1# Node-API FAQs
2
3## What should I do when "undefined/not callable" or specific error message is reported for xxx after "import xxx from libxxx.so" is executed?
4
51. Check whether the module name in the .cpp file used in module registration is the same as that in the .so file name.
6   If the module name is **entry**, the .so file name must be **libentry.so**, and the **nm_modname** field in **napi_module** must be **entry**. The module names must be of the same case.
7
82. Check whether the .so file is successfully loaded.
9
10   Check the log related to module loading during the application startup. Search for the keyword "dlopen" and check for error information. Possible causes include the following:
11
12   - The file to be loaded does not exist or is in a blocklist.
13   - The application does not have the required permission.
14   - The **nm_modname** value does not match the module name in multi-thread scenarios (such as worker threads and taskpool). The module names must be of the same case.
153. Check whether the dependency .so files are successfully loaded.
16   Check that all the dependency .so files are packaged into the application and the application has the permission to open them.
17
184. Check whether the module import mode matches the .so file path.
19   If the module is imported by using **import xxx from '\@ohos.yyy.zzz'**, you may find **libzzz.z.so** or **libzzz_napi.z.so** in **/system/lib/module/yyy** (for a 32-bit system) or **/system/lib64/module/yyy** (for a 64-bit system). If the .so file does not exist or the file names do not match, an error containing the keyword "dlopen" will be reported.
20
21
22
23| **Error Log**| **Solution**|
24| -------- | -------- |
25| module $SO is not allowed to load in restricted runtime | The module, identified by **$SO**, is not allowed for the worker thread running in a restricted environment and cannot be loaded. You are advised to delete the module.|
26| module $SO is in blocklist, loading prohibited | The module, identified by **$SO**, is in the blocklist due to the control of the widget or Extension, and cannot be loaded. You are advised to delete the module.|
27| load module failed. $ERRMSG | The dynamic library fails to be loaded. **$ERRMSG** indicates the cause of the loading failure. Possible causes include the following:<br>- The .so file to be loaded does not exist.<br>- The dependency .so file does not exist. <br>- Undefined symbol is found. <br>Locate the cause based on the error message.|
28| try to load abc file from $FILEPATH failed. | You can load either a dynamic library or an .abc file. If this log information is displayed when you attempt to load a dynamic library, ignore this message. If it is displayed when you attempt to load an .abc file, the .abc file does not exist. **$FILEPATH** indicates the module path.|
29
305. If specific error message is reported, identify the fault based on the error message.
31
32| **Error message** | **Fault Analysis & Solution**|
33| -------- | -------- |
34| First attempt: $ERRMSG | Loading the .so file with the module name of "xxx" fails. *$ERRMSG* indicates the error information.|
35| Second attempt: $ERRMSG | Loading the .so file with the module name of "xxx_napi" fails. *$ERRMSG* indicates the error information.|
36| try to load abc file from xxx failed | Loading the .abc file fails. *xxx* indicates the name of the .abc file.|
37| module xxx is not allowed to load in restricted runtime. | This module cannot be used in restricted runtime. *xxx* indicates the module name. You are advised to delete the module.|
38| module xxx is in blocklist, loading prohibited. | The module cannot be used in the current extension. *xxx* indicates the module name. You are advised to delete the module.|
39
40## What should I do when an unexpected value is returned by an API and "occur exception need return" is reported?
41
42Before the call is complete, some Node-API interfaces are checked for JavaScript (JS) exceptions in the VM. If an exception is detected, "occur exception need return" will be reported, with the line number of the code and the Node-API interface name.
43
44You can solve this problem as follows:
45
46- If the exception does not matter, clear the exception.
47  Call **napi_get_and_clear_last_exception** before "occur exception need return" is printed to clear the exception.
48
49- Throw the exception to the ArkTS layer for capture.
50  Throw the exception directly to the ArkTS layer without going through the native logic.
51
52## What are the differences between the lifecycle of napi_value and napi_ref?
53
54- **native_value** is managed by **HandleScope**. Generally, you do not need to add **HandleScope** for **native_value** (except for **complete callback** of **uv_queue_work**).
55
56- **napi_ref** must be deleted manually.
57
58## How do I locate the fault if the return value of a Node-API interface is not "napi_ok"?
59
60When a Node-API interface is successfully executed, **napi_ok** is returned. If the return value is not **napi_ok**, locate the fault as follows:
61
62- Check the result of the input parameter null check, which is performed first before a Node-API interface is executed. The code is as follows:
63
64  ```cpp
65  CHECK_ENV: null check for env.
66  CHECK_ARG: null check for other input parameters.
67  ```
68
69- Check the result of the input parameter type check, which is performed for certain Node-API interfaces. For example, **napi_get_value_double** is used to obtain a C double value from a JS number, and the type of the JS value passed in must be number. The parameter type check is as follows:
70
71  ```cpp
72  RETURN_STATUS_IF_FALSE(env, nativeValue->TypeOf() == NATIVE_NUMBER, napi_number_expected);
73  ```
74
75- Check the return value, which contains the verification result of certain interfaces. For example, **napi_call_function** is used to execute a JS function. If an exception occurs in the JS function, Node-API returns **napi_pending_exception**.
76
77  ```cpp
78  auto resultValue = engine->CallFunction(nativeRecv, nativeFunc, nativeArgv, argc);
79  RETURN_STATUS_IF_FALSE(env, resultValue != nullptr, napi_pending_exception)
80  ```
81
82- Determine the status value returned, and analyze the situation in which the status value is returned.
83
84## What should I do if memory leaks when napi_threadsafe_function is used?
85
86When **napi_threadsafe_function** (**tsfn** for short) is used, **napi_acquire_threadsafe_function** is often called to change the reference count of **tsfn** to ensure that **tsfn** is not released unexpectedly. When all the **tsfn** calls are complete, **napi_release_threadsafe_function** should be called in **napi_tsfn_release** mode in a timely manner to ensure that the reference count returns to the value before **napi_acquire_threadsafe_function** is called. **tsfn** can be correctly released only when the reference count is **0**.
87
88When **env** is about to exit but the reference count of **tsfn** is not **0**, **napi_release_threadsafe_function** should be called in **napi_tsfn_abort** mode to ensure that **tsfn** is not held or used by **env** after **env** is released. If **env** continues to hold and use **tsfn** after exiting, the application may crash.
89
90The following code shows how to register **env_cleanup** to ensure that **tsfn** is no longer held by **env** after **env** exits.
91
92```cpp
93#include <hilog/log.h> // To output logs, link libhilog_ndk.z.so.
94#include <thread> // Include the thread module to create and manage threads.
95#include <unistd.h> // Include unistd.h to suspend the execution of the calling thread.
96
97// Define the log domain and tag.
98#undef LOG_DOMAIN
99#undef LOG_TAG
100#define LOG_DOMAIN 0x2342
101#define LOG_TAG "MY_TSFN_DEMO"
102
103/*
104  To construct a scenario in which the env lifecycle is shorter than the native lifecycle,
105  the following uses worker, taskpool, and napi_create_ark_runtime to create an ArkTS
106  running environment for a worker thread and manually stop the thread in advance.
107*/
108
109
110// Define a struct to simulate the scenario where tsfn is stored.
111class MyTsfnContext {
112public:
113// MyTsfnContext is constructed only in a JS thread because Node-API is used.
114MyTsfnContext(napi_env env, napi_value workName) {
115    // Register the env_cleanup_hook function.
116    napi_add_env_cleanup_hook(env, Cleanup, this);
117    // Create a thread-safe function.
118    if (napi_create_threadsafe_function(env, nullptr, nullptr, workName, 1, 1, this,
119            TsfnFinalize, this, TsfnCallJs, &tsfn_) != napi_ok) {
120        OH_LOG_INFO(LOG_APP, "tsfn is created faild");
121        return;
122    };
123};
124
125~MyTsfnContext() { OH_LOG_INFO(LOG_APP, "MyTsfnContext is deconstructed"); };
126
127napi_threadsafe_function GetTsfn() {
128    std::unique_lock<std::mutex> lock(mutex_);
129    return tsfn_;
130}
131
132bool Acquire() {
133    if (GetTsfn() == nullptr) {
134        return false;
135    };
136    return (napi_acquire_threadsafe_function(GetTsfn()) == napi_ok);
137};
138
139bool Release() {
140    if (GetTsfn() == nullptr) {
141        return false;
142    };
143    return (napi_release_threadsafe_function(GetTsfn(), napi_tsfn_release) == napi_ok);
144};
145
146bool Call(void *data) {
147    if (GetTsfn() == nullptr) {
148        return false;
149    };
150    return (napi_call_threadsafe_function(GetTsfn(), data, napi_tsfn_blocking) == napi_ok);
151};
152
153private:
154// Ensure correct read and write of tsfn by multiple threads.
155std::mutex mutex_;
156napi_threadsafe_function tsfn_ = nullptr;
157
158// Call napi_add_env_cleanup_hook.
159static void Cleanup(void *data) {
160    MyTsfnContext *that = reinterpret_cast<MyTsfnContext *>(data);
161    napi_threadsafe_function tsfn = that->GetTsfn();
162    std::unique_lock<std::mutex> lock(that->mutex_);
163    that->tsfn_ = nullptr;
164    lock.unlock();
165    OH_LOG_WARN(LOG_APP, "cleanup is called");
166    napi_release_threadsafe_function(tsfn, napi_tsfn_abort);
167};
168
169// Callback to be invoked when tsfn is released.
170static void TsfnFinalize(napi_env env, void *data, void *hint) {
171    MyTsfnContext *ctx = reinterpret_cast<MyTsfnContext *>(data);
172    OH_LOG_INFO(LOG_APP, "tsfn is released");
173    napi_remove_env_cleanup_hook(env, MyTsfnContext::Cleanup, ctx);
174    // Cleanup releases the thread-safe function in advance. To avoid UAF, enable the caller to trigger the release.
175    if (ctx->GetTsfn() != nullptr) {
176        OH_LOG_INFO(LOG_APP, "ctx is released");
177        delete ctx;
178    }
179};
180
181// Callback sent by tsfn to the JS thread for execution.
182static void TsfnCallJs(napi_env env, napi_value func, void *context, void *data) {
183    MyTsfnContext *ctx = reinterpret_cast<MyTsfnContext *>(context);
184    char *str = reinterpret_cast<char *>(data);
185    OH_LOG_INFO(LOG_APP, "tsfn is called, data is: \"%{public}s\"", str);
186    // The service logic is omitted here.
187};
188};
189
190// Register the myTsfnDemo method with the module. The myTsfnDemo method is defined as follows:
191// export const myTsfnDemo: () => void;
192napi_value MyTsfnDemo(napi_env env, napi_callback_info info) {
193    OH_LOG_ERROR(LOG_APP, "MyTsfnDemo is called");
194    napi_value workName = nullptr;
195    napi_create_string_utf8(env, "MyTsfnWork", NAPI_AUTO_LENGTH, &workName);
196    MyTsfnContext *myContext = new MyTsfnContext(env, workName);
197    if (myContext->GetTsfn() == nullptr) {
198        OH_LOG_ERROR(LOG_APP, "faild to create tsfn");
199        delete myContext;
200        return nullptr;
201    };
202    char *data0 = new char[]{"Im call in ArkTS Thread"};
203    if (!myContext->Call(data0)) {
204        OH_LOG_INFO(LOG_APP, "call tsfn failed");
205    };
206
207    // Create a thread to simulate an asynchronous operation.
208    std::thread(
209        [](MyTsfnContext *myCtx) {
210            if (!myCtx->Acquire()) {
211                OH_LOG_ERROR(LOG_APP, "acquire tsfn faild");
212                return;
213            };
214            char *data1 = new char[]{"Im call in std::thread"};
215            // This operation is optional and used only to check whether the asynchronous tsfn is still valid.
216            if (!myCtx->Call(data1)) {
217                OH_LOG_ERROR(LOG_APP, "call tsfn failed");
218            };
219            // Suspend the thread for 5 seconds to simulate a time-consuming operation, which is not complete when env exits.
220            sleep(5);
221            // When the asynchronous operation is complete, tsfn has been released and set to nullptr.
222            char *data2 = new char[]{"Im call after work"};
223            if (!myCtx->Call(data2) && !myCtx->Release()) {
224                OH_LOG_ERROR(LOG_APP, "call and release tsfn failed");
225                delete myCtx;
226            }
227        },
228        myContext)
229        .detach();
230    return nullptr;
231};
232```
233
234The following is the main thread logic, which creates worker threads and instruct workers to execute tasks.
235
236```ts
237// Main thread.
238import worker, { MessageEvents } from '@ohos.worker';
239
240const mWorker = new worker.ThreadWorker('../workers/Worker');
241mWorker.onmessage = (e: MessageEvents) => {
242    const action: string | undefined = e.data?.action;
243    if (action === 'kill') {
244        mWorker.terminate();
245    }
246}
247
248// The registration of the triggering mode is omitted.
249mWorker.postMessage({action: 'tsfn-demo'})
250
251```
252
253The following is the worker thread logic, which triggers native tasks.
254
255```ts
256// worker.ets
257import worker, { ThreadWorkerGlobalScope, MessageEvents, ErrorEvent } from '@ohos.worker';
258import napiModule from 'libentry.so'; // libentry.so is the module name of the Node-API library.
259
260const workerPort: ThreadWorkerGlobalScope = worker.workerPort;
261
262workerPort.onmessage = (e: MessageEvents) => {
263    const action: string | undefined = e.data?.action;
264    if (action === 'tsfn-demo') {
265        // Trigger the tsfn demo in C++.
266        napiModule.myTsfnDemo();
267        // Instruct the main thread to terminate the worker.
268        workerPort.postMessage({action: 'kill'});
269    };
270}
271```
272
273## napi_get_uv_event_loop Error Codes
274
275Additional parameter verification is added to prevent use of invalid **napi_env** in **napi_get_uv_event_loop**. The return value indicates the verification result. The return values of this API are as follows:
276
2771. If **env** and/or **loop** are **nullptr**, **napi_invalid_arg** is returned.
2782. If **env** is a valid **napi_env** and **loop** is a valid pointer, **napi_ok** is returned.
2793. If **env** is not a valid **napi_env** (for example, a released **env**), **napi_generic_failure** is returned.
280
281Example:
282
283```c++
284napi_value NapiInvalidArg(napi_env env, napi_callback_info)
285{
286    napi_status status = napi_ok;
287    status = napi_get_uv_event_loop(env, nullptr); // loop is nullptr, napi_invalid_arg.
288    if (status == napi_ok) {
289        // do something
290    }
291
292    uv_loop_s* loop = nullptr;
293    status = napi_get_uv_event_loop(nullptr, &loop); // env is nullptr, napi_invalid_arg.
294    if (status == napi_ok) {
295        // do something
296    }
297
298    status = napi_get_uv_event_loop(nullptr, nullptr); // Both env and loop are nullptr, napi_invalid_arg.
299    if (status == napi_ok) {
300        // do something
301    }
302
303    return nullptr;
304}
305
306napi_value NapiGenericFailure(napi_env env, napi_callback_info)
307{
308    std::thread([]() {
309        napi_env env = nullptr;
310        napi_create_ark_runtime (&env); // Generally, the return value needs to be checked.
311        // napi_destroy_ark_runtime sets the pointer to null. Copy the pointer to simulate the problem.
312        napi_env copiedEnv = env;
313        napi_destroy_ark_runtime(&env);
314        uv_loop_s* loop = nullptr;
315        napi_status status = napi_get_uv_event_loop(copiedEnv, &loop); // env is invalid. napi_generic_failure will be returned.
316        if (status == napi_ok) {
317            // do something
318        }
319    }).detach();;
320}
321```
322