Blame - docs/jni-tips.html - platform/dalvik

blob: e2c3b8509b6cb32855c0aaaafcc7b899c4200eea [file] [log] [blame]

The Android Open Source Project	f6c3871	2009-03-03 19:28:47 -0800	[diff] [blame]	1	<html>
				2	<head>
				3	<title>Android JNI Tips</title>
				4	<link rel=stylesheet href="android.css">
				5	</head>
				6
				7	<body>
				8	<h1><a name="JNI_Tips"></a>Android JNI Tips</h1>
				9	<p>
				10	</p><p>
				11	</p><ul>
				12	<li> <a href="#What_s_JNI_">What's JNI?</a>
				13	</li>
				14	<li> <a href="#JavaVM_and_JNIEnv">JavaVM and JNIEnv</a>
				15
				16	</li>
				17	<li> <a href="#jclassID_jmethodID_and_jfieldID">jclassID, jmethodID, and jfieldID</a>
				18	</li>
				19	<li> <a href="#local_vs_global_references">Local vs. Global References</a>
				20	</li>
				21	<li> <a href="#UTF_8_and_UTF_16_strings">UTF-8 and UTF-16 Strings</a>
				22	</li>
				23	<li> <a href="#Arrays">Primitive Arrays</a>
				24	</li>
				25	<li> <a href="#RegionCalls">Region Calls</a>
				26	</li>
				27	<li> <a href="#Exceptions">Exceptions</a>
				28	</li>
				29
				30	<li> <a href="#Extended_checking">Extended Checking</a>
				31	</li>
				32	<li> <a href="#Native_Libraries">Native Libraries</a>
				33	</li>
				34	<li> <a href="#64bit">64-bit Considerations</a>
				35	</li>
				36
				37	<li> <a href="#Unsupported">Unsupported Features</a>
				38	</ul>
				39	<p>
				40	<noautolink>
				41	</noautolink></p><p>
				42	</p><h2><a name="What_s_JNI_"> </a> What's JNI? </h2>
				43	<p>
				44
				45	JNI is the Java Native Interface. It defines a way for code written in the
				46	Java programming language to interact with native
				47	code, e.g. functions written in C/C++. It's VM-neutral, has support for loading code from
				48	dynamic shared libraries, and while cumbersome at times is reasonably efficient.
				49	</p><p>
				50	You really should read through the
				51	<a href="http://java.sun.com/javase/6/docs/technotes/guides/jni/spec/jniTOC.html">JNI spec for J2SE 1.6</a>
				52	to get a sense for how JNI works and what features are available. Some
				53	aspects of the interface aren't immediately obvious on
				54	first reading, so you may find the next few sections handy.
				55	The more detailed <i>JNI Programmer's Guide and Specification</i> can be found
				56	<a href="http://java.sun.com/docs/books/jni/html/jniTOC.html">here</a>.
				57	</p><p>
				58	</p><p>
				59	</p><h2><a name="JavaVM_and_JNIEnv"> </a> JavaVM and JNIEnv </h2>
				60	<p>
				61	JNI defines two key data structures, "JavaVM" and "JNIEnv". Both of these are essentially
				62	pointers to pointers to function tables. (In the C++ version, it's a class whose sole member
				63	is a pointer to a function table.) The JavaVM provides the "invocation interface" functions,
				64	which allow you to create and destroy the VM. In theory you can have multiple VMs per process,
				65	but Android's VMs only allow one.
				66	</p><p>
				67	The JNIEnv provides most of the JNI functions. Your native functions all receive a JNIEnv as
				68	the first argument.
				69	</p><p>
				70
				71	On some VMs, the JNIEnv is used for thread-local storage. For this reason, <strong>you cannot share a JNIEnv between threads</strong>.
				72	If a piece of code has no other way to get its JNIEnv, you should share
				73	the JavaVM, and use JavaVM->GetEnv to discover the thread's JNIEnv.
				74	</p><p>
				75	The C and C++ declarations of JNIEnv and JavaVM are different. "jni.h" provides different typedefs
				76	depending on whether it's included into ".c" or ".cpp". For this reason it's a bad idea to
				77	include JNIEnv arguments in header files included by both languages. (Put another way: if your
				78	header file requires "#ifdef __cplusplus", you may have to do some extra work if anything in
				79	that header refers to JNIEnv.)
				80	</p><p>
				81	</p><p>
				82	</p><h2><a name="jclassID_jmethodID_and_jfieldID"> jclassID, jmethodID, and jfieldID </a></h2>
				83	<p>
				84	If you want to access an object's field from native code, you would do the following:
				85	</p><p>
				86	</p><ul>
				87	<li> Get the class object reference for the class with <code>FindClass</code>
				88	</li>
				89	<li> Get the field ID for the field with <code>GetFieldID</code>
				90	</li>
				91	<li> Get the contents of the field with something appropriate, e.g.
				92	<code>GetIntField</code>
				93	</li>
				94	</ul>
				95	<p>
				96	Similarly, to call a method, you'd first get a class object reference and then a method ID. The IDs are often just
				97	pointers to internal VM data structures. Looking them up may require several string
				98	comparisons, but once you have them the actual call to get the field or invoke the method
				99	is very quick.
				100	</p><p>
				101	If performance is important, it's useful to look the values up once and cache the results
				102	in your native code. Because we are limiting ourselves to one VM per process, it's reasonable
				103	to store this data in a static local structure.
				104	</p><p>
				105	The class references, field IDs, and method IDs are guaranteed valid until the class is unloaded. Classes
				106	are only unloaded if all classes associated with a ClassLoader can be garbage collected,
				107	which is rare but will not be impossible in our system. The jclassID
				108	is a class reference and <strong>must be protected</strong> with a call
				109	to <code>NewGlobalRef</code> (see the next section).
				110	</p><p>
				111	If you would like to cache the IDs when a class is loaded, and automatically re-cache them
				112	if the class is ever unloaded and reloaded, the correct way to initialize
				113	the IDs is to add a piece of code that looks like this to the appropriate class:
				114	</p><p>
				115
				116	</p><pre> /*
				117	* We use a class initializer to allow the native code to cache some
				118	* field offsets.
				119	*/
				120
				121	/*
				122	* A native function that looks up and caches interesting
				123	* class/field/method IDs for this class. Returns false on failure.
				124	*/
				125	native private static boolean nativeClassInit();
				126
				127	/*
				128	* Invoke the native initializer when the class is loaded.
				129	*/
				130	static {
				131	if (!nativeClassInit())
				132	throw new RuntimeException("native init failed");
				133	}
				134	</pre>
				135	<p>
				136	Create a nativeClassInit method in your C/C++ code that performs the ID lookups. The code
				137	will be executed once, when the class is initialized. If the class is ever unloaded and
				138	then reloaded, it will be executed again. (See the implementation of java.io.FileDescriptor
				139	for an example in our source tree.)
				140	</p><p>
				141	</p><p>
				142	</p><p>
				143	</p><h2><a name="local_vs_global_references"> Local vs. Global References </a></h2>
				144	<p>
				145	Every object that JNI returns is a "local reference". This means that it's valid for the
				146	duration of the current native method in the current thread.
				147	<strong>Even if the object itself continues to live on after the native method returns, the reference is not valid.</strong>
				148	This applies to all sub-classes of jobject, including jclass and jarray.
				149	(Dalvik VM will warn you about this when -Xcheck:jni is enabled.)
				150	</p><p>
				151
				152	If you want to hold on to a reference for a longer period, you must use a "global" reference.
				153	The <code>NewGlobalRef</code> function takes the local reference as
				154	an argument and returns a global one:
				155
				156	<p><pre>jobject* localRef = [...];
				157	jobject* globalRef;
				158	globalRef = env->NewGlobalRef(localRef);
				159	</pre>
				160
				161	The global reference is guaranteed to be valid until you call
				162	<code>DeleteGlobalRef</code>.
				163	</p><p>
				164	All JNI methods accept both local and global references as arguments.
				165	</p><p>
				166	Programmers are required to "not excessively allocate" local references. In practical terms this means
				167	that if you're creating large numbers of local references, perhaps while running through an array of
				168	Objects, you should free them manually with
				169	<code>DeleteLocalRef</code> instead of letting JNI do it for you. The
				170	VM is only required to reserve slots for
				171	16 local references, so if you need more than that you should either delete as you go or use
				172	<code>EnsureLocalCapacity</code> to reserve more.
				173	</p><p>
				174	Note: method and field IDs are just 32-bit identifiers, not object
				175	references, and should not be passed to <code>NewGlobalRef</code>. The raw data
				176	pointers returned by functions like <code>GetStringUTFChars</code>
				177	and <code>GetByteArrayElements</code> are also not objects.
				178	</p><p>
				179	One unusual case deserves separate mention. If you attach a native
				180	thread to the VM with AttachCurrentThread, the code you are running will
				181	never "return" to the VM until the thread detaches from the VM. Any local
				182	references you create will have to be deleted manually unless the thread
				183	is about to exit or detach.
				184	</p><p>
				185	</p><p>
				186	</p><p>
				187	</p><h2><a name="UTF_8_and_UTF_16_strings"> </a> UTF-8 and UTF-16 Strings </h2>
				188	<p>
				189	The Java programming language uses UTF-16. For convenience, JNI provides methods that work with "modified UTF-8" encoding
				190	as well. (Some VMs use the modified UTF-8 internally to store strings; ours do not.) The
				191	modified encoding only supports the 8- and 16-bit forms, and stores ASCII NUL values in a 16-bit encoding.
				192	The nice thing about it is that you can count on having C-style zero-terminated strings,
				193	suitable for use with standard libc string functions. The down side is that you cannot pass
				194	arbitrary UTF-8 data into the VM and expect it to work correctly.
				195	</p><p>
				196	It's usually best to operate with UTF-16 strings. With our current VMs, the
				197	<code>GetStringChars</code> method
				198	does not require a copy, whereas <code>GetStringUTFChars</code> requires a malloc and a UTF conversion. Note that
				199	<strong>UTF-16 strings are not zero-terminated</strong>, and \u0000 is allowed,
				200	so you need to hang on to the string length as well as
				201	the string pointer.
				202
				203	</p><p>
				204	<strong>Don't forget to Release the strings you Get</strong>. The string functions return <code>jchar</code> or <code>jbyte</code>, which
				205	are pointers to primitive types rather than local references. They are
				206	guaranteed valid until Release is called, which means they are not
				207	released when the native method returns.
				208	</p><p>
				209	</p><p>
				210
				211
				212	</p><h2><a name="Arrays"> </a> Primitive Arrays </h2>
				213	<p>
				214	JNI provides functions for accessing the contents of array objects.
				215	While arrays of objects must be accessed one entry at a time, arrays of
				216	primitives can be read and written directly as if they were declared in C.
				217	</p><p>
				218	To make the interface as efficient as possible without constraining
				219	the VM implementation,
				220	the <code>Get<PrimitiveType>ArrayElements</code> family of calls
				221	allows the VM to either return a pointer to the actual elements, or
				222	allocate some memory and make a copy. Either way, the raw pointer returned
				223	is guaranteed to be valid until the corresponding <code>Release</code> call
				224	is issued (which implies that, if the data wasn't copied, the array object
				225	will be pinned down and can't be relocated as part of compacting the heap).
				226	<strong>You must Release every array you Get.</strong> Also, if the Get
				227	call fails, you must ensure that your code doesn't try to Release a NULL
				228	pointer later.
				229	</p><p>
				230	You can determine whether or not the data was copied by passing in a
				231	non-NULL pointer for the <code>isCopy</code> argument. This is rarely
				232	useful.
				233	</p><p>
				234	The <code>Release</code> call takes a <code>mode</code> argument that can
				235	have one of three values. The actions performed by the VM depend upon
				236	whether it returned a pointer to the actual data or a copy of it:
				237	<ul>
				238	<li><code>0</code>
				239	<ul>
				240	<li>Actual: the array object is un-pinned.
				241	<li>Copy: data is copied back. The buffer with the copy is freed.
				242	</ul>
				243	<li><code>JNI_COMMIT</code>
				244	<ul>
				245	<li>Actual: does nothing.
				246	<li>Copy: data is copied back. The buffer with the copy
				247	<strong>is not freed</strong>.
				248	</ul>
				249	<li><code>JNI_ABORT</code>
				250	<ul>
				251	<li>Actual: the array object is un-pinned. Earlier
				252	writes are <strong>not</strong> aborted.
				253	<li>Copy: the buffer with the copy is freed; any changes to it are lost.
				254	</ul>
				255	</ul>
				256	</p><p>
				257	One reason for checking the <code>isCopy</code> flag is to know if
				258	you need to call <code>Release</code> with <code>JNI_COMMIT</code>
				259	after making changes to an array -- if you're alternating between making
				260	changes and executing code that uses the contents of the array, you may be
				261	able to
				262	skip the no-op commit. Another possible reason for checking the flag is for
				263	efficient handling of <code>JNI_ABORT</code>. For example, you might want
				264	to get an array, modify it in place, pass pieces to other functions, and
				265	then discard the changes. If you know that JNI is making a new copy for
				266	you, there's no need to create another "editable" copy. If JNI is passing
				267	you the original, then you do need to make your own copy.
				268	</p><p>
				269	Some have asserted that you can skip the <code>Release</code> call if
				270	<code>*isCopy</code> is false. This is not the case. If no copy buffer was
				271	allocated, then the original memory must be pinned down and can't be moved by
				272	the garbage collector.
				273	</p><p>
				274	Also note that the <code>JNI_COMMIT</code> flag does NOT release the array,
				275	and you will need to call <code>Release</code> again with a different flag
				276	eventually.
				277	</p><p>
				278	</p><p>
				279
				280
				281	</p><h2><a name="RegionCalls"> Region Calls </a></h2>
				282
				283	<p>
				284	There is an alternative to calls like <code>Get<Type>ArrayElements</code>
				285	and <code>GetStringChars</code> that may be very helpful when all you want
				286	to do is copy data in or out. Consider the following:
				287	<pre>
				288	jbyte* data = env->GetByteArrayElements(array, NULL);
				289	if (data != NULL) {
				290	memcpy(buffer, data, len);
				291	env->ReleaseByteArrayElements(array, data, JNI_ABORT);
				292	}
				293	</pre>
				294	<p>
				295	This grabs the array, copies the first <code>len</code> byte
				296	elements out of it, and then releases the array. Depending upon the VM
				297	policies the <code>Get</code> call will either pin or copy the array contents.
				298	We copy the data (for perhaps a second time), then call Release; in this case
				299	we use <code>JNI_ABORT</code> so there's no chance of a third copy.
				300	</p><p>
				301	We can accomplish the same thing with this:
				302	<pre>
				303	env->GetByteArrayRegion(array, 0, len, buffer);
				304	</pre>
				305	</p><p>
				306	This accomplishes the same thing, with several advantages:
				307	<ul>
				308	<li>Requires one JNI call instead of 3, reducing overhead.
				309	<li>Doesn't require pinning or extra data copies.
				310	<li>Reduces the risk of programmer error -- no need to match up
				311	<code>Get</code> and <code>Release</code> calls.
				312	</ul>
				313	</p><p>
				314	Similarly, you can use the <code>Set<Type>ArrayRegion</code> call
				315	to copy data into an array, and <code>GetStringRegion</code> or
				316	<code>GetStringUTFRegion</code> to copy characters out of a
				317	<code>String</code>.
				318
				319
				320	</p><h2><a name="Exceptions"> Exceptions </a></h2>
				321	<p>
				322	<strong>You may not call most JNI functions while an exception is pending.</strong>
				323	Your code is expected to notice the exception (via the function's return value,
				324	<code>ExceptionCheck()</code>, or <code>ExceptionOccurred()</code>) and return,
				325	or clear the exception and handle it.
				326	</p><p>
				327	The only JNI functions that you are allowed to call while an exception is
				328	pending are:
				329	<font size="-1"><ul>
				330	<li>DeleteGlobalRef
				331	<li>DeleteLocalRef
				332	<li>DeleteWeakGlobalRef
				333	<li>ExceptionCheck
				334	<li>ExceptionClear
				335	<li>ExceptionDescribe
				336	<li>ExceptionOccurred
				337	<li>MonitorExit
				338	<li>PopLocalFrame
				339	<li>PushLocalFrame
				340	<li>Release<PrimitiveType>ArrayElements
				341	<li>ReleasePrimitiveArrayCritical
				342	<li>ReleaseStringChars
				343	<li>ReleaseStringCritical
				344	<li>ReleaseStringUTFChars
				345	</ul></font>
				346	</p><p>
				347	Note that exceptions thrown by interpreted code do not "leap over" native code,
				348	and C++ exceptions thrown by native code are not handled by Dalvik.
				349	The JNI <code>Throw</code> and <code>ThrowNew</code> instructions just
				350	set an exception pointer in the current thread. Upon returning to the VM from
				351	native code, the exception will be noted and handled appropriately.
				352	</p><p>
				353	Native code can "catch" an exception by calling <code>ExceptionCheck</code> or
				354	<code>ExceptionOccurred</code>, and clear it with
				355	<code>ExceptionClear</code>. As usual,
				356	discarding exceptions without handling them can lead to problems.
				357	</p><p>
				358	There are no built-in functions for manipulating the Throwable object
				359	itself, so if you want to (say) get the exception string you will need to
				360	find the Throwable class, look up the method ID for
				361	<code>getMessage "()Ljava/lang/String;"</code>, invoke it, and if the result
				362	is non-NULL use <code>GetStringUTFChars</code> to get something you can
				363	hand to printf or a LOG macro.
				364
				365	</p><p>
				366	</p><p>
				367	</p><h2><a name="Extended_checking"> Extended Checking </a></h2>
				368	<p>
				369	JNI does very little error checking. Calling <code>SetFieldInt</code>
				370	on an Object field will succeed, even if the field is marked
				371	<code>private</code> and <code>final</code>. The
				372	goal is to minimize the overhead on the assumption that, if you've written it in native code,
				373	you probably did it for performance reasons.
				374	</p><p>
				375	Some VMs support extended checking with the "<code>-Xcheck:jni</code>" flag. If the flag is set, the VM
				376	puts a different table of functions into the JavaVM and JNIEnv pointers. These functions do
				377	an extended series of checks before calling the standard implementation.
				378
				379	</p><p>
				380	Some things that may be verified:
				381	</p><p>
				382	</p>
				383	<ul>
				384	<li> Check for null pointers where not allowed.
				385	<li>
				386	<li> Verify argument type correctness (jclass is a class object,
				387	jfieldID points to field data, jstring is a java.lang.String).
				388	</li>
				389	<li> Field type correctness, e.g. don't store a HashMap in a String field.
				390	</li>
				391	<li> Check to see if an exception is pending on calls where pending exceptions are not legal.
				392	</li>
				393	<li> Check for calls to inappropriate functions between Critical get/release calls.
				394	</li>
				395	<li> Check that JNIEnv structs aren't being shared between threads.
				396
				397	</li>
				398	<li> Make sure local references aren't used outside their allowed lifespan.
				399	</li>
				400	<li> UTF-8 strings contain valid "modified UTF-8" data.
				401	</li>
				402	</ul>
				403	<p>Accessibility of methods and fields (i.e. public vs. private) is not
				404	checked.
				405	<p>
				406	The Dalvik VM supports the <code>-Xcheck:jni</code> flag. For a
				407	description of how to enable it for Android apps, see
				408	<a href="embedded-vm-control.html">Controlling the Embedded VM</a>.
				409	It's currently enabled by default in the Android emulator and on
				410	"engineering" device builds.
				411
				412	</p><p>
				413	JNI checks can be modified with the <code>-Xjniopts</code> command-line
				414	flag. Currently supported values include:
				415	</p>
				416	<blockquote><dl>
				417	<dt>forcecopy
				418	<dd>When set, any function that can return a copy of the original data
				419	(array of primitive values, UTF-16 chars) will always do so. The buffers
				420	are over-allocated and surrounded with a guard pattern to help identify
				421	code writing outside the buffer, and the contents are erased before the
				422	storage is freed to trip up code that uses the data after calling Release.
				423	<dt>warnonly
				424	<dd>By default, JNI "warnings" cause the VM to abort. With this flag
				425	it continues on.
				426	</dl></blockquote>
				427
				428
				429	</p><p>
				430	</p><h2><a name="Native_Libraries"> Native Libraries </a></h2>
				431	<p>
				432	You can load native code from shared libraries with the standard
				433	<code>System.loadLibrary()</code> call. The
				434	preferred way to get at your native code is:
				435	</p><p>
				436	</p><ul>
				437	<li> Call <code>System.loadLibrary()</code> from a static class initializer. (See the earlier example, where one is used to call nativeClassInit().) The argument is the "undecorated" library name, e.g. to load "libfubar.so" you would pass in "fubar".
				438
				439	</li>
				440	<li> Provide a native function: <code><strong>jint JNI_OnLoad(JavaVM* vm, void* reserved)</strong></code>
				441	</li>
				442	<li>In <code>JNI_OnLoad</code>, register all of your native methods. You
				443	should declare
				444	the methods "static" so the names don't take up space in the symbol table
				445	on the device.
				446	</li>
				447	</ul>
				448	<p>
				449	The <code>JNI_OnLoad</code> function should look something like this if
				450	written in C:
				451	</p><blockquote><pre>jint JNI_OnLoad(JavaVM* vm, void* reserved)
				452	{
				453	JNIEnv* env;
				454	if ((vm)->GetEnv(vm, (void*) &env, JNI_VERSION_1_4) != JNI_OK)
				455	return -1;
				456
				457	/* get class with (env)->FindClass /
				458	/* register methods with (env)->RegisterNatives /
				459
				460	return JNI_VERSION_1_4;
				461	}
				462	</pre></blockquote>
				463	</p><p>
				464	You can also call <code>System.load()</code> with the full path name of the
				465	shared library. For Android apps, you can get the full path to the
				466	application's private data storage area from the context object.
				467	</p><p>
				468	Dalvik does support "discovery" of native methods that are named in a
				469	specific way (see <a href="http://java.sun.com/javase/6/docs/technotes/guides/jni/spec/design.html#wp615">
				470	the JNI spec</a> for details), but this is a less desirable
				471	approach. It requires more space in the shared object symbol table,
				472	loading is slower because it requires string searches through all of the
				473	loaded shared libraries, and if a method signature is wrong you won't know
				474	about it until the first time the method is actually used.
				475	</p><p>
				476
				477
				478	</p><h2><a name="64bit"> 64-bit Considerations </a></h2>
				479
				480	<p>
				481	Android is currently expected to run on 32-bit platforms. In theory it
				482	could be built for a 64-bit system, but that is not a goal at this time.
				483	For the most part this isn't something that you will need to worry about
				484	when interacting with native code,
				485	but it becomes significant if you plan to store pointers to native
				486	structures in integer fields in an object. To support architectures
				487	that use 64-bit pointers, <strong>you need to stash your native pointers in a
				488	<code>long</code> field rather than an <code>int</code></strong>.
				489
				490
				491	</p><h2><a name="Unsupported"> Unsupported Features </a></h2>
				492	<p>All JNI 1.6 features are supported, with the following exceptions:
				493	<ul>
				494	<li><code>DefineClass</code> is not implemented. Dalvik does not use
				495	Java bytecodes or class files, so passing in binary class data
				496	doesn't work. Translation facilities may be added in a future
				497	version of the VM.</li>
				498	<li><code>NewWeakGlobalRef</code> and <code>DeleteWeakGlobalRef</code>
				499	are not implemented. The
				500	VM supports weak references, but not JNI "weak global" references.
				501	These will be supported in a future release.</li>
				502	<li><code>GetObjectRefType</code> (new in 1.6) is implemented but not fully
				503	functional -- it can't always tell the difference between "local" and
				504	"global" references.</li>
				505	</ul>
				506
				507	</p>
				508
				509	<address>Copyright © 2008 The Android Open Source Project</address>
				510
				511	</body>
				512	</html>