src: preload function for Environment

This PR adds a |preload| arg to the node::CreateEnvironment to allow
embedders to set a preload function for the environment, which will run
after the environment is loaded and before the main script runs.

This is similiar to the --require CLI option, but runs a C++ function,
and can only be set by embedders.

The preload function can be used by embedders to inject scripts before
running the main script, for example:
1. In Electron it is used to initialize the ASAR virtual filesystem,
   inject custom process properties, etc.
2. In VS Code it can be used to reset the module search paths for
   extensions.

lib/internal/process/pre_execution.js

@@ -186,6 +186,9 @@ function setupUserModules(forceDefaultLoader = false) {
   initializeESMLoader(forceDefaultLoader);
   const CJSLoader = require('internal/modules/cjs/loader');
   assert(!CJSLoader.hasLoadedAnyUserCJSModule);
+  if (getEmbedderOptions().hasEmbedderPreload) {
+    runEmbedderPreload();
+  }
   // Do not enable preload modules if custom loaders are disabled.
   // For example, loader workers are responsible for doing this themselves.
   // And preload modules are not supported in ShadowRealm as well.
@@ -754,6 +757,10 @@ function initializeFrozenIntrinsics() {
   }
 }
 
+function runEmbedderPreload() {
+  internalBinding('mksnapshot').runEmbedderPreload(process, require);
+}
+
 function loadPreloadModules() {
   // For user code, we preload modules if `-r` is passed
   const preloadModules = getOptionValue('--require');

src/api/environment.cc

@@ -431,7 +431,8 @@ Environment* CreateEnvironment(
     const std::vector<std::string>& exec_args,
     EnvironmentFlags::Flags flags,
     ThreadId thread_id,
-    std::unique_ptr<InspectorParentHandle> inspector_parent_handle) {
+    std::unique_ptr<InspectorParentHandle> inspector_parent_handle,
+    EmbedderPreloadCallback preload) {
   Isolate* isolate = isolate_data->isolate();
 
   Isolate::Scope isolate_scope(isolate);
@@ -452,7 +453,8 @@ Environment* CreateEnvironment(
                                      exec_args,
                                      env_snapshot_info,
                                      flags,
-                                     thread_id);
+                                     thread_id,
+                                     std::move(preload));
   CHECK_NOT_NULL(env);
 
   if (use_snapshot) {

src/env-inl.h

@@ -438,6 +438,10 @@ inline void Environment::set_embedder_entry_point(StartExecutionCallback&& fn) {
   embedder_entry_point_ = std::move(fn);
 }
 
+inline const EmbedderPreloadCallback& Environment::embedder_preload() const {
+  return embedder_preload_;
+}
+
 inline double Environment::new_async_id() {
   async_hooks()->async_id_fields()[AsyncHooks::kAsyncIdCounter] += 1;
   return async_hooks()->async_id_fields()[AsyncHooks::kAsyncIdCounter];

src/env.cc

@@ -776,7 +776,8 @@ Environment::Environment(IsolateData* isolate_data,
                          const std::vector<std::string>& exec_args,
                          const EnvSerializeInfo* env_info,
                          EnvironmentFlags::Flags flags,
-                         ThreadId thread_id)
+                         ThreadId thread_id,
+                         EmbedderPreloadCallback preload)
     : isolate_(isolate),
       isolate_data_(isolate_data),
       async_hooks_(isolate, MAYBE_FIELD_PTR(env_info, async_hooks)),
@@ -802,7 +803,8 @@ Environment::Environment(IsolateData* isolate_data,
       flags_(flags),
       thread_id_(thread_id.id == static_cast<uint64_t>(-1)
                      ? AllocateEnvironmentThreadId().id
-                     : thread_id.id) {
+                     : thread_id.id),
+      embedder_preload_(std::move(preload)) {
   constexpr bool is_shared_ro_heap =
 #ifdef NODE_V8_SHARED_RO_HEAP
       true;

src/env.h

@@ -649,7 +649,8 @@ class Environment : public MemoryRetainer {
               const std::vector<std::string>& exec_args,
               const EnvSerializeInfo* env_info,
               EnvironmentFlags::Flags flags,
-              ThreadId thread_id);
+              ThreadId thread_id,
+              EmbedderPreloadCallback preload);
   void InitializeMainContext(v8::Local<v8::Context> context,
                              const EnvSerializeInfo* env_info);
   ~Environment() override;
@@ -1002,6 +1003,8 @@ class Environment : public MemoryRetainer {
   inline const StartExecutionCallback& embedder_entry_point() const;
   inline void set_embedder_entry_point(StartExecutionCallback&& fn);
 
+  inline const EmbedderPreloadCallback& embedder_preload() const;
+
   inline void set_process_exit_handler(
       std::function<void(Environment*, ExitCode)>&& handler);
 
@@ -1208,6 +1211,7 @@ class Environment : public MemoryRetainer {
 
   builtins::BuiltinLoader builtin_loader_;
   StartExecutionCallback embedder_entry_point_;
+  EmbedderPreloadCallback embedder_preload_;
 
   // Used by allocate_managed_buffer() and release_managed_buffer() to keep
   // track of the BackingStore for a given pointer.

src/node.h

@@ -691,19 +691,32 @@ struct InspectorParentHandle {
   virtual ~InspectorParentHandle() = default;
 };
 
+using EmbedderPreloadCallback =
+    std::function<void(Environment* env,
+                       v8::Local<v8::Value> process,
+                       v8::Local<v8::Value> require)>;
+
 // TODO(addaleax): Maybe move per-Environment options parsing here.
 // Returns nullptr when the Environment cannot be created e.g. there are
 // pending JavaScript exceptions.
 // `context` may be empty if an `EmbedderSnapshotData` instance was provided
 // to `NewIsolate()` and `CreateIsolateData()`.
+//
+// The |preload| function will run before executing the entry point, which
+// is usually used by embedders to inject scripts. The function is executed
+// with preload(process, require), and the passed require function has access
+// to internal Node.js modules. The |preload| function is inherited by worker
+// threads and thus will run in work threads, so make sure the function is
+// thread-safe.
 NODE_EXTERN Environment* CreateEnvironment(
     IsolateData* isolate_data,
     v8::Local<v8::Context> context,
     const std::vector<std::string>& args,
     const std::vector<std::string>& exec_args,
     EnvironmentFlags::Flags flags = EnvironmentFlags::kDefaultFlags,
     ThreadId thread_id = {} /* allocates a thread id automatically */,
-    std::unique_ptr<InspectorParentHandle> inspector_parent_handle = {});
+    std::unique_ptr<InspectorParentHandle> inspector_parent_handle = {},
+    EmbedderPreloadCallback preload = nullptr);
 
 // Returns a handle that can be passed to `LoadEnvironment()`, making the
 // child Environment accessible to the inspector as if it were a Node.js Worker.

src/node_options.cc

@@ -1304,6 +1304,12 @@ void GetEmbedderOptions(const FunctionCallbackInfo<Value>& args) {
           .IsNothing())
     return;
 
+  if (ret->Set(context,
+               FIXED_ONE_BYTE_STRING(env->isolate(), "hasEmbedderPreload"),
+               Boolean::New(isolate, env->embedder_preload() != nullptr))
+          .IsNothing())
+    return;
+
   args.GetReturnValue().Set(ret);
 }
 

src/node_snapshotable.cc

@@ -1453,6 +1453,13 @@ static void RunEmbedderEntryPoint(const FunctionCallbackInfo<Value>& args) {
   }
 }
 
+static void RunEmbedderPreload(const FunctionCallbackInfo<Value>& args) {
+  Environment* env = Environment::GetCurrent(args);
+  CHECK(env->embedder_preload());
+  CHECK_EQ(args.Length(), 2);
+  env->embedder_preload()(env, args[0], args[1]);
+}
+
 void CompileSerializeMain(const FunctionCallbackInfo<Value>& args) {
   CHECK(args[0]->IsString());
   Local<String> filename = args[0].As<String>();
@@ -1577,6 +1584,7 @@ void CreatePerIsolateProperties(IsolateData* isolate_data,
                                 Local<ObjectTemplate> target) {
   Isolate* isolate = isolate_data->isolate();
   SetMethod(isolate, target, "runEmbedderEntryPoint", RunEmbedderEntryPoint);
+  SetMethod(isolate, target, "runEmbedderPreload", RunEmbedderPreload);
   SetMethod(isolate, target, "compileSerializeMain", CompileSerializeMain);
   SetMethod(isolate, target, "setSerializeCallback", SetSerializeCallback);
   SetMethod(isolate, target, "setDeserializeCallback", SetDeserializeCallback);
@@ -1590,6 +1598,7 @@ void CreatePerIsolateProperties(IsolateData* isolate_data,
 
 void RegisterExternalReferences(ExternalReferenceRegistry* registry) {
   registry->Register(RunEmbedderEntryPoint);
+  registry->Register(RunEmbedderPreload);
   registry->Register(CompileSerializeMain);
   registry->Register(SetSerializeCallback);
   registry->Register(SetDeserializeCallback);

src/node_worker.cc

@@ -63,6 +63,7 @@ Worker::Worker(Environment* env,
       thread_id_(AllocateEnvironmentThreadId()),
       name_(name),
       env_vars_(env_vars),
+      embedder_preload_(env->embedder_preload()),
       snapshot_data_(snapshot_data) {
   Debug(this, "Creating new worker instance with thread id %llu",
         thread_id_.id);
@@ -366,7 +367,8 @@ void Worker::Run() {
             std::move(exec_argv_),
             static_cast<EnvironmentFlags::Flags>(environment_flags_),
             thread_id_,
-            std::move(inspector_parent_handle_)));
+            std::move(inspector_parent_handle_),
+            std::move(embedder_preload_)));
         if (is_stopped()) return;
         CHECK_NOT_NULL(env_);
         env_->set_env_vars(std::move(env_vars_));

src/node_worker.h

@@ -114,6 +114,7 @@ class Worker : public AsyncWrap {
 
   std::unique_ptr<MessagePortData> child_port_data_;
   std::shared_ptr<KVStore> env_vars_;
+  EmbedderPreloadCallback embedder_preload_;
 
   // A raw flag that is used by creator and worker threads to
   // sync up on pre-mature termination of worker  - while in the

test/cctest/test_environment.cc

@@ -778,3 +778,31 @@ TEST_F(EnvironmentTest, RequestInterruptAtExit) {
 
   context->Exit();
 }
+
+TEST_F(EnvironmentTest, EmbedderPreload) {
+  v8::HandleScope handle_scope(isolate_);
+  v8::Local<v8::Context> context = node::NewContext(isolate_);
+  v8::Context::Scope context_scope(context);
+
+  node::EmbedderPreloadCallback preload = [](node::Environment* env,
+                                             v8::Local<v8::Value> process,
+                                             v8::Local<v8::Value> require) {
+    CHECK(process->IsObject());
+    CHECK(require->IsFunction());
+    process.As<v8::Object>()->Set(
+        env->context(),
+        v8::String::NewFromUtf8Literal(env->isolate(), "prop"),
+        v8::String::NewFromUtf8Literal(env->isolate(), "preload")).Check();
+  };
+
+  std::unique_ptr<node::Environment, decltype(&node::FreeEnvironment)> env(
+      node::CreateEnvironment(isolate_data_, context, {}, {},
+                              node::EnvironmentFlags::kDefaultFlags, {}, {},
+                              preload),
+      node::FreeEnvironment);
+
+  v8::Local<v8::Value> main_ret =
+      node::LoadEnvironment(env.get(), "return process.prop;").ToLocalChecked();
+  node::Utf8Value main_ret_str(isolate_, main_ret);
+  EXPECT_EQ(std::string(*main_ret_str), "preload");
+}