|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
*/ |
|
|
|
/** |
|
* This package provides a mechanism for defining and |
|
* inserting tracepoints into Java-technology based applications, which |
|
* can then be monitored by the tracing tools available on the system. |
|
* <p> |
|
* To add tracepoints to a program, you must first decide where to place the |
|
* tracepoints, what the logical names are for these points, what information |
|
* will be available to the tracing mechanisms at each point, and decide upon |
|
* any logical grouping. |
|
* <p> |
|
* You add instrumentation to a program in three steps: |
|
* <ul> |
|
* <li>First, declare tracepoints by creating interfaces to define |
|
* them, and include these interfaces in the program definition. |
|
* The declared interfaces are standard Java technology-based |
|
* interfaces and are compiled with the program.</li> |
|
* <li>Second, add code in the application to create an instance of the |
|
* interface at some point during the initialization of the application, |
|
* using a factory class provided by the system. The reference to the |
|
* instance can be stored as a global static, or passed as context to all |
|
* the places where it is needed.</li> |
|
* <li>Finally, add the actual tracepoints to the desired locations in the |
|
* application by inserting a call to one of the methods defined in the |
|
* interface, via the factory-created reference.</li> |
|
* </ul> |
|
* <p> |
|
* The method calls representing the tracepoints have no logical |
|
* impact on the program. The side effect of the call is that any |
|
* activated tracing mechanisms will be notified that the tracepoint has |
|
* been hit, and will take whatever actions are appropriate (for example, |
|
* logging the tracepoint, or triggering a DTrace probe, etc.). In most |
|
* cases, the impact on performance of adding tracepoints to the application |
|
* will be minimal. |
|
* <p> |
|
* Each logical grouping of tracepoints should be defined in a common |
|
* interface, called a <i>provider</i>. An application can have one or many |
|
* providers. Each provider is independent and can be created whenever |
|
* it is appropriate for that provider, for example, when a subsytem is |
|
* initialized. Providers should be disposed of when they are no longer |
|
* needed, to free up any associated system resources. Each tracepoint |
|
* in a provider is represented by a method in that interface. These methods |
|
* are referred to as <i>probes</i>. The method signature determines the probe |
|
* parameters. A call to the method with the specified parameters triggers |
|
* the probe and makes its parameter values visible to any associated tracing |
|
* mechanism. |
|
* <p> |
|
* User-defined interfaces which represent providers must extend the |
|
* {@code Provider} interface. To activate the system-defined |
|
* tracing mechanisms, you must obtain an instance of the |
|
* {@code ProviderFactory} class, and pass the class of the provider to |
|
* the {@code createProvider()} method. The returned instance is then used to |
|
* trigger the probes later in the application. |
|
* <p> |
|
* In addition to triggering the probes, the provider instance can be used |
|
* to obtain direct references to the {@code Probe} objects, which can be used |
|
* directly for triggering, or can be queried to determine whether the probe is |
|
* currently being traced. The {@code Provider} interface also defines a |
|
* {@code Provider.dispose()} method which is used to free up any resources |
|
* that might be associated with that provider. |
|
* <p> |
|
* When a probe is triggered, any activated tracing system will be given |
|
* the provider name, the probe name, and the values of the probe arguments. |
|
* The tracing system is free to consume this data is whatever way is |
|
* appropriate. |
|
* By default, the provider name is the same as the class name of the interface |
|
* that defines the provider. Similarly, the probe name is |
|
* the name of the method that defines the probe. These default values |
|
* can be over-ridden by annotations. The provider definition can be |
|
* annotated with the {@code @ProviderName} annotation, whose value will |
|
* indicate the provider name that the tracing system will use. Similarly, |
|
* the {@code @ProbeName} annotation annotates a declared method and |
|
* indicates the probe name that should be used in the place of the |
|
* method name. These annotations can be used to define providers and |
|
* probes with the same name, in cases where the semantics of the Java language |
|
* may prevent this. |
|
* <p> |
|
* Here is a very small and simple usage example: |
|
* <p> |
|
* |
|
<PRE> |
|
import com.sun.tracing.Provider; |
|
import com.sun.tracing.ProviderFactory; |
|
|
|
interface MyProvider extends Provider { |
|
void startProbe(); |
|
void finishProbe(int value); |
|
} |
|
|
|
public class MyApplication { |
|
public static void main(String argv[]) { |
|
ProviderFactory factory = ProviderFactory.getDefaultFactory(); |
|
MyProvider trace = factory.createProvider(MyProvider.class); |
|
|
|
trace.startProbe(); |
|
int result = foo(); |
|
trace.finishProbe(result); |
|
|
|
trace.dispose(); |
|
} |
|
} |
|
</PRE> |
|
* <p> |
|
* The Java Development Kit (JDK) currently only includes one system-defined |
|
* tracing framework: DTrace. DTrace is enabled automatically whenever an |
|
* application is run on a system and a JDK release that supports it. When |
|
* DTrace is enabled, probes are made available for listing and matching by |
|
* DTrace scripts as soon as the provider is created. At the tracepoint, an |
|
* associated DTrace script is informed of the creation of the provider, and |
|
* it takes whatever action it is designed to take. Tracepoints in the |
|
* program have the following DTrace probe names:<br> |
|
* {@code <provider><pid>:<module>:<function>:<probe>} |
|
* Where: |
|
* <ul> |
|
* <li>{@code <provider>} the provider name as specified by the application</li> |
|
* <li>{@code <pid>} the operating system process ID</li> |
|
* <li>{@code <module>} undefined, unless specified by the application</li> |
|
* <li>{@code <function>} undefined, unless specified by the application</li> |
|
* <li>{@code <probe>} the probe name as specified by the application</li> |
|
* </ul> |
|
* <p> |
|
* The {@code com.sun.tracing.dtrace} package contains additional |
|
* annotations that can be used to control the names used for the |
|
* <code>module</code> and <code>function</code> fields, as well as annotations |
|
* that can be added to the provider to control probe stability and dependency |
|
* attributes. |
|
* <p> |
|
* Integer, float and string probe parameters are made available to DTrace |
|
* using |
|
* the built-in argument variables, {@code arg0 ... arg_n}. Integer-types |
|
* are passed by value (boxed values are unboxed), floating-point types are |
|
* passed as encoded integer |
|
* arguments, and {@code java.lang.String} objects are converted |
|
* to UTF8 strings, so they can be read into the DTrace script using the |
|
* {@code copyinstr()} intrinsic. Non-string and non-boxed primitive |
|
* reference arguments are only |
|
* placeholders and have no value. |
|
* <p> |
|
* Using the example above, with a theoretical process ID of 123, these are |
|
* the probes that can be traced from DTrace: |
|
<PRE> |
|
MyProvider123:::startProbe |
|
MyProvider123:::finishProbe |
|
</PRE> |
|
* When {@code finishProbe} executes, {@code arg0} will contain the |
|
* value of {@code result}. |
|
* <p> |
|
* The DTrace tracing mechanism is enabled for all providers, apart from in the |
|
* following circumstances: |
|
* <ul> |
|
* <li>DTrace is not supported on the underlying system.</li> |
|
* <li>The property {@code com.sun.tracing.dtrace} is set to "disable".</li> |
|
* <li>The RuntimePermission {@code com.sun.tracing.dtrace.createProvider} |
|
* is denied to the process.</li> |
|
* </ul> |
|
* <p> |
|
*/ |
|
|
|
package com.sun.tracing; |