Java<a href="#*">*</a> objects </ul> The thread manager has a global lock table to map between the lock ID and the appropriate fat monitor, as follows: <img src="images/inflated_thin_monitor.gif" alt= "Inflated thin monitor" border="0"> Figure 9: Thin and Fat Monitor Relationship <a href="#Top">Back to Top</a> <h3> <a name="AcquireMonitor"></a>5.6.3 Acquiring a Monitor </h3> The process of acquiring a monitor with the help of the <code>hythread_thin_monitor_try_enter()</code> function is shown on the following diagram: <img src="images/Lock_reservation.gif" alt= "Lock reservation"> Figure 10: Acquiring a Thin Lock First, the thread uses the reservation bit to check whether the required lock is owned by this thread. If yes, the thread increases the recursion count by 1 and the function exits. This makes the fast path of the monitor enter operation for a single-threaded application. The fast path involves only a few assembly instructions and does no expensive atomic compare-and-swap (CAS) operations. If the lock is not yet reserved, then it is checked for being occupied. The free lock is set to be reserved and acquired simultaneously with a single CAS operation. If the lock becomes busy then, the system checks whether the lock is fat. The lock table holds a mapping between the fat lock ID and the actual monitor. Fat monitors are extracted from the lock table and acquired. If the lock is not fat and reserved by another thread, then this thread suspends the execution of the lock owner thread, removes the reservation, and resumes the owner thread. After that, the lock acquisition is tried again. <a href="#Top">Back to Top</a> <h1> <a name="Usage_Scenarios"></a>6. Usage scenarios </h1> This section contains various scenarios of thread manipulation. <h2> <a name="Thread_lifecycle"></a>6.1 Java<a href="#*">*</a> thread life cycle </h2> The Java<a href="#*">*</a> thread creation procedure consists of the following key stages: <ol> <li> After creating a new thread, the <code>Thread()</code> constructor creates a new native thread and initializes <a href="#Thread_Control_Structure">thread control structures</a> <code>HyThread</code> and <code>JVMTIThread</code>. <li> The user application then calls the <code>java.lang.Thread.start()</code> method of kernel classes within the VM core component. <li> The <code>java.lang.Thread.start()</code> method delegates the call to the <code>jthread_create()</code> function of the Java<a href="#*">*</a> layer in the thread manager via the <code>java.lang.VMThreadManager.start()</code> function. <li> The function <code>jthread_create()</code> calls <code>jthread_create_with_function()</code>. <li> The <code>jthread_create_with_function()</code> function calls <code>hythread_create()</code> supplying <code>wrapper_proc</code> as a new thread body procedure. <li> The <code>hythread_create()</code> function delegates the call to the <code>apr_thread_create()</code> function of the APR porting layer, which does the actual fork and creates a new thread. The <code>wrapper_proc</code> argument is supplied as the thread body at this step. <li> The newly created thread begins to execute <code>thread_start_proc()</code>, which, in its turn, runs the <code>wrapper_proc()</code> function. <li> The function <code>wrapper_proc()</code> performs the required registration of the new thread with other components by sending the <code>vm_attach</code> event to VM core and by calling the <code>jvmti_send_thread_start_end_event()</code> function, which sends the <code>JVMTI_EVENT_THREAD_START</code> event. <li> The function <code>wrapper_proc()</code> calls the <code>java.lang.Thead.run()</code> method, which makes the user-defined body of the new Java<a href="#*">*</a> thread. <li> After <code>Thread.run()</code> has finished, the thread is unregistered with other components by calling the <code>jvmti_send_thread_start_end_event()</code> function, which sends the <code>JVMTI_EVENT_THREAD_END</code> and then the <code>vm_detach</code> events. </ol> The following figure illustrates the detailed sequence of thread creation and completion: <img alt="Thread operation from creation to finalization" border="0" src="images/ThreadCreation.gif"> Figure 11: Java<a href="#*">*</a> Thread Life Cycle <a name="TM.END"></a>Note The native thread control structures (such as, <code>HyThread</code> and <code>JVMTIThread</code>) are not de-allocated once the new thread body is finished. The thread manager creates a weak reference for each <code>java.lang.Thread</code> object supplying its internal reference queue. The garbage collector places a reference into that queue when a specific <code>java.lang.Thread</code> object is garbage-collected. Before allocating native resources for new threads, the thread manager seeks for the weak references in the queue. In case the weak references queue is not empty, the thread manager extracts the first available reference and re-uses its native resources for the newly created thread. <a href="#Top">Back to Top</a> <h2> <a name="Thread_suspension"></a>6.2 Thread Suspension </h2> One of the important features that the native layer adds for APR threading is safe suspension. This mechanism ensures that the suspended thread can be safely explored by the garbage collector during the enumeration of live references. If a thread holds some system-critical locks, such as the locks associated with the native heap memory, safe suspension can keep it running even during the enumeration. Otherwise, doing the system or “hard” call to suspend the thread may result in deadlocks in case system locks are requested by other parts of the VM. The algorithm of safe suspension describes the protocol of communication between two threads, for example, thread T1 and thread T2, where is T1 safely suspends thread T2. The T1 thread calls the <code>hythread_suspend(T2)</code> function to suspend thread T2. The procedure goes in the following stages: <ol> <li> The <code>hythread_suspend(T2)</code> function increments the flag for the T2 thread indicating a request for suspension. Depending on the current state of thread T2, the <code>hythread_suspend(T2)</code> function activates one of the following mechanisms: <ol> <li> If thread T2 is currently running in a safe code region, the <code>hythread_suspend(T2)</code> call immediately returns, see Figure 12. <li> If thread T2 is currently in an unsafe region, then the <code>hythread_suspend()</code> function gets blocked until thread T2 reaches the beginning of a safe region or a safe point. </ol> <li> Thread T2 runs to the end of the safe region and gets blocked until T1 resumes it by calling <code>hythread_resume(T2)</code>. </ol> The T2 thread undergoes the following: <ol start="1" type="1"> <li> Thread T2 periodically calls the <code>hythread_safe_point()</code> function to designate the points of safe suspension. If a suspension request has been set previously for T2, this method notifies the T1 thread and then waits until T1 resumes it by calling <code>hythread_resume(T2)</code>.� <li> When the T2 thread enters a safe region, it calls the <code>hythread_suspend_ensable()</code> function, which decrements the <code>suspend_disable_count</code> state flag. If a suspension request has been previously set for T2, T1 is notified about T2 having reached a safe region. <li> When T2 leaves a safe region, it calls the <code>hythread_suspend_disable()</code> function. This function increments the <code>suspend_disable_count</code> state flag. </ol> A typical example of the safe suspension scenario takes place when the garbage collector suspends a Java<a href= "#*">*</a> thread to enumerate live references. Figure 12 illustrates the case when the GC uses the thread manager to suspend the Java<a href="#*">*</a> thread while it is running in the safe code region. <img alt="Safe region during thread execution" border="0" src="images/safeRegion.gif"> Figure 12: Suspension: Safe Region To understand the safe thread suspension algorithm better, think of each thread as having a lock associated with it. Thread T2 releases the lock when it enters a safe region and acquires the lock when it leaves the safe region. To suspend thread T2, acquire the lock associated with it. Resuming thread T2 is equivalent to releasing the lock associated with it. A straight-forward implementation of the safe suspension algorithm reserves a single-thread optimized lock (that is, the thin monitor) for each thread and uses it for suspending and resuming that thread. Another safe suspension case is when a GC thread hits a Java<a href="#*">*</a> thread while it is in an unsafe region of code, as shown in Figure 13. <img alt="Safe Point in Thread Execution" border="0" src= "images/safePoint.gif"> Figure 13: Safe Point Consider the <code>hythread_safe_point()</code> operation as a wait operation performed over the monitor associated with the thread. In this case, the <code>hythread_resume()</code> operation is equivalent to notifying that monitor. <a href="#Top">Back to Top</a> <h2> <a name="Stop-the-world_Suspension"></a>6.3 Stop-the-world Thread Suspension </h2> The stop-the-world thread suspension happens when the garbage collector needs to enumerate the live object references for all threads of a given thread group. Figure 14 illustrates the case when only a GC thread an indefinite number of Java<a href="#*">*</a> threads are running, so that the GC needs to suspend all Java<a href="#*">*</a> threads. <img alt="stop-the-world suspension" border="0" src= "images/SuspendAll.gif"> Figure 14: Suspending a Group of Threads First, the garbage collector calls the thread manager interface function <code>hythread_suspend_all()</code> to suspend every thread running within the given group (in this scenario, all Java<a href="#*">*</a> threads). The thread manager then returns the iterator for traversing the list of suspended threads. GC uses this iterator to analyze each Java<a href="#*">*</a> thread with respect to live references and then does a garbage collection. After it is complete, GC instructs the thread manager to resume all suspended threads. <a href="#Top">Back to Top</a> <h2> <a name="Thread_locking"></a>6.4 Thread Locking </h2> Locking with the thread manager can be done by means of a mutex or a thin monitor. The mutex is preferable in case of high contention, while thin monitors are better optimized for space. This section describes a scenario when the VM core attempts to lock a resource from multiple threads, T1 and T2. The major stages of the process of locking and unlocking are shown in Figure 15. <img alt="locking and unlocking a mutex" border="0" src= "images/locking.gif"> Figure 15: Locking with a Mutex Initially, the mutex is not occupied, that is, the label lock is set to zero. Thread T1 calls the <code>hymutex_lock()</code> function, which instructs the thread manager to mark the mutex as locked by T1. T2 can also call the <code>hymutex_lock()</code> function later, and if it happens to call on a lock already occupied, then T2 is placed into the internal waiting queue associated with the mutex and gets blocked until T1 unlocks the mutex. The T1 thread calls <code>hymutex_unlock()</code> to release the mutex, which enables the mutex to extract T2 from the queue, transfer the lock ownership to this thread, and to notify T2 that it can wake up. <a href="#Top">Back to Top</a> <h2> <a name="Monitor_scenarios"></a>6.5 Monitor enter and exit </h2> Locking Java<a href="#*">*</a> monitors implies interaction between the thread manager and the VM core since the thread manager requires the memory address within the Java<a href= "#*">*</a> object where it keeps the lock data. The process of locking Java<a href="#*">*</a> monitors is shown on Figure 16 below. When a synchronized section occurs in Java<a href="#*">*</a> code, the following steps are taken: <ol start="1" type="1"> <li> The code generated by the JIT compiler calls the <code>hythread_monitor_enter()</code> helper function provided by the thread manager. The helper function provides a chunk of code (stub) that can be in-lined by the JIT compiler directly into the generated assembly code. <li> The <code>hythread_monitor_enter()</code> helper calls the <code>vm_object_get_lockword_addr()</code> function of VM core component to find out the physical address of the lock word within the Java<a href="#*">*</a> object. <li> The helper calls the <code>thin_monitor_try_lock()</code> function in order to acquire the lock associated with the object. <li> In case the lock is acquired, the helper returns. This is the fast path to acquiring the Java<a href="#*">*</a> monitor. In this scenario, the helper does not need to switch between Java<a href="#*">*</a> and native frames and the <code>thin_monitor_try_enter()</code> function is called directly involving no Java<a href="#*">*</a> objects. Otherwise, the helper code enters a slow path by switching between Java<a href="#*">*</a> and native code (see the actions of pushing an M2nFrame and creating the local handle in the figure below). <li> The helper calls the <code>jthread_monitor_enter()</code> function, which works with the Java<a href="#*">*</a> object as with JNI code. </ol> <img alt="slow and fast paths to locking a Java* monitor" src="images/Monitors.gif"> Figure 16: Locking Java<a href="#*">*</a> Monitors <a href="#Top">Back to Top</a> <h1> <a name="References"></a>7. References </h1> This section lists the resources used in this document and other related documents. [<a name="J2SESpec"></a>1] J2SE 1.5.0 specification, <a href="http://java.sun.com/j2se/1.5.0/docs/api/" target= "_blank">http://java.sun.com/j2se/1.5.0/docs/api/</a> [<a name="JVMTI_ref"></a>2] JVM Tool Interface Specification, <a href= "http://java.sun.com/j2se/1.5.0/docs/guide/jvmti/jvmti.html" target= "_blank">http://java.sun.com/j2se/1.5.0/docs/guide/jvmti/jvmti.html</a> [<a name="JNI_ref"></a>3] Java<a href="#*">*</a> Native Interface Specification, <a href= "http://java.sun.com/j2se/1.5.0/docs/guide/jni/spec/jniTOC.html" target= "_blank">http://java.sun.com/j2se/1.5.0/docs/guide/jni/spec/jniTOC.html</a> [<a name="APR_ref"></a>4] Apache Portable Runtime project, <a href="http://apr.apache.org/" target= "_blank">http://apr.apache.org/</a> [<a name="posix_ref"></a>5] POSIX standard in threading, <a href= "http://www.opengroup.org/onlinepubs/009695399/idx/threads.html" target= "_blank">http://www.opengroup.org/onlinepubs/009695399/idx/threads.html</a> [<a name="thin_lock_ref"></a>6] David F. Bacon, Ravi Konuru, Chet Murthy, Mauricio Serrano, Thin locks: featherweight synchronization for Java, <a href= "http://portal.acm.org/citation.cfm?id=277734" target= "_blank">http://portal.acm.org/citation.cfm?id=277734</a> [<a name="lock_overhead_ref"></a>7] Kiyokuni Kawachiya Akira Koseki Tamiya Onodera, Lock Reservation: Java Locks Can Mostly Do Without Atomic Operation, <a href= "http://portal.acm.org/citation.cfm?id=582433" target= "_blank">http://portal.acm.org/citation.cfm?id=582433</a> [<a name="hythread_ref"></a>8] HyThread documentation, <a href= "http://svn.apache.org/viewvc/incubator/harmony/enhanced/classlib/trunk/doc/vm_doc/html/group__Thread.html" target= "_blank">http://svn.apache.org/viewvc/incubator/harmony/enhanced/classlib/trunk/doc/vm_doc/html/group__Thread.html</a> � � <a href="#top">Back to Top</a> (C) Copyright 2006 Intel Corporation. All rights reserved. <a name="*"></a>* Other brands and names are the property of their respective owners. </body> </html>

vm/thread/doc/ThreadManager.htm (1,912 lines of code) (raw):